
Sign up to save your podcasts
Or


Mike Nicoletti, Hunt Lawrence, and Jason Wallace walk this week’s Cash Flow Memo: oil stuck in triple digits, a 10-year Treasury headed toward 6%, SpaceX’s $40 billion Nvidia build, the agent threat to Apple’s App Store, and what a trillion-dollar valuation buys at Eli Lilly.
The Cashflow Memo
Key Takeaways
* SpaceX is borrowing $40 billion with Nvidia credit support for what is close to a turnkey rack build, and Mike puts SpaceX’s Nvidia spend at $180 to $280 billion next calendar year against a Street number near $400 billion for Nvidia’s fiscal 2027, so one customer could equal half of the prior year’s revenue.
* Brent sits at $100 and WTI in the low $90s with Hunt calling the Iran standoff an indefinite stalemate, and the 12-month backwardation has narrowed from about $20 to $10 to $12 as the 2027 strip rises; natural gas is holding in the $3s despite Permian associated gas driving two-thirds of the supply increase.
* Hunt expects the $1.55 trillion deficit shown in Exhibit A to land closer to $1.9 to $2 trillion and, with the 10-year Treasury at 5.5%, predicts 6% by the first half of December, a direct hit to mortgage rates and to Lennar, Home Depot, and Lowe’s.
* Jason’s Apple bear case is that Meta’s Muse agent moves transactions off Apple hardware: a two-core cloud instance lists at 4 cents an hour and likely costs Meta about $1 per user per month, so $7 of transaction volume at Apple’s 15% fee covers it, which threatens App Store services growth while Apple pushes into low-margin home hardware with LG.
* Lilly trades at a $1 trillion equity value on $23 billion of free cash flow (Exxon generates $50 billion, Apple $130 billion), which Jason defends on follow-on GLP-1 patents and out-of-cycle pipeline acquisitions; he expects Grail’s Galleri test to win a limited-cancer approval after a 7 to 2 advisory vote on risk-benefit and a split vote on effectiveness.
Show Notes
[00:33] Oil Exhibit C is unchanged. Brent is at $100 and WTI in the low $90s, and Hunt sees the Iran standoff as a stalemate with no end date.
[01:42] Backwardation The 2027 oil price keeps rising, cutting the 12-month backwardation from about $20 to $10 to $12. The market is pricing the stalemate as lasting.
[02:13] Gas Natural gas is holding in the $3s as new LNG trains come on, but high oil prices keep pushing associated gas out of the Permian.
[02:55] Deficit Exhibit A needs a rework, with the deficit likely closer to $2 trillion. Hunt predicts a 6% 10-year Treasury by the first half of December and flags the hit to homebuilders and home improvement retailers.
[05:42] SpaceX’s $40 Billion SpaceX is borrowing $40 billion with Nvidia credit support for a near-turnkey rack build. Mike sizes SpaceX’s Nvidia spend at $180 to $280 billion next calendar year.
[08:02] Nvidia’s Pace Why SpaceX builds on Nvidia alone: a yearly generation leap that no one else matches, plus the cooling, power, and architecture that let Colossus run 100,000 coherent GPUs.
[10:33] Apple Jason was an Apple bull for about a day. The home hardware push with LG is a low-margin bet in a crowded market, and the company needs a software leader.
[12:32] Agents vs. the App Store Meta’s Muse gives every user a cloud server for roughly $1 a month. Transactions that run there never touch Apple hardware or Apple’s 15% fee.
[17:11] Lilly at a Trillion Hunt lines up the 12 trillion-dollar tech companies against Exxon, Lilly, JP Morgan, and Walmart. Jason makes the case for Lilly’s follow-on GLP-1 patents and pipeline acquisitions.
[20:33] Grail’s FDA Vote The advisory panel split on Galleri’s effectiveness and voted 7 to 2 that the benefit outweighs the risk. Jason expects approval for a limited set of cancers.
[23:00] Hospital Price Letters The FTC sent warning letters to a couple dozen large hospital groups over price transparency rules that have gone unenforced for years.
[23:57] After the Midterms What a Democratic House and Senate would mean for the administration’s healthcare agenda, and why food supply reform and price transparency look bipartisan.
[26:44] Food Additives How the generally-recognized-as-safe system lets manufacturers self-certify new ingredients, and how a 1997 change made notifying the FDA voluntary.
Download this week’s Cash Flow Memo at telltales.us and subscribe. More Apple and more on food additives next week.
Cashtags
$AAPL $AMD $AMZN $GOOGL $GRAL $HD $ILMN $INTC $JPM $LEN $LLY $LOW $META $MU $NVDA $TSLA $VRTX $WMT $XOM
This post and the information herein are intended for informational purposes only. The views expressed herein are the author’s alone and do not constitute an offer to sell, or a recommendation to purchase, or a solicitation of an offer to buy, any security, nor a recommendation for any investment product or service. While certain information contained herein has been obtained from sources believed to be reliable, neither the author nor any of his employers or their affiliates have independently verified this information, and its accuracy and completeness cannot be guaranteed. Accordingly, no representation or warranty, express or implied, is made as to, and no reliance should be placed on, the fairness, accuracy, timeliness or completeness of this information. The author and all employers and their affiliated persons assume no liability for this information and no obligation to update the information or analysis contained herein in the future.
▶ Explore this week’s Tape — live, sortable, drill-down →
The Tape — W2640
Universe of 94 cashflow-memo names, snap dates 2026-09-25 → 2026-10-02. Composite is rank-sum percentile of FCF Yield + NTM Revenue Growth (higher = better balance). Banks and finance-book names shown separately.
Telltales Yield — Top 10
Telltales Yield — Bottom 10
This Week’s Reporters
No universe names reporting in the coming 7 days.
Sector Medians
Debt / FCF Watch (highest leverage on TTM FCF)
Weekly Price Movement
Top 5 (week-over-week price)
Bottom 5 (week-over-week price)
Banks (shown separately — FCF metric not meaningful)
Finance-book — FCF not comparable
Customer-float / captive-finance / reserve businesses (IBKR broker float, KMX CarMax Auto Finance, PYPL customer funds, CRCL stablecoin reserves). The memo’s operating-FCF method overstates their FCF, so they are held off the ranked leaderboard pending the P&L-waterfall rebuild.
Data Gaps
91 of 92 ranked-eligible names ranked. 1 dropped for missing FCF yield or NTM revenue growth; 7 shown separately (banks + finance-book, FCF not comparable).
Source: cashflow-memo master_2026-10-02.csv. NTM growth from analyst-estimates consensus. Composite is a percentile rank, not a recommendation.
The Issue — This Week's Brief
Reports on Broadcom’s financing.
The Cashflow Memo
Micron’s Fourth Quarter and Broadcom’s Financing Commitments
Micron’s fiscal fourth quarter, the financing commitments in Broadcom’s 10-Q, and NextEra’s Texas power project
The Telltales Weekend Update. Ava Cabot reports what the companies in the Cash Flow Memo disclosed this week and what is scheduled next. About 9 minutes.
Download the memo at telltales.us. Hunt, Jason, and Mike are back Wednesday on episode E2641.
Chapter markers
* Time | Segment
* 0:00 | Intro: what the episode covers
* 0:40 | Theme: AI infrastructure financing (Broadcom, NextEra)
* 3:20 | Deep dive: Micron’s fiscal fourth quarter
* 5:50 | Rapid-fire: Regeneron, UnitedHealth, Interactive Brokers, the earnings calendar
* 8:00 | Close
* 8:20 | Disclaimer
Full transcript
Intro
Ava: Welcome to the Telltales Weekend Update. I’m Ava Cabot.
Ava: This week, from the companies in the Cash Flow Memo: reports on Broadcom’s financing for Anthropic, NextEra’s Texas power project, Micron’s fiscal fourth quarter, and the Sanofi and Regeneron expansion.
Ava: The show is produced entirely with AI tools, and the voice you’re hearing is AI-generated. The following conversation is intended for informational purposes only. You should always do your own work to determine if an investment is suitable for you.
Theme: AI infrastructure financing
Ava: Reuters reported on Thursday, citing Anthropic’s confidential IPO filing, that Broadcom agreed to lend Anthropic up to $42 billion in convertible notes[^reuters-avgo-anthropic-20261001]. Per Reuters, that is about a third of a $125 billion, five-year commitment by Anthropic to lease TPU computing capacity[^reuters-avgo-anthropic-20261001].
Ava: Broadcom’s 10-Q, filed September 10, describes a backstop and convertible notes, each tied to customer lease agreements[^10q-avgo-note10-20260910]. It does not name a customer. The backstop was given to a financial partner, on a customer’s five-year lease obligations. If the customer defaults on a lease, Broadcom owes the difference between 85% of the outstanding lease amounts and the value received from selling the AI racks[^10q-avgo-note10-20260910]. The 10-Q states the limit this way.
Broadcom 10-Q (quoted): Our maximum potential liability under the Backstop upon the deployment of all AI racks, on an undiscounted basis, was approximately $29 billion.[^10q-avgo-note10-20260910]
Ava: The 10-Q says no amounts have been paid under the backstop[^10q-avgo-note10-20260910]. The 10-Q also describes convertible notes. Under certain circumstances, and if needed, a customer may issue Broadcom up to $42 billion of them, to be used solely for the customer’s obligations under the lease agreements. As of August 2, none had been issued[^10q-avgo-note10-20260910].
Ava: On Friday, Bloomberg reported that Blackstone and a group of banks are assembling $60 billion of financing for AI chips[^bbg-avgo-syndicate-20261002]. Bloomberg reported that it includes a $42 billion senior secured tranche, that Broadcom will provide residual value support on that tranche, and that the deal has not been announced[^bbg-avgo-syndicate-20261002].
Ava: Reuters also reported this week, from the same Anthropic filing, that Anthropic expects to spend at least $518 billion on infrastructure over the next decade, across six partners, with about 80% of it non-cancelable or payable whether or not the capacity is used[^reuters-anthropic-518-20260929].
Ava: The Cash Flow Memo dated September 28 has Broadcom at 51 times free cash flow, on $34 billion of free cash flow[^memo-avgo-20260928].
Ava: Next, NextEra. On Wednesday, NextEra Energy, Lewis Energy Group and Related Digital said the U.S. Department of Commerce and the Republic of Korea selected Related Companies and NextEra Energy Resources, in partnership with Lewis Energy Group, to develop Project Star in Encinal, Texas[^pr-nee-projectstar-20260930]. Per the companies’ release, it is a $22 billion natural gas power campus with about 6.5 gigawatts of generation, built to support a 5-gigawatt data center campus on an adjacent site[^pr-nee-projectstar-20260930].
Ava: The release says the campus will be owned jointly by the Republic of Korea and the United States, and built and operated by the Related and NextEra Energy Resources joint venture[^pr-nee-projectstar-20260930]. Initial generation is expected as early as 2029, subject to permitting and approvals[^pr-nee-projectstar-20260930].
Deep dive: Micron
Ava: Micron reported its fiscal fourth quarter on Wednesday. Per the 8-K, revenue was $54 billion, against $41 billion in the prior quarter and $11 billion a year earlier[^8k-mu-ex991-20260930]. Net income was about $38 billion[^8k-mu-ex991-20260930].
Ava: For the fiscal year, revenue was $133 billion, against $37 billion the year before[^8k-mu-ex991-20260930]. Adjusted free cash flow for the year was $62 billion, after $27 billion of capital expenditures, net of government incentives[^8k-mu-ex991-20260930].
Ava: Micron’s guidance for the current quarter is revenue of $61.5 billion, plus or minus $1.5 billion, at a gross margin of about 86%[^8k-mu-ex991-20260930].
Ava: In its prepared remarks for the earnings call, Micron said fiscal fourth-quarter DRAM revenue was about $40 billion, 73% of total revenue[^mu-remarks-20260930].
Ava: Micron said it has signed 26 strategic customer agreements, which it estimates at more than 35% of its revenue through 2030[^mu-remarks-20260930]. Three-quarters of that estimated revenue has a defined pricing framework, and the remaining quarter is priced periodically, based on market prices[^mu-remarks-20260930].
Ava: Micron put its remaining performance obligations at about $150 billion[^mu-remarks-20260930]. Customer cash deposits received in the quarter were about $12 billion[^mu-remarks-20260930].
Ava: For calendar 2027 and 2028, Micron expects industry DRAM bit shipments to grow in the low-20s percent range and NAND in the mid-20s, and expects the industry to remain supply constrained in both years[^mu-remarks-20260930]. Chief executive Sanjay Mehrotra, in the prepared remarks, on DRAM supply.
Sanjay Mehrotra (quoted): Even with additional industry DRAM cleanroom space plans, with robust demand trends including new upside requests from customers, we do not have line of sight to when supply and demand will return to balance.[^mu-remarks-20260930]
Ava: Micron said it plans to increase capital expenditures in fiscal 2027 versus its prior plans[^mu-remarks-20260930].
Ava: The Cash Flow Memo dated September 28 has Micron at 28 times free cash flow, on $43 billion of free cash flow, figures that predate this report[^memo-mu-20260928].
Rapid-fire
Ava: Page 19 of the memo. Sanofi and Regeneron announced on Thursday an expansion of their antibody collaboration[^pr-regn-sanofi-20261001]. Per the release, Regeneron receives $1 billion upfront, with up to $7 billion more in milestone payments[^pr-regn-sanofi-20261001].
Ava: The companies will co-develop and co-commercialize four new long-acting antibodies, and share costs and any profits equally[^pr-regn-sanofi-20261001]. Regeneron also has an option to include Sanofi’s lunsekimig, exercisable when its phase 3 studies in chronic obstructive pulmonary disease are complete[^pr-regn-sanofi-20261001].
Ava: Also on page 19, UnitedHealth. Bloomberg reported on Thursday that UnitedHealth is dropping 2027 Medicare Advantage plans covering about 390,000 members[^bbg-unh-maplans-20261001]. UnitedHealth reports earnings on Tuesday, October 13[^earn-unh].
Ava: Interactive Brokers released its September metrics on Thursday. Per the release, client accounts were 5.6 million, 35% higher than a year earlier, and daily average revenue trades were 4.1 million, 6% higher[^pr-ibkr-metrics-20261001]. Interactive Brokers reports earnings on Thursday, October 15[^earn-ibkr].
Ava: On page 13 of the memo, the banks. The SEC on Friday censured two JPMorgan units in a settled order[^sec-jpm-order-20261002]. The order finds that they permitted a person subject to a statutory disqualification to be involved in security-based swap transactions, and failed to supervise that person[^sec-jpm-order-20261002]. The sanctions are a censure and a cease-and-desist order[^sec-jpm-order-20261002]. CNBC, citing The Wall Street Journal, reported that Goldman Sachs’s board has discussed president John Waldron succeeding David Solomon as chief executive as early as next year[^cnbc-gs-succession-20260929]. CNBC also reported that a Goldman spokesman said there is no definitive timeline for succession[^cnbc-gs-succession-20260929].
Ava: The earnings calendar. Goldman Sachs[^earn-gs] and JPMorgan[^earn-jpm] report on Tuesday, October 13. Morgan Stanley[^earn-ms], ASML[^earn-asml] and Fastenal[^earn-fast] report on Wednesday, October 14. Taiwan Semiconductor reports on Thursday, October 15[^earn-tsm].
Close
Ava: That’s the Weekend Update. Hunt, Jason, and Mike are back Wednesday on episode 2641. Download this week’s Cash Flow Memo at telltales.us, and send us a note through the Substack.
Disclaimer
Ava: This podcast is intended for informational purposes only. You should always do your own work to determine if an investment is suitable for you. The views expressed on this podcast are the host alone and do not constitute an offer to sell or a recommendation to purchase, or a solicitation of an offer to buy any security, nor a recommendation for any investment product or service. While certain information contained herein has been obtained from sources believed to be reliable, neither the host nor any of their employers or their affiliates have independently verified this information, and its accuracy and completeness cannot be guaranteed. Accordingly, no representation or warranty, express or implied, is made as to, and no reliance should be placed on, the fairness, accuracy, timeliness, or completeness of this information. The host and all employers and their affiliated persons assume no liability for this information and no obligation to update the information or analysis contained herein in the future, and may or may not hold positions in the securities mentioned.
Sources
* Bloomberg. (2026, October 1). UnitedHealth, Humana Drop Medicare Advantage Plans to Boost Profit. Bloomberg. https://www.bloomberg.com/news/articles/2026-10-01/health-insurers-ditch-medicare-advantage-plans-to-boost-profit
* Bloomberg. (2026, October 2). Broadcom Amassing $60 Billion to Fund Chips for Anthropic. Bloomberg. https://www.bloomberg.com/news/articles/2026-10-02/broadcom-starts-amassing-60-billion-to-fund-chips-for-anthropic
* Broadcom Inc. (2026, September 10). Form 10-Q for the quarterly period ended August 2, 2026 [10-Q]. SEC EDGAR. https://www.sec.gov/Archives/edgar/data/1730168/000173016826000080/avgo-20260802.htm
* CNBC. (2026, September 29). Goldman Sachs CEO succession planning faces one big problem. CNBC. https://www.cnbc.com/2026/09/29/goldman-sachs-ceo-succession-planning.html
* Interactive Brokers Group, Inc. (2026, October 1). Interactive Brokers Group reports brokerage metrics and other financial information for September 2026, includes Reg.-NMS execution statistics [Press release]. Business Wire. https://finance.yahoo.com/markets/stocks/articles/interactive-brokers-group-reports-brokerage-161500811.html
* Micron Technology, Inc. (2026, September 30). Fiscal Q4 2026 earnings call prepared remarks. Micron Investor Relations. https://investors.micron.com/files/doc_financials/2026/q4/Q4-FY26-Prepared-Remarks.pdf
* Micron Technology, Inc. (2026, September 30). Micron Technology, Inc. reports record fiscal fourth-quarter and full-year 2026 results [8-K, Exhibit 99.1]. SEC EDGAR. https://www.sec.gov/Archives/edgar/data/723125/000072312526000018/a2026q4ex991-pressrelease.htm
* NextEra Energy, Inc., Lewis Energy Group, & Related Digital. (2026, September 30). The United States Department of Commerce and the Government of the Republic of Korea announce strategic investment in Project Star, a $22.3 billion energy infrastructure campus in Encinal, Texas [Press release]. https://www.investor.nexteraenergy.com/news-and-events/news-releases/2026/09-30-2026-231920318
* Reuters. (2026, October 1). Broadcom to lend Anthropic up to $42 billion to lease its chips, filing says. CNBC. https://www.cnbc.com/2026/10/01/broadcom-lending-anthropic-42-billion-chips-reuters.html
* Reuters. (2026, September 29). Anthropic’s $518 billion AI buildout hinges largely on deals that cannot be canceled, filing shows. KSL. https://www.ksl.com/article/51629885/anthropics-518-billion-ai-buildout-hinges-largely-on-deals-that-cannot-be-canceled-filing-shows
* Sanofi & Regeneron Pharmaceuticals, Inc. (2026, October 1). Sanofi and Regeneron expand global Alliance with multiple next-generation, long-acting immunology antibodies [Press release]. https://www.sanofi.com/en/media-room/press-releases/2026/2026-10-01-05-00-00-3372517
* U.S. Securities and Exchange Commission. (2026, October 2). In the Matter of J.P. Morgan Securities LLC and JPMorgan Chase Bank, N.A. (Exchange Act Release No. 106569). https://www.sec.gov/files/litigation/admin/2026/34-106569.pdf
Internal data
Internal data is provided on a best efforts basis.
Hunt, Mike, and Jason walk the Cash Flow Memo as Hormuz flows recover, the fiscal year closes near a $2 trillion deficit, and the 10-year pushes toward 6%. Then Hunt’s case that the 12 trillion-dollar tech companies compound at three times nominal GDP, what $70 billion run rates at OpenAI and Anthropic say about agents, and whether AI makes drug development cheaper for everyone or just for Lilly.
The Cashflow Memo
Key Takeaways
* Hormuz flows have recovered to roughly 15M barrels a day (vs ~20M before February) via shuttle tankers and the Saudi and Abu Dhabi bypass pipelines, but the 2027 strip trading up to $77 against $91 spot says the market prices the disruption as durable.
* Hunt expects the fiscal-year deficit closing today near $2T against the $1.55T in Exhibit A, driven by spending (defense above $950B, all other not falling), with the 10-year at ~5.30% possibly reaching 6% by December and 7.5% mortgages the main casualty.
* Hunt’s working thesis on the 12 tech companies above $1T is revenue growth of ~3x nominal GDP (~15%), enough to double in five years; Mike and Jason accept the direction but expect one or two laggards and structurally thinner AI software margins because every user consumes tokens.
* Leaked figures put OpenAI and Anthropic each near a $70B revenue run rate from near zero ~18 months ago, and consumer agents (Meta’s Muse, OpenAI’s Dots, SpaceX’s Grokbot) now follow the Claude Code / Codex business inflection.
* Lilly, the largest non-tech name in the memo at ~50x free cash flow, gains from AI-accelerated R&D, but Mike argues equal access across pharma likely competes those gains away to consumers, leaving scale distribution and proprietary data (Lilly’s GLP-1 base) as the moat.
Show Notes
[00:00] Intro
[00:27] Oil: shuttle tankers through Hormuz Flows are back to roughly 15M barrels a day from 20M before February, via shuttle tankers and Gulf bypass pipelines. US inventories are drawing about 1M barrels a day.
[03:07] Backwardation narrows Spot at $91 against a 2027 strip of $77. The back end trading up says buyers see the disruption as intractable.
[03:47] Gas Permian associated gas keeps a lid on prices.
[04:06] Deficit and the 10-year Exhibit A shows $1.55T; Hunt expects closer to $2T, and the problem is spending. The 10-year near 5.30% could reach 6% by December, with mortgages already at 7.5%.
[09:34] An industrial revolution parallel Sour consumer sentiment in a strong economy has happened before.
[10:16] Three times GDP Hunt’s work on the 12 companies above $1T: 15% growth, three times ~5% nominal GDP, doubles in five years. Mike and Jason expect lumpy growth, a laggard or two, and thinner AI software margins.
[16:33] $70B run rates Leaked figures put OpenAI and Anthropic each near a $70B revenue run rate, from near zero 18 months ago.
[17:13] The consumer agent race Claude Code and Codex were the business inflection; Muse, Dots, and Grokbot are the consumer one. Why the winners may not be invented yet.
[21:20] A PhD hour for a fraction of a penny A benchmark task worth an hour of PhD work cost 30-40 cents in 2025 and costs a fraction of a penny today. Could AI step GDP growth up again?
[23:58] Lilly at 50x free cash flow The largest non-tech name in the memo, ahead of Walmart, JPMorgan, and Exxon. Does AI-driven R&D accrue to Lilly or get competed away across pharma?
[29:13] Moderna, BioNTech, and drug pricing AI for cancer medicines and FDA filings, why biology is harder to simulate than physics, and how much of US drug spend most-favored-nation pricing covers.
Download this week’s Cash Flow Memo at telltales.us and subscribe. More healthcare next week.
Cashtags
$AAPL $AMD $AMZN $BNTX $GOOGL $JPM $LLY $META $MRNA $MSFT $MU $NVDA $TSLA $XOM
This post and the information herein are intended for informational purposes only. The views expressed herein are the author’s alone and do not constitute an offer to sell, or a recommendation to purchase, or a solicitation of an offer to buy, any security, nor a recommendation for any investment product or service. While certain information contained herein has been obtained from sources believed to be reliable, neither the author nor any of his employers or their affiliates have independently verified this information, and its accuracy and completeness cannot be guaranteed. Accordingly, no representation or warranty, express or implied, is made as to, and no reliance should be placed on, the fairness, accuracy, timeliness or completeness of this information. The author and all employers and their affiliated persons assume no liability for this information and no obligation to update the information or analysis contained herein in the future.
▶ Explore this week’s Tape — live, sortable, drill-down →
Berkshire Gave Up the Quiet Exit
Berkshire’s real commitment to Lennar this month was the paperwork, more than the two hundred and twelve million dollars. On Thursday, September seventeenth, Berkshire crossed ten percent ownership of Lennar¹. Past that line the rules of owning a stock change. Every share it buys or sells now goes public within two business days, and any profit on a purchase and sale inside six months gets handed back to Lennar².
That is a strange thing for Berkshire to walk into on purpose. This is a company that asked the SEC to keep its Chubb stake off the public record for two full quarters while it was still buying, and only showed its hand in amended filings in May of twenty-twenty-four³. Berkshire likes to build in the dark.
It built most of Lennar in the dark too. Until that Thursday the only public record was the quarterly 13F, which lands a month and a half after each quarter closes. It shows about seven million Class A shares in mid-twenty-twenty-five, left alone for two quarters, then ten million by March and thirteen million by June⁴. By the morning Berkshire crossed the line, the Form 3 put the count at twenty-one million⁵. Eight million shares went on the books over a summer when nobody outside Omaha could see them.
Then Lennar reported, on Wednesday the sixteenth: the miss, the guide cut⁶. Berkshire bought the next three sessions, crossed ten percent on the first of them, and filed Monday night⁷. The stock popped Tuesday on the disclosure⁸.
The audio stopped there. The filings kept going. Wednesday through Friday, Berkshire bought again, about another hundred and thirty-six million dollars per a second Form 4 filed Friday⁹, and it paid about four dollars a share more than on the first batch. The first batch’s own disclosure had moved the price. Berkshire paid up for its own signal. A buyer who cares mainly about price stops at nine point nine percent, where the buying stays private. A buyer who crosses the line and keeps going has decided that size matters more than the last four dollars.
The part that deserves the most attention runs the other way. Above ten percent, selling is public too. Every share Berkshire trims will print within two business days, and a holder that has just taught the market to refresh EDGAR for its name cannot leave quietly. Add the six-month short-swing rule, and nothing Berkshire bought this week can be sold at a profit before late March without the gain going to Lennar¹⁰. Berkshire gave up the quiet exit. That is a duration commitment written into securities law, which binds harder than any shareholder letter.
It is also a Lennar bet, not a housing basket. The same June 13F that shows thirteen million Lennar shares shows a few thousand shares of D.R. Horton, the country’s largest builder by homes closed¹¹. Berkshire picked one balance sheet to wait out the housing cycle with.
The Cash Flow Memo’s Lennar page was drafted in July, off the last filed quarter, and Berkshire’s second batch was bought against a worse one. The cashflow read is in Marcus’s column below. Short version: the screen that builds our leaderboard would never have surfaced this trade.
What changes the read. Three filings, in order. The next Form 4: another week of buying says Berkshire has not reached its size; silence says it has. Lennar’s 10-Q, due in October, re-anchors the free cash flow math the memo is still carrying from before the miss. And Berkshire’s third-quarter 13F, due mid-November, shows whether that conviction stops at Lennar or spreads to builders it can still buy in private. The thesis breaks on a single letter. A transaction code of S on a Berkshire Form 4 for Lennar, a sale made in public before new orders turn, would say the duration call was a trade.
Wall Street’s consensus on Lennar: five Underperform-or-Underweight ratings and three Holds in the week and a half around the print, not a Buy among them, with the last month’s price targets averaging about seventy-one dollars¹². Three of those bearish calls were reiterated on the Friday Berkshire bought its biggest single block of the month¹³. The Street published an opinion. Berkshire filed one with the SEC.
The Tape — W2639
Universe of 94 cashflow-memo names, snap dates 2026-09-18 → 2026-09-25. Composite is rank-sum percentile of FCF Yield + NTM Revenue Growth (higher = better balance). Banks and finance-book names shown separately.
Telltales Yield — Top 10
From the Cashflow Desk — Marcus Graham
The composite is built to miss the week’s biggest trade. It scores FCF yield plus NTM revenue growth, and a homebuilder near the bottom of an order cycle carries consensus NTM growth of 1.7% — which is why Lennar appears in neither table above. Going into the print, the memo had Lennar at 16x trailing FCF on $1.8B, 10-Q confirmed; the Q3 release is not a filing, so we re-anchor when the 10-Q lands in October. The leaderboard rewards growth consensus can already see. Berkshire is paying for growth consensus has written down to zero. I read it as a rate call, and the screen has no column for rates. The test is Lennar’s Q4 print in December: if orders are still falling high single digits with mortgage rates unchanged, the 16x anchor comes down and the screen was right to ignore it.
Telltales Yield — Bottom 10
This Week’s Reporters
Sector Medians
Debt / FCF Watch (highest leverage on TTM FCF)
Weekly Price Movement
Top 5 (week-over-week price)
Bottom 5 (week-over-week price)
Banks (shown separately — FCF metric not meaningful)
Finance-book — FCF not comparable
Customer-float / captive-finance / reserve businesses (IBKR broker float, KMX CarMax Auto Finance, PYPL customer funds, CRCL stablecoin reserves). The memo’s operating-FCF method overstates their FCF, so they are held off the ranked leaderboard pending the P&L-waterfall rebuild.
Data Gaps
91 of 92 ranked-eligible names ranked. 1 dropped for missing FCF yield or NTM revenue growth; 7 shown separately (banks + finance-book, FCF not comparable).
Source: cashflow-memo master_2026-09-25.csv. NTM growth from analyst-estimates consensus. Composite is a percentile rank, not a recommendation.
The Issue — This Week's Brief
The Cashflow Memo
Nobody Paid for the Quarter
Costco beat and nobody cared, Lennar missed and Berkshire bought ten percent
The Telltales Weekend Update. Ava Cabot and analyst Marcus Graham walk through what happened this week — and what’s coming next — across the 99 companies in the Cash Flow Memo. About 14 minutes. No filler.
Download the memo at telltales.us. Hunt, Jason, and Mike are back Wednesday on episode E2640.
Chapter markers
* Time | Segment
* 0:00 | Cold open — nobody paid for the quarter
* 0:45 | Theme — pharma buys what it doesn’t own (Merck, Lilly)
* 4:45 | Deep dive — Lennar, and the $212 million disagreement
* 8:45 | Rapid-fire — Costco, Generac, Micron, Nike
* 11:45 | Close — Consensus Watch and the forward week
* 12:45 | Disclaimer
Full transcript
Cold open
Ava: Nobody paid for the quarter this week. They paid for the one after next.
Ava: Costco beat on both lines and the stock went nowhere[^news-cost-flat-20260925]. Lennar missed its quarter and cut its full-year guide[^news-len-guidance-20260924] — and Berkshire bought $212 million of it anyway[^news-len-berkshire-20260924]. Eli Lilly wrote a $2.8 billion check for a clinical-stage company[^news-lly-ataibeckley-20260925], a business with nothing on a pharmacy shelf. Every one of those is the same trade. The number in the release was not what anybody was pricing. Not this quarter — the one after next.
Ava: Telltales Weekend Update. I’m Ava Cabot, with Marcus Graham at the cashflow desk.
Theme — Pharma buys what it doesn’t own
Ava: Start with the drug companies, because they explain the rest of the week. Merck had one of the stranger seven-day stretches you will ever see out of a company that size — page 15 of the memo, and it spent the week buying, winning, and killing. It agreed to buy EyeBio — $1.3 billion up front, up to $3 billion with milestones[^news-mrk-eyebio-20260924]. It won an FDA approval with Eisai for a kidney-cancer combination[^news-mrk-eisai-20260925]. It hit the primary endpoint in a pivotal trial for diabetic macular edema, per the company’s own release[^news-mrk-brunello-20260925]. And then it walked away from something: Merck and Daiichi Sankyo pulled their accelerated-approval application in small-cell lung cancer, saying the evidence wasn’t sufficient[^news-mrk-daiichi-20260925].
Ava: Three different verdicts on three different assets, inside five days. Marcus, the cashflow take.
Marcus: Merck is the cheapest large-cap drug company the memo prices, and it spent the week buying something that doesn’t sell yet. The memo has Merck at 23x free cash flow[^memo-mrk-evfcf-20260831], on $18 billion of trailing free cash flow[^memo-mrk-fcf-20260831], against $44 billion of net debt[^memo-mrk-netdebt-20260831]. A multiple that low is the market saying it has already marked down the back half of the decade. So the EyeBio check and the Daiichi withdrawal are the same decision made twice — fund the pipeline you believe, stop funding the one you don’t. The withdrawal is the part I’d weight. A company pulling its own filing is repricing its pipeline before the agency does it for them.
Ava: Pulling your own filing. That is not a press release anybody enjoys writing.
Ava: Now Lilly. Different problem entirely. Lilly did four things in five days. It agreed to buy AtaiBeckley, clinical stage, for $2.8 billion[^news-lly-ataibeckley-20260925]. It signed a research and license agreement with China’s InnoCare worth up to $3.35 billion[^news-lly-innocare-20260925]. It took an FDA approval for Onswik, a once-weekly basal insulin for type 2 diabetes[^news-lly-onswik-20260924]. And it broke ground on a $6.5 billion manufacturing plant in Houston[^news-lly-houston-20260924]. And then Dave Ricks went on television and said the thing that actually matters. 700,000 new seniors have started GLP-1s since Medicare coverage opened in July — and 70% of them picked Lilly[^news-lly-medicare-20260921]. He put the share of new pill starts going to Lilly’s Foundayo at about a third[^news-lly-foundayo-20260921].
Marcus: At Lilly’s multiple, the deals and the approval don’t move it. The memo has Lilly at 50x free cash flow[^memo-lly-evfcf-20260824]. At 50x, a new plant and an insulin approval are rounding. What you are being asked to pay for is the Medicare cohort — that 70% of 700,000 new starts[^news-lly-medicare-20260921], and the pill share Ricks quoted alongside it points the same direction. Hold that and the multiple has a story behind it. Lose it and nothing else on this week’s list closes the gap.
Ava: One at 23x, one at 50x, doing the same thing in the same week.
Marcus: Same act, opposite reasons. Merck is buying because at 23x the market has already written its pipeline down. Lilly is buying because at 50x the price assumes a pipeline that has to keep arriving. Neither check is about this quarter. What I’d watch into year-end is whether either of them has to fund the next one with borrowed money. Merck is already carrying $44 billion of net debt.
Deep dive — Lennar
Ava: The sharpest disagreement in the market this week was about a homebuilder, and it was not close. Lennar, page 17 of the memo, reported. Earnings of $1.23 against a consensus of $1.29, with revenue down 8%[^news-len-q3miss-20260923]. New orders fell 9%, on deliveries of 20,840 homes[^news-len-orders-20260924]. Then management took about 2,000 homes off both ends of the full-year delivery guide, down to 80,000-81,000[^news-len-guidance-20260924]. That is a bad quarter under any reading of it. And while it was landing, Berkshire Hathaway bought $212 million of the stock across three September sessions and took the position past 10% of the company[^news-len-berkshire-20260924].
Marcus: Berkshire is not underwriting this quarter. Nobody buys 10% of a homebuilder for a delivery guide. The memo has Lennar at 16x free cash flow going into this print[^memo-len-evfcf-20260706], on $1.8 billion of trailing free cash flow[^memo-len-fcf-20260706] — and that page was drafted off the last filed quarter, so we re-anchor when the 10-Q lands. At a stake that size you are buying the land and the balance sheet, and you are accepting an ugly income statement for as long as mortgage rates sit where they sit.
Ava: So the print is almost beside the point.
Marcus: The print is the entry price. But separate them, because they measure different things. Taking 2,000 homes off the guide is a statement about what Lennar can build and close[^news-len-guidance-20260924]. Orders down 9% is a statement about who is walking into the sales office[^news-len-orders-20260924]. Management controls the first one. The bond market controls the second. Orders are what I’d track into the next print, because that is the number that tells you whether this is a cycle bottom or a step down.
Ava: And here is the part that doesn’t fit the headline. In the same week it cut the delivery guide, Lennar declared its regular 50-cent quarterly dividend[^news-len-dividend-20260923], and it opened a brand-new division in Iowa — five communities around Des Moines[^news-len-iowa-20260924].
Marcus: That tells you what kind of downturn management thinks this is. You don’t stand up a new division in a market you’re leaving. The memo has Lennar carrying $2.5 billion of net debt[^memo-len-netdebt-20260706] against that $1.8 billion of trailing free cash flow[^memo-len-fcf-20260706] — that is the balance sheet paying for a new geography and a 50-cent dividend in the same week it guides deliveries down. Read those together and you get a company slowing into a rate problem while it keeps buying optionality for the recovery.
Ava: And Berkshire has taken the other side of that.
Marcus: The other side of the next three years of it. I’d weight it as a duration call rather than a quarter call — call it a position that the mortgage rate is a cyclical number and the housing shortage is a structural one. What would change my read is that order line. If orders are still falling high single digits two prints from now with rates unchanged, the structural half of the argument is weaker than it looks, and the 16x anchor comes down with it.
Ava: $212 million, into a miss and a guidance cut. That is the trade of the week, and it isn’t about the week.
Rapid-fire
Ava: Four to close, starting with Costco, and this one is the whole show in a single print. Fiscal-year revenue of $303 billion, up 10%; earnings of $20.76 a share, up 15% — straight from the company[^news-cost-q4-20260924]. It beat, and the stock sat there[^news-cost-flat-20260925]. The memo has Costco at 52x free cash flow[^memo-cost-evfcf-20260727], and that is the answer. At 52x, a good quarter is the baseline, not the news. The rest of Costco’s week is more interesting than the beat was. 33 new warehouses are coming in fiscal 2027[^news-cost-expansion-20260925]. $184 million of tariff refunds are going back into prices rather than into margin[^news-cost-tariff-20260925]. And the food-court churro is back[^news-cost-flat-20260925]. Costco knows precisely which of those three the members will notice.
Ava: One follow-up on last week’s lead. Analysts now have Generac’s data-center revenue topping $3 billion by 2028, on the back of the Amazon deal[^news-gnrc-datacenters-20260925].
Ava: Two forward-week names and we’re done. Micron reports Wednesday[^earn-mu]. Two things sit underneath that print. A Munich court handed China’s YMTC two injunctions over 3D NAND patents that could restrict Micron’s supply into Germany[^news-mu-ymtc-20260924]. And Micron’s two-year buyback moratorium, the one attached to its CHIPS Act money, expires in December[^news-mu-buyback-20260923]. The memo has Micron at 24x free cash flow[^memo-mu-evfcf-20260831] on $43 billion of trailing free cash flow[^memo-mu-fcf-20260831]. A company that size with the buyback handcuffs coming off before year-end is a capital-allocation story wearing a memory print. Micron also named Deirdre Hanford to run its research labs and put out a 512GB DDR5 server module it calls the first of its kind[^news-mu-ddr5-20260925].
Ava: Nike reports Thursday[^earn-nke], and walks in off a downgrade. Bank of America moved Nike to Underperform and cut its price target from $47 to $30[^news-nke-bofa-20260925]. The same week, Kylian Mbappé ended a 20-year relationship with Nike and signed his boot deal with On[^news-nke-mbappe-20260922]. The memo has Nike at 16x free cash flow[^memo-nke-evfcf-20260706]. Cheap, if the brand is intact. And CarMax opens the week Tuesday[^earn-kmx].
Close
Ava: That’s the show. Wall Street’s consensus this week was written in price targets: Bank of America took Nike from $47 down to $30[^news-nke-bofa-20260925]. That’s a forecast. Berkshire’s $212 million is a position.
Ava: Three prints in the forward week, and on this week’s evidence not one of them is what will set the price. Watch the 90 minutes after each release, not the release.
Ava: On Wednesday’s show, Hunt put the 30-year mortgage at 7.15, 7.20, and said plainly that it hurts housing[^ep-e2639]. That same week, Lennar cut its guide and Berkshire crossed 10%. Both of those are true at once, and that is the whole episode: the quarter is not what anybody is buying.
Ava: Hunt, Jason, and Mike are back Wednesday on episode 2640. Download this week’s Cash Flow Memo at telltales.us, and send us a note through the Substack — we read them.
Ava: The show is produced entirely with AI tools, and both voices you’re hearing are AI-generated.
Disclaimer
Ava: This podcast is intended for informational purposes only. You should always do your own work to determine if an investment is suitable for you. The views expressed on this podcast are the host alone and do not constitute an offer to sell or a recommendation to purchase, or a solicitation of an offer to buy any security, nor a recommendation for any investment product or service. While certain information contained herein has been obtained from sources believed to be reliable, neither the host nor any of their employers or their affiliates have independently verified this information, and its accuracy and completeness cannot be guaranteed. Accordingly, no representation or warranty, express or implied, is made as to, and no reliance should be placed on, the fairness, accuracy, timeliness, or completeness of this information. The host and all employers and their affiliated persons assume no liability for this information and no obligation to update the information or analysis contained herein in the future, and may or may not hold positions in the securities mentioned.
Sources
* CNBC. (2026, September 21). Eli Lilly CEO says 700,000 new seniors have started GLP-1s after Medicare coverage, and 70% are on Lilly drugs. CNBC. https://www.cnbc.com/2026/09/21/eli-lilly-ceo-david-ricks-glp-1s-medicare-coverage.html
* CNBC. (2026, September 21). Eli Lilly CEO tells CNBC one-third of new GLP-1 pill patients are taking Foundayo, as drugmaker ramps up production. CNBC. https://www.cnbc.com/2026/09/21/lilly-is-ramping-up-future-manufacturing-for-foundayo-obesity-pill.html
* CNBC. (2026, September 22). Soccer superstar Kylian Mbappé on ending 20-year journey with Nike: It was time to change. CNBC. https://www.cnbc.com/2026/09/21/kylian-mbappe-nike-exit-on-soccer-footwear.html
* CNBC. (2026, September 23). History rhymes for Meta, and our plan for Micron ahead of earnings. CNBC. https://www.cnbc.com/2026/09/23/history-rhymes-for-meta-and-our-plan-for-micron-ahead-of-earnings.html
* CNBC. (2026, September 23). Lennar shares pop as Berkshire builds almost a 10% stake in beleaguered homebuilder. CNBC. https://www.cnbc.com/2026/09/22/lennar-shares-pop-as-berkshire-builds-almost-a-10percent-stake-in-beleaguered-homebuilder.html
* CNBC. (2026, September 25). Risks are rising for Nike’s turnaround, Bank of America says. Sell shares now. CNBC. https://www.cnbc.com/2026/09/25/risks-are-rising-for-nikes-turnaround-bank-of-america-says.html
* Construction Dive. (2026, September 24). Eli Lilly breaks ground on $6.5B Houston manufacturing plant. Construction Dive. https://www.constructiondive.com/news/eli-lilly-breaks-ground-houston-manufacturing-plant/831307/
* Costco Wholesale Corporation. (2026, September 24). Costco Wholesale Corporation reports fourth quarter and fiscal year 2026 operating results [Press release]. Costco Investor Relations. https://investor.costco.com/news/news-details/2026/Costco-Wholesale-Corporation-Reports-Fourth-Quarter-and-Fiscal-Year-2026-Operating-Results/default.aspx
* Eastern Herald. (2026, September 24). Lennar (NYSE:LEN) stock falls 1.22% to $82.03 on September 23: Berkshire crosses 10%. Eastern Herald. https://easternherald.com/2026/09/24/lennar-len-stock-berkshire-hathaway-10-percent-september-23/
* HousingWire. (2026, September 24). Lennar’s new Iowa division opens 5 Des Moines-area communities. HousingWire. https://www.housingwire.com/articles/lennar-enters-des-moines-market/
* Investing.com. (2026, September 24). Eli Lilly gets FDA approval for once-weekly diabetes insulin. Investing.com. https://www.investing.com/news/stock-market-news/eli-lilly-gets-fda-approval-for-onceweekly-diabetes-insulin-93CH-4914987
* Lennar Corporation. (2026, September 23). Lennar Corporation declares quarterly dividends [Press release]. Lennar Newsroom. https://newsroom.lennar.com/2026-09-23-Lennar-Corporation-Declares-Quarterly-Dividends
* MarketScreener. (2026, September 25). Merck, Eisai renal cell cancer drug combination gets US FDA approval. MarketScreener. https://www.marketscreener.com/news/merck-eisai-renal-cell-cancer-drug-combination-gets-us-fda-approval-ce785adfdc80f726
* Merck & Co. (2026, September 25). Merck’s remigromig, a tri-specific agonist of the Wnt pathway, met primary endpoint in the pivotal Phase 2b/3 BRUNELLO study of adults with diabetic macular edema [Press release]. https://www.merck.com/news/mercks-remigromig-a-tri-specific-agonist-of-the-wingless-related-integration-site-wnt-pathway-met-primary-endpoint-in-the-pivotal-phase-2b-3-brunello-study-of-adults-with-diabetic-macular/
* Mugglehead. (2026, September 25). Merck, Daiichi withdraw ifinatamab FDA application. Mugglehead. https://mugglehead.com/merck-daiichi-sankyo-withdraw-ifinatamab-deruxtecan-bla/
* MyNorthwest. (2026, September 25). Costco receives $184M in tariff refunds, plans to pass savings to shoppers. MyNorthwest. https://mynorthwest.com/local/costco-tariff-refunds/4279798
* Ophthalmology Times. (2026, September 24). Merck to acquire EyeBio for upfront payment of $1.3 billion. Ophthalmology Times. https://www.ophthalmologytimes.com/view/merck-to-acquire-eyebio-for-upfront-payment-of-1-3-billion
* PharmExec. (2026, September 25). Eli Lilly to acquire AtaiBeckley for $2.8 billion. Pharmaceutical Executive. https://www.pharmexec.com/view/lilly-acquire-ataibeckley-2-billion
* PharmExec. (2026, September 25). InnoCare enters $3.35 billion research collaboration and license agreement with Eli Lilly. Pharmaceutical Executive. https://www.pharmexec.com/view/innocare-3-billion-research-collaboration-license-agreement-eli-lilly
* Pomegra News. (2026, September 24). Lennar Q3 2026 miss deepens housing crisis signal. Pomegra. https://pomegra.io/news/lennar-q3-2026-miss-deepens-housing-crisis-signal
* StocksToTrade. (2026, September 25). GNRC soars as Generac lands $8B Amazon data center deal. StocksToTrade. https://stockstotrade.com/news/generacholdlingsinc-gnrc-news-2026_09_25-2/
* The Hill. (2026, September 25). 33 new Costco stores announced. Here’s what we know about their locations. The Hill. https://thehill.com/business/6112042-33-new-costco-locations-announced-heres-what-we-know-about-their-locations/
* TheStreet. (2026, September 24). Berkshire buys $212 million of tumbling housing stock. TheStreet. https://www.thestreet.com/investing/stocks/berkshire-hathaway-buys-212-million-lennar-len-stock
* Tom’s Hardware. (2026, September 24). China’s YMTC wins patent battle against Micron in ongoing 3-year legal war over memory patents; new injunctions could restrict Micron’s supply into Germany. Tom’s Hardware. https://www.tomshardware.com/pc-components/storage/chinas-ymtc-wins-patent-battle-against-micron-in-ongoing-3-year-legal-war-over-memory-patents-new-injunctions-could-restrict-microns-supply-into-germany
* Yahoo Finance. (2026, September 25). Costco confirms its food court churro is coming back — COST stock flat despite Q4 beat. Yahoo Finance. https://finance.yahoo.com/markets/stocks/articles/costco-confirms-food-court-churro-043258079.html
* Yahoo Finance. (2026, September 25). Is Micron (MU) quietly recasting its AI memory strategy with new leadership and 512GB DDR5? Yahoo Finance. https://sg.finance.yahoo.com/news/micron-mu-quietly-recasting-ai-021239442.html
Note: reference 22 is cited twice in the episode — once for the Amazon supply agreement terms and once for the sell-side $3 billion 2028 data-center revenue estimate carried in the same article.
Internal data
Internal data is provided on a best efforts basis.
Hunt, Mike, and Jason walk the Cash Flow Memo with WTI near $101, a $20 discount to next year’s strip, and the 10-year above 5%. Then the data against AI job doomers, and what it would cost Meta to serve its new Muse agent at scale.
The Cashflow Memo
Key Takeaways
* Hunt holds oil flat on Exhibit C: Iranian supply has only slipped from 3.0 to 2.1 million bbl/day under embargo because domestic use absorbs it, and China’s real demand is off ~400K bbl/day (16.2 to 15.8M), not the 2-3M press figure, since the cuts were re-exported product.
* WTI at ~$101 against a ~$77 2027 strip leaves $20+ of backwardation that Hunt expects to keep rolling forward as a Hormuz risk premium; low-debt producers can hedge less and sell more barrels at ~$90 while reinvestment decisions key off ~$72-73.
* With the 10-year up ~10bp in a day to ~5.1% and 30-year mortgages ~7.15-7.2%, Hunt asks what it will take for Washington to treat the deficit as a problem and suspects it needs an ’07-’08-style capital-markets closure.
* Mike and Jason answer AI job doomers with data: Medicare-enrolled radiologists and radiologic technologists have risen since the 2016-17 radiology is done call, programmer counts still sit above the pre-COVID trend, and construction hiring (led by electrical contractors) keeps climbing on data center builds.
* Meta’s Muse, an OpenClaw-style always-on personal agent at ~640K users, would need roughly 265,000 Graviton CPUs to serve 100M users, about $50M/day ($1.5B/month) at AWS on-demand rates before inference; separately, Grail meets the FDA on Galleri today after its failed trial, and the Anthropic S-1 is expected any day.
Show Notes
[00:00] Intro
[00:27] Oil: supply and China demand Iran still pumps ~2.1M barrels a day under embargo, and China’s real consumption is down ~400K barrels a day, not the millions in the headlines.
[02:16] The $20 backwardation Why the gap between spot and the 2027 strip keeps rolling forward, and how low-debt producers can take advantage of it.
[04:22] Deficit and the 10-year Rates jump again, mortgages top 7%, and Hunt asks what it will take for Washington to treat the deficit as a problem.
[07:06] AI and jobs: the data Radiology was declared dead in 2017; radiologist and technologist counts are up since. Software jobs remain above the pre-COVID trend.
[12:08] 10,000 agents, three days OpenAI’s swarm approach to century-old math problems as a picture of the productivity multiplier.
[13:35] Construction keeps hiring Electrical contractors, site prep, and plumbing keep rising on data center builds.
[15:57] Meta Muse and OpenClaw A short history of open source, and how an open-source personal agent became Meta’s consumer product.
[20:12] Serving 100 million users Napkin math: ~265,000 Graviton CPUs and ~$50M a day at on-demand rates before inference.
[25:21] Frontier labs vs. consumer agents Why Muse and Grokbot target a different market than OpenAI and Anthropic, and why Google is surprisingly late.
[27:59] Agents in healthcare Clinical-trial adherence as the obvious use case, once HIPAA is solved.
[28:58] Grail at the FDA Galleri’s FDA meeting after a failed trial, and why multi-cancer detection is coming regardless of who gets there first.
[30:26] Anthropic: wet lab and S-1 AI-designed drugs in a San Francisco wet lab, and the S-1 everyone is waiting to read.
Download this week’s Cash Flow Memo at telltales.us and subscribe for next week’s episode, including the Anthropic S-1 if it drops.
Cashtags
$AMZN $GH $GOOGL $GRAL $META $MSFT
This post and the information herein are intended for informational purposes only. The views expressed herein are the author’s alone and do not constitute an offer to sell, or a recommendation to purchase, or a solicitation of an offer to buy, any security, nor a recommendation for any investment product or service. While certain information contained herein has been obtained from sources believed to be reliable, neither the author nor any of his employers or their affiliates have independently verified this information, and its accuracy and completeness cannot be guaranteed. Accordingly, no representation or warranty, express or implied, is made as to, and no reliance should be placed on, the fairness, accuracy, timeliness or completeness of this information. The author and all employers and their affiliated persons assume no liability for this information and no obligation to update the information or analysis contained herein in the future.
▶ Explore this week’s Tape — live, sortable, drill-down →
Read the Vesting Schedule
Amazon signed two documents with Generac on Tuesday. Everybody covered the smaller one.
The supply agreement is the small document: backup generators for Amazon data centers, initial deliveries expected to total two point four billion dollars across twenty-twenty-seven and twenty-twenty-eight. The warrant is the large one. Amazon took the right to buy about 1.7 million Generac shares at a fixed strike, roughly eighteen percent of it vesting the day it was signed. The rest vests in tranches, against — Generac’s words, in its own 8-K — aggregate gross payments, net of certain offsets, received by the Company and its global affiliates from or on behalf of Amazon and its affiliates for backup power generators for Amazon data centers, up to a total of $8 billion.¹
Eight billion.
Nobody has committed to eight billion, and that is exactly why the number is worth something. A press release carries the figure a company is confident enough to print. A vesting ladder carries the figure two sets of lawyers negotiated to, because Amazon does not build itself a staircase it cannot climb and Generac does not hand over equity against volume it cannot picture shipping. The ceiling in the filing is the honest one. It runs to nearly two years of Generac’s entire trailing revenue², from a single customer, at a company that until Tuesday was understood as selling generators to homeowners in a storm.
None of this is improvised. Amazon has run the structure at least twice before, and the tell is always the same clause. Air Transport Services Group, 2016: warrants for up to nineteen point nine percent of the company, the unvested portion vesting as ATSG delivers additional aircraft leased under the ATSA, or as the Company achieves specified revenue targets³. Plug Power, 2017: fifty-five million shares, vesting based on Amazon’s payment of up to $600 million to the Company in connection with Amazon’s purchase of goods and services from the Company⁴. In every version the supplier funds the customer’s equity out of the supplier’s own order book. Amazon does not write a check for the stake. It buys product, and the stake arrives.
Note where the strike gets set. ATSG’s was nine seventy-three, the closing price a month before that deal was announced⁵. Generac’s sits below where the stock closed Friday. The customer prices its option before the market finds out what the customer is about to do.
And it is an option, not a marriage. When Stonepeak took ATSG private in an all-cash deal at twenty-two fifty a share⁶, Amazon — which by then held about nineteen and a half percent, per FreightWaves — ended up with no stake at all in the business it had supplied for eight years⁷. A warrant on public equity cannot survive into a private company. It converts to cash, or it lapses.
Which reframes what happened to Generac this week. The stock did not move because a generator company won an order. It moved because a generator company disclosed, in a filing, the maximum size of its new relationship, and the market read the ceiling as the forecast. Getting there is a manufacturing problem, not a demand problem. The cashflow read is in Marcus’s column below; short version, the tape priced the order book and skipped the factory.
What changes the read. The test is the Q3 print, and it is the capital expenditure line, not the revenue line. Eight billion dollars of generators has to be built before any of it can be collected, and the Cash Flow Memo’s page on this name was drafted before Amazon walked in the door — the multiple sitting on it belongs to a storm-demand business, not a contracted-infrastructure supplier. Guide capital expenditure up hard, and fund it, and the ladder is live. Leave capital expenditure where it is and the ladder is decoration. The second thing to watch is quieter: every filing from here reports how many warrant shares vested, which is a quarterly readout on what Amazon actually paid. Generac has agreed to publish its largest customer’s purchase orders. Not many suppliers do.
Wall Street’s consensus on Generac after the Amazon deal: the stock has further to go. Right direction, wrong document. Consensus is pricing the press release. Amazon priced the vesting schedule.
The Tape — W2638
Universe of 94 cashflow-memo names, snap dates 2026-09-11 → 2026-09-18. Composite is rank-sum percentile of FCF Yield + NTM Revenue Growth (higher = better balance). Banks and finance-book names shown separately.
Telltales Yield — Top 10
From the Cashflow Desk — Marcus Graham
Generac got repriced on an order book, and nobody has priced what filling it costs. Capex ran $168.9M TTM — a company sized to sell generators into storms, not to stand up capacity for a hyperscaler. The warrant vests against Amazon payments running up to $8B, and consensus is reading that ceiling as a demand forecast; it is a negotiated ceiling, and it only pays Amazon if Generac builds. Debt/FCF at 2.7x says there is balance-sheet room for the first tranche. It does not say there is room for the ladder. The test on the Q3 print is the capital expenditure guide, not the revenue line. We re-anchor the multiple when the Q3 10-Q files.
Telltales Yield — Bottom 10
This Week’s Reporters
Sector Medians
Debt / FCF Watch (highest leverage on TTM FCF)
Weekly Price Movement
Top 5 (week-over-week price)
Bottom 5 (week-over-week price)
Banks (shown separately — FCF metric not meaningful)
Finance-book — FCF not comparable
Customer-float / captive-finance / reserve businesses (IBKR broker float, KMX CarMax Auto Finance, PYPL customer funds, CRCL stablecoin reserves). The memo’s operating-FCF method overstates their FCF, so they are held off the ranked leaderboard pending the P&L-waterfall rebuild.
Data Gaps
91 of 92 ranked-eligible names ranked. 1 dropped for missing FCF yield or NTM revenue growth; 7 shown separately (banks + finance-book, FCF not comparable).
Source: cashflow-memo master_2026-09-18.csv. NTM growth from analyst-estimates consensus. Composite is a percentile rank, not a recommendation.
The Issue — This Week's Brief
The Cashflow Memo
The AI Gold Rush Paid the Shovel Sellers and Billed the Diggers
Amazon handed Generac a $2.4 billion order and took warrants on its stock. Oracle got a congressional subpoena. Same buildout, opposite ends of it.
The Telltales Weekend Update. Ava Cabot and analyst Marcus Graham walk through what happened this week — and what’s coming next — across the 86 companies in the Cash Flow Memo. About 14 minutes. No filler.
Download the memo at telltales.us. Hunt, Jason, and Mike are back Wednesday on episode E2639.
Chapter markers
* Time | Segment
* 0:00 | Cold open
* 0:45 | Theme — Who’s getting paid (Generac, Lennar)
* 4:45 | Deep dive — Oracle
* 8:45 | Rapid-fire (Merck, Lilly, Salesforce, Google, Tesla)
* 11:45 | Close
* 12:40 | Closing disclaimer
Full transcript
Cold open
Ava: This week the AI gold rush paid the shovel-sellers and billed the diggers. Generac, a company that makes backup generators, signed a $2.4 billion supply deal with Amazon and watched its stock run more than 19%[^news-gnrc-amazon-supply-20260916][^news-gnrc-shares-20260917]. Oracle, which is actually building the thing, spent the week getting its founder subpoenaed by Congress and calling off the sale of $7.5 billion of his own stock[^news-orcl-subpoena-20260917][^news-orcl-ellison-20260913]. Same buildout. Very different end of it.
Ava: Telltales Weekend Update. I’m Ava Cabot, with Marcus Graham at the cashflow desk.
Theme — Who’s getting paid
Ava: Generac spent 20 years selling generators to people whose power goes out in a storm. This week it found a customer whose power can never go out at all. Amazon signed a $2.4 billion long-term supply agreement for backup generators at its data centers, first deliveries expected in 2027 and 2028[^news-gnrc-amazon-supply-20260916]. And then Amazon did something more interesting than paying. It took warrants, the right to buy up to $340 million of Generac stock[^news-gnrc-amazon-warrant-20260916]. Shares ran more than 19%[^news-gnrc-shares-20260917]. Marcus, who got the better end of that bargain?
Marcus: Amazon just bought a supplier’s upside with the supplier’s own order book. The Cash Flow Memo has Generac generating about $600 million of free cash flow a year[^memo-gnrc-fcf-20260831]. That $2.4 billion order is roughly four years of everything the company makes, landing in one contract[^news-gnrc-amazon-supply-20260916]. The warrant is Amazon telling you it knows exactly how big that order is against the company that has to fill it. What I’d watch is the 2027 delivery date, because between here and there Generac has to build capacity for an order that size, and nobody has shown us what that costs yet.
Ava: And Generac’s own multiple?
Marcus: 29x free cash flow in the memo[^memo-gnrc-evfcf-20260831], and that price was set before Amazon walked in the door. It is not a bargain for a generator company. What you’re being asked to underwrite now is whether a storm-demand business with lumpy revenue turns into a contracted-infrastructure supplier with a visible order book, because those two things do not trade at the same multiple. The 19% move was the market starting that argument[^news-gnrc-shares-20260917]. It did not settle it.
Ava: Lennar told you two things about the American housing market this week, and they don’t agree with each other. Orders fell 9%, to just under 21,000 homes, short of its own guide[^news-len-orders-20260917]. Full-year deliveries cut to 80,000-81,000, down from 82,000-83,000[^news-len-guidance-20260917]. Margins actually got better, about 15.8% gross, with buyer incentives down to 12%[^news-len-margins-20260918]. And then, in the same week it was cutting guidance, Lennar bought 444 lots for $60 million, right next to Taiwan Semi’s $265 billion project in Phoenix[^news-len-tsmc-land-20260917]. Marcus, what’s a homebuilder doing in a fab town?
Marcus: Lennar is pulling back across the whole business and buying into one specific town. The memo has Lennar at 16x free cash flow on $1.8 billion[^memo-len-evfcf-20260831][^memo-len-fcf-20260831], and that is what a homebuilder looks like late in a cycle: a cheap-looking multiple bolted to a shrinking order book. $60 million of dirt does not move that math. It’s a rounding error against the book[^news-len-tsmc-land-20260917]. What it tells you is where this company thinks the next household formation comes from, and it isn’t the general market. The test is the next order print. If orders keep falling and they keep buying land, that’s conviction. If they stop buying land, that’s the signal.
Ava: Two companies, one buildout. One of them is selling the picks. The other is digging a hole in Phoenix and hoping the town shows up.
Deep dive — Oracle
Ava: Last week on this show, we said Oracle’s demand question was answered and its funding question was wide open[^ep-w2637]. Oracle spent this week answering the funding question three separate times, and not one of those answers came from the bond market. Quick context for anyone joining: Oracle printed on September 11, revenue $19.3 billion, up 30%, and the company led its own release with triple-digit growth in cloud infrastructure[^news-orcl-q1-20260911]. We did that print here last week. What happened after it is the story. Two days later, Larry Ellison called off a plan to sell up to 50 million shares, call it $7.5 billion[^news-orcl-ellison-20260913]. On September 17, Oracle signed a 433 MW power purchase agreement with RWE[^news-orcl-rwe-20260917]. And on that same day, the House Veterans Affairs Committee subpoenaed Ellison and Mike Sicilia over the $27 billion VA health-records contract[^news-orcl-subpoena-20260917]. Marcus, three items, one week. What do they have in common?
Marcus: All three are about the same question: who pays for the rest of this build, and on what terms. Oracle funded a big piece of last quarter with equity, not the bond market, which we walked through here last week[^ep-w2637]. The Cash Flow Memo already carries $126 billion of net debt on this name against a $561 billion enterprise value[^memo-orcl-netdebt-20260831][^memo-orcl-ev-20260831]. Roughly a fifth of what you’re buying here is already borrowed. That’s the constraint on this story. It was never the demand.
Ava: Start with Ellison, then.
Marcus: Ellison calling off $7.5 billion of stock sales is the most interesting item of the three, and not for the reason it’s being covered[^news-orcl-ellison-20260913]. Oracle has been funding this build partly by issuing stock. The largest holder deciding not to put 50 million shares into that same market is a fact about supply. His reasons are not disclosed and I’m not going to invent them for him. What it tells you is that the equity channel is still the open channel, and the people closest to it are not leaning on it.
Ava: And the power deal?
Marcus: That’s the one I’d put at the top of the list. 433 megawatts, contracted, announced this week[^news-orcl-rwe-20260917]. Power is the binding constraint on this buildout, not chips, and companies about to slow down do not go and lock up electricity years ahead of needing it. Set it next to capital spending running about $76 billion over the trailing twelve months[^data-orcl-capex-20260918] and it’s the same message told twice: Oracle is behaving like a company that believes its own order book. That’s the leg I’d watch into the next print, and it’s cheap to watch, because power agreements get announced.
Ava: Which leaves the subpoena.
Marcus: The subpoena is the one that could actually move the numbers. $27 billion of health-records revenue with a political counterparty[^news-orcl-subpoena-20260917]. That is not a fraud allegation and I’m not going to dress it up as one — it’s a committee asking two executives to come and explain a contract. But contracts with political counterparties get renegotiated, and that one is big enough to matter against a forward revenue line running near 46% growth[^data-orcl-ntmgrowth-20260918]. I’d weight it about 60/40 that the growth holds through the next two prints, and the VA line is where I’d expect the first crack if it doesn’t.
Ava: So 60/40, not a lock.
Marcus: Not a lock. The VA risk is real, and the funding is working. Both of those are true at once.
Ava: Oracle sits on page 2 of the memo, next to Salesforce. This week those two are the same bet told two different ways: one is building the capacity, the other is selling the software that runs on top of it.
Rapid-fire
Ava: Right. The rest of the week, quickly.
Ava: The FDA had a busy week and Merck had the best of it. Merck and Moderna’s Phase 3 melanoma trial, a personalized cancer vaccine given alongside Keytruda, hit both its primary and key secondary endpoints in patients whose melanoma had been fully resected[^news-mrk-interpath-20260917]. Same week, the FDA expanded the label on Merck’s Winrevair in pulmonary arterial hypertension, on Phase 3 ZENITH data[^news-mrk-winrevair-20260917]. For a company that has spent three years being asked what replaces Keytruda when the patents run out, that’s two answers inside seven days, and one of them is a mechanism Merck does not have to invent on its own.
Ava: Eli Lilly, meanwhile, got an approval that has nothing to do with weight loss. The FDA cleared Inluriyo, Lilly’s oral estrogen-receptor degrader, in combination with Verzenio for breast cancer, which pushes Lilly further ahead in the oral SERD race[^news-lly-serd-20260918].
Ava: Salesforce did the thing every software company is trying to do right now, and actually put a number on it. Fiscal 2030 revenue guidance of more than $63 billion, ahead of what the street was carrying[^news-crm-guidance-20260916]. And the detail underneath it: 20,000 Salesforce employees are now in roles that did not exist a year ago[^news-crm-redeploy-20260918]. Same people, different jobs. That is what a software company looks like when it decides the AI story is an internal reorganisation before it is a revenue line.
Ava: Google kept its ad-tech business and lost control of it. A federal judge declined to order a breakup, and instead requires Google to appoint an internal antitrust compliance officer for six years[^news-googl-remedy-20260916]. The unsealed opinion is the part that bites: Google has to expose Prebid-equivalent interfaces from its exchange and its ad server to rival publisher ad servers on identical terms, share winning and losing bid data, and end AdWords direct bidding into its own ad server[^news-googl-prebid-20260916]. PubMatic’s chief executive reckons building that takes 12 to 15 months, against obligations that take effect 60 days from final judgment[^news-googl-timeline-20260918]. 60 days to comply. 12 to 15 months to build the thing that complies. That gap is the whole story.
Ava: And Tesla has to explain its robotaxi under oath. NHTSA gave Tesla until September 30 to explain, under oath, how the driverless Cybercab meets federal safety rules[^news-tsla-cybercab-20260917]. Separately, Tesla began auditing Chinese suppliers to size up an Optimus production ramp[^news-tsla-optimus-20260918]. The memo carries Tesla at 180x free cash flow[^memo-tsla-evfcf-20260831], which is a number that only works if both of those go right.
Ava: And the two weeks ahead. Page 8 of the memo has two reporters: Costco on Thursday, consensus around $94.9 billion of revenue[^earn-cost], and CarMax the following Tuesday[^earn-kmx]. Micron reports Wednesday the 30th, about $50.6 billion consensus revenue, and the memo has it at 24x free cash flow going in[^earn-mu][^memo-mu-evfcf-20260831]. Nike closes it out Thursday the 1st, sharing page 17 with Lennar[^earn-nke].
Close
Ava: That’s the show. Wall Street’s consensus on Generac after the Amazon deal: the stock has further to go[^news-gnrc-shares-20260917]. Maybe. It also has a $2.4 billion order it now has to actually build, first deliveries 2027[^news-gnrc-amazon-supply-20260916]. Consensus is pricing the contract. Somebody still has to fill it.
Ava: The shovel-sellers got paid this week. The diggers got the bill. Both of those are true at the same time, and the Cash Flow Memo is where you tell them apart. Download it at telltales.us.
Ava: On Wednesday’s show, Mike made the case that the data-center capacity these companies already own is worth far more than what it’s carried at on the books, because it keeps getting harder to build and the demand keeps rising[^ep-e2638]. This week Generac put a price on a piece of that, and Oracle showed you what building it costs.
Ava: Hunt, Jason, and Mike are back Wednesday on episode 2639, taking up the data-center jobs fight they promised at the end of this week’s show[^ep-e2638].
Ava: If something in here was wrong, or you want more of one thing and less of another, send it through the Substack. We do change the show based on what you send.
Ava: The show is produced entirely with AI tools, and both voices you’re hearing are AI-generated.
Closing disclaimer
Ava: This podcast is intended for informational purposes only. You should always do your own work to determine if an investment is suitable for you. The views expressed on this podcast are the host alone and do not constitute an offer to sell or a recommendation to purchase, or a solicitation of an offer to buy any security, nor a recommendation for any investment product or service. While certain information contained herein has been obtained from sources believed to be reliable, neither the host nor any of their employers or their affiliates have independently verified this information, and its accuracy and completeness cannot be guaranteed. Accordingly, no representation or warranty, express or implied, is made as to, and no reliance should be placed on, the fairness, accuracy, timeliness, or completeness of this information. The host and all employers and their affiliated persons assume no liability for this information and no obligation to update the information or analysis contained herein in the future, and may or may not hold positions in the securities mentioned.
Sources
* ABC15. (2026, September 17). Lennar snags 444 lots near TSMC’s $265B Phoenix project for $60M. ABC15 Arizona. https://www.abc15.com/news/business/lennar-snags-444-lots-near-tsmcs-265b-phoenix-project-for-60m
* Alphastreet. (2026, September 17). Lennar releases Q3 2026 financial results. Alphastreet. https://news.alphastreet.com/lennar-releases-q3-2026-financial-results/
* Builder Magazine. (2026, September 18). Lennar chooses volume over margin in tougher housing market. Builder. https://www.builderonline.com/builder-100/strategy/lennar-chooses-volume-over-margin-in-tougher-housing-market/
* CNBC. (2026, September 16). Amazon obtains right to buy stock in Generac, boosting power company’s share price. CNBC. https://www.cnbc.com/2026/09/16/amazon-obtains-right-to-buy-up-to-340m-of-generac-boosting-stock-.html
* CNBC. (2026, September 16). Salesforce issues revenue target of over $63 billion for fiscal 2030, beating estimates. CNBC. https://www.cnbc.com/2026/09/16/salesforce-issues-revenue-target-of-63-billion-for-fiscal-2030.html
* CNBC. (2026, September 17). Generac shares surge on big Amazon deal. Wall Street thinks the generator stock has more to go. CNBC. https://www.cnbc.com/2026/09/17/generac-shares-surge-on-amazon-deal-wall-street-thinks-it-has-more-to-go-.html
* FiercePharma. (2026, September 18). Eli Lilly escalates oral SERD battle with FDA approval for Inluriyo, Verzenio combo. FiercePharma. https://www.fiercepharma.com/pharma/eli-lilly-escalates-oral-serd-battle-fda-approval-inluriyo-verzenio-combo
* Healthcare IT News. (2026, September 17). Oracle leaders receive subpoenas to appear before House VA Committee. Healthcare IT News. https://www.healthcareitnews.com/news/oracle-leaders-receive-subpoenas-appear-before-house-va-committee
* HousingWire. (2026, September 18). Lennar defends even-flow, land banking strategy as risks build. HousingWire. https://www.housingwire.com/articles/lennar-q3-2026-earnings/
* Merck & Co. (2026, September 17). Merck and Moderna announce Phase 3 INTerpath-001 trial of intismeran autogene plus KEYTRUDA met endpoints of recurrence-free survival (RFS) and distant metastasis-free survival (DMFS) in patients with completely resected Stage IIB-IV melanoma [Press release]. https://www.merck.com/news/merck-and-moderna-announce-phase-3-interpath-001-trial-of-intismeran-autogene-plus-keytruda-met-endpoints-of-recurrence-free-survival-rfs-and-distant-metastasis-free-survival-dmfs-in-patient/
* Oracle Corporation. (2026, September 11). Oracle announces Q1 results driven by triple digit growth in cloud infrastructure revenues [Press release]. Oracle Investor Relations. https://investor.oracle.com/investor-news/news-details/2026/Oracle-Announces-Q1-Results-Driven-by-Triple-Digit-Growth-in-Cloud-Infrastructure-Revenues/default.aspx
* Oracle Corporation. (2026, September 13). Larry Ellison cancels his plan to sell Oracle stock [Press release]. Oracle Investor Relations. https://investor.oracle.com/investor-news/news-details/2026/Larry-Ellison-Cancels-His-Plan-to-Sell-Oracle-Stock/default.aspx
* Pharmaceutical Technology. (2026, September 17). FDA expands indication for Merck’s Winrevair in pulmonary arterial hypertension. Pharmaceutical Technology. https://www.pharmtech.com/view/fda-merck-winrevair-pulmonary-arterial-hypertension
* PPC Land. (2026, September 18). PubMatic CEO says Google needs up to 15 months to open AdX to Prebid. PPC Land. https://ppc.land/pubmatic-ceo-says-google-needs-up-to-15-months-to-open-adx-to-prebid/
* Reuters. (2026, September 16). Generac, Amazon strike $2.4 billion long-term generator supply deal. Reuters. https://www.reuters.com/business/energy/generac-amazon-strike-24-billion-long-term-generator-supply-deal-2026-09-16/
* Reuters. (2026, September 16). Google should appoint antitrust compliance officer, US judge says in ad tech case. Reuters. https://www.reuters.com/world/google-should-appoint-antitrust-compliance-officer-us-judge-says-ad-tech-case-2026-09-16/
* RWE. (2026, September 17). RWE signs 433 MW virtual power purchase agreement with Oracle [Press release]. RWE Americas. https://www.rwe.com/en/press/rwe-americas/2026-09-17-rwe-signs-433-mw-virtual-power-purchase-agreement-with-oracle/
* Salesforce Ben. (2026, September 18). 20,000 Salesforce employees are now in roles that didn’t exist a year ago. Salesforce Ben. https://www.salesforceben.com/20000-salesforce-employees-are-now-in-roles-that-didnt-exist-a-year-ago/
* South China Morning Post. (2026, September 18). Tesla auditing Chinese suppliers ahead of Optimus roll-out. South China Morning Post. https://www.scmp.com/business/companies/article/3368053/tesla-auditing-chinese-suppliers-ahead-optimus-roll-out-sources
* Teslarati. (2026, September 17). NHTSA just escalated its Tesla Cybercab investigation in a big way. Teslarati. https://www.teslarati.com/nhtsa-just-escalated-its-tesla-cybercab-investigation-in-a-big-way/
* The New York Times. (2026, September 16). Judge orders Google to open ad tech tools to rivals. The New York Times. https://www.nytimes.com/2026/09/16/technology/google-ad-tech-remedies.html
Internal data
Internal data is provided on a best efforts basis.
Earnings slate
Forward earnings dates, times and consensus figures are taken from the episode’s earnings slate, pulled 2026-09-18. See 04. Publishing/shows/weekend-update/W2638/dryrun/earnings_slate.md.
* COST - Costco, 2026-09-24 (Thursday). Consensus EPS 6.53, consensus revenue $94.9B.
* KMX - CarMax, 2026-09-29 (Tuesday). Consensus EPS 0.72, consensus revenue $6.9B.
* MU - Micron, 2026-09-30 (Wednesday). Consensus EPS 31.27, consensus revenue $50.6B.
* NKE - Nike, 2026-10-01 (Thursday). Consensus EPS 0.44, consensus revenue $11.3B.
Hunt, Mike, and Jason walk the Cash Flow Memo with oil back near $100, the 10-year above 5%, and the AI labs talking about extinction. Plus notes from the All-In Summit and a Vertex update.
The Cashflow Memo
Key Takeaways
* Hunt holds oil around $90 with the 12-month strip ~$20 lower despite WTI briefly topping $100; the real squeeze is diesel (~$200/bbl globally) from Ukrainian strikes on Russian refineries, Chinese product-export limits, and uninsurable Gulf product tankers.
* Gas has settled at ~$3.50 for both 2026 and 2027 as higher oil prices drive Permian associated-gas supply while power-sector demand stays flat on coal-plant life extensions and subsidized solar.
* With the 10-year above 5% and the Fed’s unanimous 25bp hike to 3.75-4%, Hunt still argues most investable assets belong in equities, about half in the leading dozen names, since the Mag 7 trades ~21x versus ~40x two years ago while growing sales at roughly 3x nominal GDP.
* Jason and Mike frame AI risk as an engineering problem (human gating, cyber rather than physical threat) and read the labs’ public doom talk as fundraising PR and a push for regulatory capture against open-source models; Anthropic told a small group of shareholders it expects positive adjusted operating income (excluding stock-based compensation) for a second straight quarter, per the FT.
* Jason sees the main threat to Vertex’s CF franchise receding after Sionna Therapeutics’ SION-719 missed its Phase 2a endpoint as a Trikafta add-on (Sionna is still advancing a dual combination), while non-opioid pain uptake improves and the FDA let Vertex add a blood-type-O second-generation T1D cell therapy as a new arm in the existing trial.
Show Notes
[00:27] Oil: the diesel squeeze Why the East-West pipeline shutdown and Hormuz headlines matter less for crude than for diesel, now ~$200/bbl worldwide, and why backwardation keeps the strip ~$20 under spot.
[05:15] Gas settles at $3.50 Higher oil prices mean more Permian associated gas, while flat power demand keeps prices anchored for 2026 and 2027.
[07:01] The 10-year above 5% A unanimous Fed hike, $2 trillion deficits, and hyperscaler debt push rates higher, yet the S&P trades 19x and the Mag 7 just 21x.
[10:59] Data center capacity is worth more than book Mike argues existing capacity is undervalued as new builds get harder and demand keeps rising.
[12:01] Engineering, not extinction Jason on why language models are software, where the real cyber risk sits, and why the labs’ doom talk looks like PR and a push to regulate open source out.
[20:08] Tap water, bottled water NVIDIA’s Hugging Face acquisition, Jensen’s All-In analogy for cheap versus premium intelligence, and whether hosted Grok agents replace always-on Mac minis.
[23:00] Healthcare: Vertex Sionna Therapeutics’ Phase 2a miss eases the competitive threat to Vertex’s cystic fibrosis franchise, non-opioid pain uptake improves, and the FDA approves a new T1D trial arm.
[27:10] All-In notes SpaceX and xAI on hiring and turnover, AI-lab monthly revenue as a tracked signal, and Anthropic’s second straight quarter of positive adjusted operating income.
[29:47] Data center politics Meta’s Louisiana case study, and why candidates are running against data centers heading into November.
Download this week’s Cash Flow Memo at telltales.us and subscribe for next week’s episode on the data center jobs debate.
Cashtags
$AAPL $AMZN $GOOGL $META $MSFT $NVDA $VRTX
This post and the information herein are intended for informational purposes only. The views expressed herein are the author’s alone and do not constitute an offer to sell, or a recommendation to purchase, or a solicitation of an offer to buy, any security, nor a recommendation for any investment product or service. While certain information contained herein has been obtained from sources believed to be reliable, neither the author nor any of his employers or their affiliates have independently verified this information, and its accuracy and completeness cannot be guaranteed. Accordingly, no representation or warranty, express or implied, is made as to, and no reliance should be placed on, the fairness, accuracy, timeliness or completeness of this information. The author and all employers and their affiliated persons assume no liability for this information and no obligation to update the information or analysis contained herein in the future.
▶ Explore this week’s Tape — live, sortable, drill-down →
The Biggest AI Builder Just Chose Dilution
Four weeks ago the argument in this column was that the AI buildout had turned into a credit trade — that what prices the complex from here is whether somebody else’s lender says yes. Oracle answered the question this week, in a filing, and the answer was not a lender. It was the shareholders.
Start with the numbers everybody quoted. Oracle earned a record twenty-three billion dollars of operating cash in the quarter and spent twenty-eight and a half billion on capital expenditures, per the 10-Q filed Thursday¹². Free cash flow, in Oracle’s own release, negative five billion³. A record cash quarter that did not cover its own construction.
Now the line that ran in the same statement of cash flows and made no headlines. Oracle sold roughly twenty billion dollars of stock through an at-the-market program during those same three months⁴.
Four times the gap.
That ratio is the story, because it separates two completely different acts. Covering a five-billion-dollar shortfall with five billion of equity is housekeeping. Raising twenty against it is a decision about the quarters that have not happened yet — get the money in the door now, on these terms, before the terms move. Management then put a sentence in the release confirming that, given how the new contracts are structured, there is no incremental impact on its plans to raise capital⁵. Translation: there are plans to raise capital, and we would like everyone to have finished absorbing that before anybody asks it as a question.
The more interesting thing is the door Oracle did not use. The Cash Flow Memo already carries a hundred twenty-six billion dollars of net debt on this name⁶. You do not sell twenty billion dollars of your own stock into the best operating quarter in company history if long paper is cheap to you. Equity is the expensive money in every textbook and the patient money in every build — it cannot be called, it cannot be repriced, and it is indifferent to how long the conversion takes. Oracle bought the instrument that survives a slow cycle instead of the one that is cheaper in a fast one. That is a statement about duration, and it is being paid for in share count.
Which changes how the backlog reads. Six hundred sixty-four billion dollars of remaining performance obligations, up two hundred nine billion year over year⁷, against a full-year guide of at least ninety billion of revenue⁸ — the order book runs past seven years of sales at the pace the company is underwriting this year. Fewer as the pace rises, but the shape holds. A backlog that long is not only a revenue promise. It is a capital expenditure commitment. Seven years of contracted compute has to be bought before any of it can be collected, and the buying leads the collecting by years. The funding line, not the bookings line, is what decides how this ends.
The cashflow read is in Marcus’s column below — short version, the leaderboard is scoring cash and expectation as though they were the same asset.
Every forward-sold infrastructure cycle has broken on this question, and it has never been demand. Long-haul fiber in 1999 sold capacity years ahead of delivery, booked the contracts, and financed the trenching in the capital markets. The backlog was real. The capacity got built. What failed was the assumption that money stays available at one price for the length of a build. Oracle starts somewhere materially better — a profitable software business with real earnings sitting underneath the construction — but the dependency is identical, and it is now on public display in a single filing.
What changes the read. The tell on the next print is the funding line, not the revenue line. Bridge the following gap with more equity and this read holds: management is buying duration and paying in dilution, and every per-share figure Oracle guides to — the eight-dollar-ten non-GAAP EPS number included⁹ — sits on a denominator that is deliberately moving. Bridge it with debt against the hundred twenty-six billion already there¹⁰ and the arithmetic on the whole backlog changes, and the credit trade comes back on. The second thing to watch is conversion: eight hundred fifty megawatts of capacity delivered in the quarter, more than three hundred thousand GPUs since the end of May¹¹¹². Contracted revenue turns into cash at the rate the megawatts arrive. If the order book keeps climbing while the megawatts flatten, the spread between sold and deliverable is exactly where these cycles have always come apart.
Wall Street’s consensus on the AI build: the binding constraint is silicon. Oracle sold twenty billion dollars of its own stock this quarter to say the binding constraint is capital, and it is the only one of these companies that filed the receipt.
The Tape — W2637
Universe of 94 cashflow-memo names, snap dates 2026-09-04 → 2026-09-11. Composite is rank-sum percentile of FCF Yield + NTM Revenue Growth (higher = better balance). Banks and finance-book names shown separately.
Telltales Yield — Top 10
From the Cashflow Desk — Marcus Graham
The composite is flattening two different assets into one score this week. Most of the top ten earn their rank on cash yield — energy names and a couple of mature software businesses throwing off mid-to-high single digits. Nvidia earns its rank on growth alone: 68.2% NTM revenue against a 2.4% FCF yield, the only name in the top ten above 40x EV/FCF. Rank-sum scores them the same. They do not break the same way. The cash names break on a commodity price, which prints weekly. Nvidia breaks on an expectation, which does not print at all. And the DOJ inquiry into how the Groq license was structured is a question about deal structure, not about demand — worth separating before either gets read as a signal on the other. The test is whether the next print holds the growth the tape is already carrying.
Telltales Yield — Bottom 10
This Week’s Reporters
Sector Medians
Debt / FCF Watch (highest leverage on TTM FCF)
Weekly Price Movement
Top 5 (week-over-week price)
Bottom 5 (week-over-week price)
Banks (shown separately — FCF metric not meaningful)
Finance-book — FCF not comparable
Customer-float / captive-finance / reserve businesses (IBKR broker float, KMX CarMax Auto Finance, PYPL customer funds, CRCL stablecoin reserves). The memo’s operating-FCF method overstates their FCF, so they are held off the ranked leaderboard pending the P&L-waterfall rebuild.
Data Gaps
91 of 92 ranked-eligible names ranked. 1 dropped for missing FCF yield or NTM revenue growth; 7 shown separately (banks + finance-book, FCF not comparable).
Source: cashflow-memo master_2026-09-11.csv. NTM growth from analyst-estimates consensus. Composite is a percentile rank, not a recommendation.
The Issue — This Week's Brief
The Cashflow Memo
Oracle’s Record Cash Quarter Didn’t Cover Its Own Build
Oracle earned a record $23 billion of cash this quarter, spent $28.5 billion, and raised guidance anyway. The companies actually generating cash spent the week getting sued.
The Telltales Weekend Update. Ava Cabot and analyst Marcus Graham walk through what happened this week — and what’s coming next — across the 86 companies in the Cash Flow Memo. About 14 minutes. No filler.
Download the memo at telltales.us. Hunt, Jason, and Mike are back Wednesday on episode E2638.
Chapter markers
* Time | Segment
* 0:00 | Cold open
* 0:45 | Theme — Healthcare’s two halves
* 4:45 | Deep dive — Oracle
* 8:45 | Rapid-fire
* 11:45 | Close
* 12:40 | Closing disclaimer
Full transcript
Cold open
Ava: Oracle earned a record $23 billion of cash in a single quarter this week[^news-orcl-ocf-20260911]. It spent $28.5 billion in the same three months[^news-orcl-capex-20260911]. And then it raised the full-year revenue guide[^news-orcl-fy27guide-20260910]. That is the week in three numbers. The AI build is now big enough that a record cash quarter doesn’t cover it. Meanwhile four healthcare names in the Cash Flow Memo spent the week getting sued by hospitals, staring down a September 30 deadline in New York, and buying pipelines they can’t price yet. Nobody applauded any of that.
Ava: Telltales Weekend Update. I’m Ava Cabot, with Marcus Graham at the cashflow desk.
Theme — Healthcare’s two halves
Ava: Healthcare split clean down the middle this week, and the two halves did opposite things with money. One half went shopping. The other half went to war with its own providers.
Ava: Start with Vertex, because it did the biggest thing. Vertex closed the largest acquisition in the company’s history — Crinetics Pharmaceuticals, $8.8 billion, done on September 5[^news-vrtx-crinetics-20260905]. Three days later it published data on zimislecel, an islet cell therapy that restored patients’ own insulin secretion in Type 1 diabetes in the Phase 1/2 FORWARD trial[^news-vrtx-zimislecel-20260908]. Restored their own insulin production. Goldman put the stock on its Conviction List the day after that, with a $653 price target[^news-vrtx-gs-20260909].
Ava: On page 19, Eli Lilly closed one too. AtaiBeckley, $6.75 a share, roughly $2.8 billion, mental-health therapies, Nasdaq delisting initiated immediately[^news-lly-atai-20260911]. And the more interesting Lilly development came out of a competitor’s failure — Novartis missed on its lipoprotein-A cholesterol program, which raises the stakes on the competing shots Lilly and Amgen have at that same market[^news-lly-lpa-20260908]. Marcus — is the memo paying for any of this yet?
Marcus: Both of these companies are buying revenue that doesn’t exist yet, and the memo is already paying full price for it. The memo has Vertex at 33x free cash flow, on $4 billion of it[^memo-vrtx-evfcf-20260911][^memo-vrtx-fcf-20260911]. Lilly’s at 50x[^memo-lly-evfcf-20260911]. Those aren’t cheap numbers for companies whose next leg of growth is sitting in a clinical trial. What I’d watch on Vertex is conversion. A closed acquisition and a mid-stage readout are both promises, and the memo is pricing them like they already cleared.
Ava: Now the other half of that page, where nobody is buying anything. CVS is getting sued by more hospitals. Safety-net providers allege the company withheld 340B drug payments — that’s the federal program that forces drugmakers to sell discounted drugs to hospitals serving low-income patients[^news-cvs-340b-20260910]. Same week, CVS launched a joint product with Cigna called Cigna Health Works, routing Cigna-administered benefits into CVS pharmacies and MinuteClinics[^news-cvs-cigna-20260911]. Sued by its providers and partnered with a rival insurer, inside seven days.
Ava: UNH, same page, has a deadline. NewYork-Presbyterian, the largest hospital system in New York, goes out of network for most UnitedHealthcare commercial plans on September 30 unless the two sides reach a deal[^news-unh-nyp-20260910]. And UnitedHealth sold an ownership interest in its Florida WellMed clinics to the private equity firm TPG[^news-unh-wellmed-20260911]. Selling the clinics on one coast, fighting over the hospitals on the other.
Marcus: Managed care is the only part of tonight’s show where the price actually reflects the fight. The memo has CVS at 15x free cash flow on $12 billion of it[^memo-cvs-evfcf-20260911][^memo-cvs-fcf-20260911], and UnitedHealth at 22x[^memo-unh-evfcf-20260911]. Those are the lowest multiples on anything that gets a beat tonight, and the reason is sitting right there in the news. When your providers are suing you and the largest hospital system in your biggest market is weeks from walking, the market discounts the cash flow instead of extrapolating it. That’s the honest version of cheap. What settles it is the network contract, not the multiple.
Ava: Marcus, put the two halves next to each other.
Marcus: The spread is the point. One half of this sector is paying between 33x and 50x free cash flow for assets that are still in trials[^memo-vrtx-evfcf-20260911][^memo-lly-evfcf-20260911]. The other half is being priced in the teens on cash it already collects[^memo-cvs-evfcf-20260911]. That gap isn’t a mistake, and it isn’t an opportunity by itself either. It’s the market saying it will pay up for optionality and demand a discount for a fight. Whether that’s the right trade gets settled by two things this autumn: whether the Vertex and Lilly pipelines convert, and whether managed care renews its contracts without giving away margin.
Ava: Four healthcare names, four completely different problems, and not one of them needs a data center. Which brings us to the company that does.
Deep dive — Oracle
Ava: Oracle reported Thursday, and it published the most useful document anybody put out this week — because for once you can read the order book and the bill in the same filing. Total revenue up 30% to $19.3 billion[^news-orcl-revenue-20260910]. Cloud infrastructure up 121%[^news-orcl-iaas-20260910]. Remaining performance obligations — contracted revenue Oracle hasn’t delivered yet — up $209 billion year over year, to $664 billion[^news-orcl-rpo-20260910]. More than $30 billion of new AI cloud contracts booked in the quarter alone[^news-orcl-bookings-20260910]. And the full-year guide went to at least $90 billion of revenue[^news-orcl-fy27guide-20260910]. Marcus — what does an order book that size cost to carry?
Marcus: Oracle just published the price of an AI backlog, and it’s the cleanest version of that number we’ve had all year. Record operating cash flow for the quarter, $23 billion[^news-orcl-ocf-20260911]. Capital expenditures in the same three months, $28.5 billion[^news-orcl-capex-20260911]. A record cash quarter that still didn’t cover the build. Oracle’s own release puts free cash flow at negative $5 billion for the quarter[^news-orcl-fcf-20260910].
Ava: Negative $5 billion. On a record.
Marcus: The part that didn’t make the headlines is how they funded the gap. Oracle sold stock — $19.9 billion of net proceeds from an at-the-market equity program during the quarter[^news-orcl-atm-20260911]. The memo already carries $126 billion of net debt on this name[^memo-orcl-netdebt-20260911], so the bond market wasn’t the obvious door. And management put a sentence in the release saying the new contracts create no incremental impact on its plans to raise capital[^news-orcl-capital-20260910]. I’d take that as getting in front of the question before anyone had to ask it.
Ava: They pre-answered it. And two other things came out of the same filings that the headlines skipped. Oracle put 850 MW of additional data center capacity into service in the quarter[^news-orcl-capacity-20260910], and delivered more than 300,000 GPUs since the end of May, almost triple the prior quarter’s pace[^news-orcl-gpus-20260910]. Software revenue, the old business, went down 3%[^news-orcl-software-20260910]. And in the 10-Q, the restructuring plan Oracle put up to $2.1 billion behind as of August 31 got supplemented by roughly $700 million more after the quarter closed[^news-orcl-restructuring-20260911]. The filing attributes that plan to operational efficiencies, including adopting AI across certain functions. So the AI is doing the restructuring too.
Marcus: The most useful thing the memo does on this name is refuse to give you a multiple. Oracle’s in the memo at negative $32 billion of free cash flow[^memo-orcl-fcf-20260911]. When the denominator is negative, a multiple isn’t a number, it’s a category error — so valuation isn’t the frame on this name at all. Delivery is. 850 MW and 300,000 GPUs in a quarter is the conversion rate on the order book, and conversion is the only thing that turns contracted revenue into cash.
Ava: So what breaks it?
Marcus: Not the demand. The order book is real and it’s growing — $209 billion added year over year[^news-orcl-rpo-20260910]. What isn’t settled is whether the capital keeps arriving on these terms. They funded this quarter with equity[^news-orcl-atm-20260911]. The same gap funded with debt, against the net debt the memo already carries here[^memo-orcl-netdebt-20260911], is a materially worse outcome for the people who own it. I’d weight the demand question as largely answered and the funding question as wide open, and it’s the funding line I’d go to first on the next print.
Ava: So the loudest number in the market this week was $664 billion of contracted revenue[^news-orcl-rpo-20260910]. And the number that decides whether any of it is worth something is megawatts. One of those was in every headline. The other one you had to go find.
Rapid-fire
Ava: Lightning round.
Ava: Bloom Energy joins the S&P 500 on September 21[^news-be-sp500-20260904], and that’s the good news. Here’s the rest of it. Bloom is now trading more options than SpaceX, with implied volatility above 90% — the highest of any energy name in the index it’s about to join[^news-be-options-20260909]. And multiple securities class actions landed this week alleging Bloom failed to disclose that its scandium, a rare metal in its fuel cells, was being sourced through intermediaries from China[^news-be-lawsuit-20260909]. The memo carries Bloom at $200 million of free cash flow against a $60 billion enterprise value[^memo-be-fcf-20260911][^memo-be-ev-20260911]. Index inclusion changes who owns it. It does not change that arithmetic.
Ava: Amazon and Qualcomm signed a multi-generation AI chip partnership this week, under which Amazon could buy up to $60 billion of Qualcomm’s data center chips[^news-amzn-qualcomm-20260908]. Up to. That’s a ceiling, not a commitment, and it’s worth holding next to what Amazon is already spending — the memo has Amazon at negative $1 billion of free cash flow[^memo-amzn-fcf-20260911], and the dataset behind it puts trailing capex at $173 billion[^data-amzn-capex-20260911]. Also this week, Kevin Mandia, who founded the security firm Mandiant, joined Amazon’s board[^news-amzn-mandia-20260909].
Ava: Page 3 of the memo had the busiest week of anybody. Taiwan Semi posted record August revenue of about $16 billion, up 53% year over year, and still can’t keep pace with AI chip demand[^news-tsm-august-20260910]. It also signed up with ASML, Samsung and Intel behind a 12-inch photomask line for the next generation of lithography[^news-tsm-asml-20260908]. A record month, described by the company’s own run rate as insufficient.
Ava: Intel, same page. Intel is reportedly preparing to raise CPU prices by about 10% ahead of a major product launch next March, on weak supply and strong demand[^news-intc-cpuprice-20260908]. Reportedly — but a price increase is not a thing a struggling chipmaker floats casually. The stock gave back 5.6% this week to a $100 close, after a five-session run to its highest level in more than two years[^news-intc-stock-20260911].
Ava: Nvidia, same page. The Justice Department is investigating whether Nvidia structured its $20 billion licensing deal with the AI chip startup Groq specifically to avoid antitrust review[^news-nvda-groq-20260910]. That’s an inquiry into deal structure, not into chips, and those are very different problems to have.
Ava: Forward week, and it is thin. Page 17 carries both names — Lennar reports Wednesday[^earn-len], FedEx Thursday[^earn-fdx]. Costco the Thursday after that[^earn-cost]. Three reports in two weeks. And the memo has FedEx at 11x free cash flow going in[^memo-fdx-evfcf-20260911], which is the cheapest thing on the board tonight.
Close
Ava: That’s the show. Oracle earned a record $23 billion of cash this quarter and spent $28.5 billion building the thing that’s supposed to earn the next one[^news-orcl-ocf-20260911][^news-orcl-capex-20260911]. The healthcare names that already generate the cash spent the week in litigation and contract standoffs. The market is paying up for the order book and marking down the cash flow, and both of those are decisions, not facts. Wall Street’s consensus on Oracle this week: the $664 billion order book is the story. It is. It’s also in the same filing as negative $5 billion of free cash flow[^news-orcl-fcf-20260910].
Ava: On Wednesday’s show, Hunt closed the technology block by saying they’d leave Nvidia for next week[^ep-e2637]. The Justice Department has since given them something to open with. Hunt, Jason, and Mike are back Wednesday on episode 2638.
Ava: Download the memo at telltales.us, and send us a note through the Substack — we read them.
Ava: The show is produced entirely with AI tools, and both voices you’re hearing are AI-generated.
Closing disclaimer
Ava: This podcast is intended for informational purposes only. You should always do your own work to determine if an investment is suitable for you. The views expressed on this podcast are the host alone and do not constitute an offer to sell or a recommendation to purchase, or a solicitation of an offer to buy any security, nor a recommendation for any investment product or service. While certain information contained herein has been obtained from sources believed to be reliable, neither the host nor any of their employers or their affiliates have independently verified this information, and its accuracy and completeness cannot be guaranteed. Accordingly, no representation or warranty, express or implied, is made as to, and no reliance should be placed on, the fairness, accuracy, timeliness, or completeness of this information. The host and all employers and their affiliated persons assume no liability for this information and no obligation to update the information or analysis contained herein in the future, and may or may not hold positions in the securities mentioned.
Sources
* Amazon. (2026, September 9). Cybersecurity expert Kevin Mandia joins Amazon’s Board of Directors [Press release]. Amazon Newsroom. https://www.aboutamazon.com/news/company-news/kevin-mandia-amazon-board-of-directors
* AOL. (2026, September 10). Taiwan Semiconductor September revenue surges 53% and can’t keep up with demand. AOL. https://www.aol.com/articles/taiwan-semiconductor-september-revenue-surges-151224000.html
* Bloomberg. (2026, September 4). Bloom Energy, Illumina, Everpure to join S&P 500 in September. Bloomberg. https://www.bloomberg.com/news/articles/2026-09-04/bloom-energy-illumina-everpure-to-join-s-p-500-in-september
* Bloomberg. (2026, September 8). ASML, TSMC, Samsung, Intel back 12-inch masks for AI chips. Bloomberg. https://www.bloomberg.com/news/articles/2026-09-08/asml-tsmc-samsung-intel-back-12-inch-masks-for-ai-chips
* Bloomberg. (2026, September 10). DOJ probes Nvidia’s $20 billion license deal with Groq on antitrust concerns. Bloomberg. https://www.bloomberg.com/news/articles/2026-09-10/doj-probes-nvidia-s-license-deal-with-groq-on-antitrust-concerns
* The Boston Globe. (2026, September 5). Vertex makes some big people moves, completes its largest acquisition. The Boston Globe. https://www.bostonglobe.com/2026/09/05/business/vertex-pharma-crinetics-executive-moves/
* CGTlive. (2026, September 8). Vertex’s islet cell therapy zimislecel (VX-880) restores endogenous insulin secretion in Type 1 diabetes. CGTlive. https://www.cgtlive.com/view/vertex-islet-cell-therapy-vx-880-restores-endogenous-insulin-secretion-type-1-diabetes
* CNBC. (2026, September 8). Novartis trial failure raises stakes for Amgen and Eli Lilly in Lp(a) drug race. CNBC. https://www.cnbc.com/2026/09/08/novartis-cholesterol-setback-drug-race-eli-lilly-amgen.html
* CNBC. (2026, September 9). Bloom Energy is trading more options than SpaceX. CNBC. https://www.cnbc.com/2026/09/08/bloom-energy-is-trading-more-options-than-spacex.html
* Healthcare Finance News. (2026, September 11). UnitedHealth sells stake in Florida WellMed clinics. Healthcare Finance News. https://www.healthcarefinancenews.com/news/unitedhealth-sells-stake-florida-wellmed-clinics
* Investing.com. (2026, September 11). Eli Lilly completes acquisition of AtaiBeckley, initiates delisting from Nasdaq. Investing.com. https://www.investing.com/news/sec-filings/eli-lilly-completes-acquisition-of-ataibeckley-initiates-delisting-from-nasdaq-93CH-4898037
* Jon Peddie Research. (2026, September 11). Amazon bets $60 billion on Qualcomm. Jon Peddie Research. https://www.jonpeddie.com/news/amazon-bets-60-billion-on-qualcomm/
* Live Insurance News. (2026, September 10). New York’s largest hospital and UnitedHealthcare have 20 days to make a deal. Live Insurance News. https://www.liveinsurancenews.com/new-york-hospital-unitedhealth/8576058/
* Modern Healthcare. (2026, September 10). CVS Health faces more hospital 340B payment lawsuits. Modern Healthcare. https://www.modernhealthcare.com/providers/mh-cvs-health-hospital-340b-payment-lawsuits/
* NAI 500. (2026, September 9). Goldman Sachs adds Vertex to Conviction List with $653 target. NAI 500. https://nai500.com/blog/2026/09/goldman-sachs-adds-vertex-to-conviction-list-with-653-target/
* Newsfile Corp. (2026, September 9). Robbins LLP reminds investors that a securities class action was filed against Bloom Energy Corporation after the company revealed how much it relied on scandium from China [Press release]. Newsfile Corp. https://www.newsfilecorp.com/release/313407/Robbins-LLP-Reminds-Investors-That-a-Securities-Class-Action-was-Filed-Against-Bloom-Energy-Corporation-After-the-Company-Revealed-How-Much-it-Relied-on-Scandium-from-China
* Oracle Corporation. (2026, September 10). Oracle announces Q1 results driven by triple digit growth in cloud infrastructure revenues [Press release, Form 8-K Exhibit 99.1]. U.S. Securities and Exchange Commission. https://www.sec.gov/Archives/edgar/data/1341439/000119312526387905/orcl-ex99_1.htm
* Oracle Corporation. (2026, September 11). Form 10-Q for the quarterly period ended August 31, 2026 [10-Q]. U.S. Securities and Exchange Commission. https://www.sec.gov/Archives/edgar/data/1341439/000119312526389274/orcl-20260831.htm
* Pharmacy Times. (2026, September 11). Cigna and CVS Health launch Cigna Health Works, a new model for customer-enabled design for affordable, convenient access to pharmacy and health care. Pharmacy Times. https://www.pharmacytimes.com/view/cigna-and-cvs-health-launch-cigna-health-works-a-new-model-for-customerenabled-design-for-affordable-convenient-access-to-pharmacy-and-health-care
* Tom’s Hardware. (2026, September 8). Intel reportedly set to hike CPU prices by 10% ahead of major annual product launch in March 2027. Tom’s Hardware. https://www.tomshardware.com/pc-components/cpus/intel-reportedly-set-to-hike-cpu-prices-by-10-percent-ahead-of-major-annual-product-launch-in-march-2027-report-says-amd-will-follow-up-between-june-and-july
* TS2 Tech. (2026, September 11). Intel stock drops 5.6% — why $100 is the real test after its $20 billion raise. TS2 Tech. https://ts2.tech/en/intel-stock-drops-5-6-why-100-is-the-real-test-after-its-20-billion-raise/
Primary-source note. The episode’s load-bearing claim — that Oracle’s record quarterly operating cash flow did not cover its quarterly capital spending, and that the gap was funded with equity — traces to Oracle’s own Form 8-K Exhibit 99.1 (reference 17) and Form 10-Q (reference 18), not to secondary coverage. Oracle’s investor-relations site was unreachable to an automated fetch; both documents were retrieved from SEC EDGAR.
Internal data
Internal data is provided on a best efforts basis.
Hunt, Mike, and Jason walk the Cash Flow Memo across an escalating Iran-US oil standoff, the widening federal deficit, a wave of new AI agent products from xAI and Meta, and a Montana law reshaping the economics of cancer immunotherapy.
The Cashflow Memo
Key Takeaways
* Iran-US tensions escalated in the Persian Gulf (US Navy sinking Iranian shadow fleet tankers after a failed missile strike on the USS Abraham Lincoln) and pushed Brent over $100 intraday, but Hunt reads the ~$20 12-month backwardation as the market pricing de-escalation and holds his call that oil isn’t sustainable much above $90.
* The $2 trillion US deficit and the $15 trillion US repo market (matched by ~$15 trillion in Europe) are the macro risk to watch: Hunt wants a credible 24-month path from Treasury Secretary Bessent to cut the deficit to $1 trillion or expects repo-market stress reminiscent of 2008’s Bear Stearns/JPMorgan rescue; Jason sees no political will to act absent a crisis.
* Grokbot (xAI) is emerging as the household/enterprise AI-agent leader on $20-$300/month subscription tiers, with Meta building a comparable agent monetized through frictionless in-app purchases (e.g. Instagram) rather than subscriptions; Mike frames AI as a sustaining innovation for Microsoft’s knowledge-work franchise, while Jason argues Apple risks just selling screens if it doesn’t ship an equivalent iOS agent.
* Montana’s expanded right to try law is opening a roughly $50,000 cash-pay alternative to Moderna’s roughly $500,000 personalized mRNA cancer immunotherapy, riding now-cheap DNA sequencing and AI-driven mutation targeting; Mike and Jason see no patent-infringement exposure since the underlying science is federally funded and public, and flag insurance reimbursement — not the FDA pathway — as the real gate on Moderna’s pricing power.
* China’s share of global pharma licensing deals jumped to 42% in H1 2026 (from ~20% in 2023, near zero before that), with a Chinese-developed bispecific antibody now showing head-to-head superiority over Merck’s Keytruda on the PD-1 pathway; Jason and Mike flag drug development running 3x faster and 30-50% cheaper in China as a structural threat to US/Europe biotech economics.
Show Notes
[00:00:27] Oil: Why $90 Still Holds Iran’s failed missile strike on the USS Abraham Lincoln and the US Navy’s response — sinking Iran’s shadow fleet tankers — pushed Brent over $100 intraday, but Hunt reads the ~$20 12-month backwardation as the market betting this de-escalates rather than a new sustained regime.
[00:04:27] Gas: Riding the Permian’s Coattails Higher oil prices mean more Permian drilling and more associated gas supply, a headwind Hunt is weighing against otherwise supportive gas fundamentals.
[00:04:47] Deficit: A $15 Trillion Repo Market Under Strain The $2 trillion federal deficit and a $15 trillion US repo market (matched by roughly $15 trillion in Europe) are where Hunt expects the next crack to show, unless Treasury Secretary Bessent delivers a credible 24-month path to cut the deficit in half; Jason doubts the political will exists absent a crisis.
[00:08:47] A New Framework: Two Years Back, Three Years Forward Hunt introduces his personal method for underwriting a five-year plan — two years of real backward-looking analysis plus three years forward — starting with SpaceX’s capital-intensive, data-center-driven growth, and previews Amazon/Google-style capital spenders against Apple/Nvidia-style cash generators.
[00:11:23] Agents: Grokbot Goes Mainstream xAI’s Grokbot is already automating Mike’s household admin (registering appliances, renewing car registration, booking hotels) on $20-to-$300-a-month tiers, with Meta building a comparable agent monetized through frictionless in-app purchases rather than subscriptions.
[00:17:50] Microsoft and Apple: Sustaining Innovation vs. Just Selling Screens Mike argues AI is a sustaining innovation for Microsoft’s knowledge-work franchise if it ships a good enterprise agent; Jason warns Apple risks becoming a screen-seller if it doesn’t integrate an equivalent agent into iOS, even after Apple’s same-day Apple Intelligence event.
[00:21:02] Montana: A $50,000 Alternative to a $500,000 Cancer Treatment Montana’s expanded right to try law lets clinics offer personalized mRNA cancer immunotherapy for roughly $50,000, versus Moderna’s roughly $500,000 approved treatment, using now-cheap DNA sequencing and AI-driven mutation targeting; Mike and Jason see no patent issue since the underlying science is public and federally funded, with insurance reimbursement the real constraint on Moderna’s pricing.
[00:28:14] China’s Pharma Surge China’s share of global pharma licensing deals hit 42% in H1 2026 (up from ~20% in 2023 and near zero before that), including a Chinese-developed antibody now beating Merck’s Keytruda head-to-head, as China develops drugs roughly 3x faster and 30-50% cheaper than the US and Europe.
Download this week’s Cash Flow Memo at telltales.us and join us next Wednesday for more on Nvidia’s frontier-model ambitions and China’s pharma push.
Cashtags
$AAPL $AMZN $GOOGL $JPM $META $MRK $MRNA $MSFT $NVDA
This post and the information herein are intended for informational purposes only. The views expressed herein are the author’s alone and do not constitute an offer to sell, or a recommendation to purchase, or a solicitation of an offer to buy, any security, nor a recommendation for any investment product or service. While certain information contained herein has been obtained from sources believed to be reliable, neither the author nor any of his employers or their affiliates have independently verified this information, and its accuracy and completeness cannot be guaranteed. Accordingly, no representation or warranty, express or implied, is made as to, and no reliance should be placed on, the fairness, accuracy, timeliness or completeness of this information. The author and all employers and their affiliated persons assume no liability for this information and no obligation to update the information or analysis contained herein in the future.
▶ Explore this week’s Tape — live, sortable, drill-down →
ONEOK Bought a Ratio
ONEOK announced a four-point-four-billion-dollar acquisition last weekend and raised nine billion dollars to pay for it. The gap between those two numbers is the story, and it is not what the headlines carried. Four-point-four billion buys Brazos Midstream’s Permian gathering and processing assets. Five billion retires debt¹. Page ten of the Cash Flow Memo carries ONEOK at about thirty-three billion dollars of net debt², so five billion is a real dent. The interesting part is which five billion, and what replaced it.
Start with what Apollo actually bought, because nine-billion-dollar minority equity investment conceals more than it reports. Apollo gets a Class B interest in a newly formed holding company, ONEOK Holdings, L.L.C. It is nonvoting. No board seat, no liquidation preference, structurally subordinate to every existing ONEOK bond. It is entitled to fifteen percent of the operating company’s quarterly cash flow, and its total return is capped at a seven percent internal rate of return for the first nine years. ONEOK can buy the whole thing back beginning on the eighth anniversary of closing, at that same capped return³. Apollo took a bond’s return while standing in an equity’s place in the stack.
Now what ONEOK bought. One sentence in the release explains the other twelve: the investment has been reviewed with ONEOK’s credit rating agencies, all of which consider the transaction as credit-enhancing, and ONEOK expects to receive full equity credit⁴. Every concession in the paragraph above is priced to earn that clause. This instrument was assembled by people who knew precisely which boxes a rating methodology checks. Pro forma 2027 leverage lands near three and a quarter times debt to EBITDA, and not one common share was issued⁵. Apollo, in the same release, called it flexible, high-grade capital solutions at scale, structured around ONEOK’s long-term strategic objectives. Translation: we found the boxes and built something that ticks all of them.
Then the tender offer, filed the same day and buried under the acquisition headline. ONEOK is offering to repurchase up to two billion dollars of its own notes across twenty series, and the acceptance priority order tells you what it wants back. Level one is the three-point-nine-five percent notes due 2050. Level two, four-point-two percent due 2047. The list runs long-dated and low-coupon at the top, and the company adds, in a parenthetical, that most of the targeted notes currently trade below par⁶ — below face value, because those coupons sit under what the market charges today. So ONEOK is retiring some of the lowest cash coupons on its balance sheet and funding the gap with money that costs seven percent. The cash cost of the capital went up. The ratio the agencies quote went down. Both by design.
None of that makes it a bad trade. Seven percent capped is genuinely cheaper than ONEOK’s cost of common equity, every dollar of value above the cap accrues to the shareholders already there, the capital account amortizes as distributions run past the cap, and the call option hands the whole claim back inside a decade. Against issuing stock or thirty-year paper at today’s yields, it is the cheapest of the three doors. It is a well-built instrument. What it is not is a reduction in the claims on ONEOK’s cash. The leverage did not leave the company. It moved to a line the leverage ratio does not read.
Midstream has built this shape before, and ONEOK has built it twice. On June 30, 2017, the company acquired every ONEOK Partners unit it did not already own⁷, collapsing a structurally separate claim on the same operating cash flow back into the common, because by then that claim had become the expensive way to fund a build. Nine years later there is a new structurally separate claim on the same operating cash flow, with the buy-back written into the document on day one. Call that a cycle rather than a criticism. Funding structures get cleverest at exactly the point where assets are dearest and internal cash flow is furthest from covering the plan. Oracle reports Thursday running the same trade with none of the engineering. The cashflow read is in Marcus’s column below; short version, the cash line on that page is the least useful number on it.
What changes the read. The near test has a clock on it. The early tender deadline is 5 p.m. New York time on September 14, with the offers expiring September 29⁸. Watch how much of the two-billion-dollar cap fills and where those 2050s clear. Fill it at a discount and the five-billion extinguishment is cheap and quick. Fall short and ONEOK finishes the job through make-whole calls, which is the expensive door. After that: Hart-Scott-Rodino clearance and a fourth-quarter close on Brazos, then the first print carrying a noncontrolling-interest line where none existed. The read breaks if the agencies deliver anything less than the full equity credit the company says it expects. Every number in this structure is priced off that one word.
Wall Street’s consensus on ONEOK: a balance-sheet repair with a growth acquisition attached, and Wells Fargo took its target to a hundred and six dollars on the EBITDA outlook⁹. The Brazos EBITDA is real, bought at roughly seven and a half times the assets’ estimated 2027 number, with fourteen rigs running on the acreage¹⁰. The deleveraging is a ratio.
The Tape — W2636
Universe of 94 cashflow-memo names, snap dates 2026-09-02 → 2026-09-04. Composite is rank-sum percentile of FCF Yield + NTM Revenue Growth (higher = better balance). Banks and finance-book names shown separately.
Telltales Yield — Top 10
From the Cashflow Desk — Marcus Graham
Oracle is the only name in this universe reporting before the next issue, and the leaderboard cannot price it, because the cash leg is negative by construction. Capex ran $55.7B TTM against $67.4B of revenue. Eighty-three cents of every revenue dollar, into the build. That is the cost of the build, not the flag the screens read it as, and no multiple belongs on a denominator management is deliberately spending through. What prices Oracle right now is conversion, and 45.6% NTM revenue growth is the market underwriting a backlog it has not yet seen turn into cash. Reading the negative cash line as deterioration is the screen standing in for the work. The test on the September 10 print is whether the capex guide steps up again alongside the revenue guide, or capex flattens while revenue accelerates. Only the second one shortens the payback.
Telltales Yield — Bottom 10
This Week’s Reporters
Sector Medians
Debt / FCF Watch (highest leverage on TTM FCF)
Weekly Price Movement
Top 5 (week-over-week price)
Bottom 5 (week-over-week price)
Banks (shown separately — FCF metric not meaningful)
Finance-book — FCF not comparable
Customer-float / captive-finance / reserve businesses (IBKR broker float, KMX CarMax Auto Finance, PYPL customer funds, CRCL stablecoin reserves). The memo’s operating-FCF method overstates their FCF, so they are held off the ranked leaderboard pending the P&L-waterfall rebuild.
Data Gaps
91 of 92 ranked-eligible names ranked. 1 dropped for missing FCF yield or NTM revenue growth; 7 shown separately (banks + finance-book, FCF not comparable).
Source: cashflow-memo master_2026-09-04.csv. NTM growth from analyst-estimates consensus. Composite is a percentile rank, not a recommendation.
The Issue — This Week's Brief
The Cashflow Memo
A Quiet Earnings Week Is When You Find Out What Management Believes
Three new CEOs, four energy checks, and two guides the market read in opposite directions
The Telltales Weekend Update. Ava Cabot and analyst Marcus Graham walk through what happened this week — and what’s coming next — across the 86 companies in the Cash Flow Memo. About 13 minutes. No filler.
Download the memo at telltales.us. Mike, Jason, and Hunt are back Wednesday on episode E2637.
Chapter markers
* Time | Segment
* 0:00 | Cold open
* 0:45 | Theme — Energy writes the checks
* 4:45 | Deep dive — Broadcom vs. Snowflake
* 8:45 | Rapid-fire
* 11:45 | Close
* 12:40 | Disclaimer
Full transcript
Cold open
Ava: A quiet earnings week is when you find out what management actually believes. Only three companies in the Cash Flow Memo report over the next two weeks[^earn-orcl][^earn-len][^earn-fdx]. So this week didn’t hand us prints. It handed us decisions. Apple replaced Tim Cook[^news-aapl-ceo-20260901]. Albemarle went outside the company for its next chief executive[^news-alb-ceo-20260903]. Four energy companies committed serious capital, including a merger that would create the largest regulated electric utility in the country[^news-nee-merger-20260904]. And two enterprise software names reported on the very same Wednesday and got read in completely opposite directions. Nobody had to guess what these managements think this week. They showed you.
Ava: Telltales Weekend Update. I’m Ava Cabot, with Marcus Graham at the cashflow desk.
Theme — Energy writes the checks
Ava: Start with the loudest thing that happened in a quiet week. Four energy companies, across four completely different parts of the business, all committed capital inside seven days. Power, upstream, midstream, and LNG. Nobody was reporting. Everybody was spending.
Ava: NextEra first. Shareholders on both sides approved the $66.8 billion all-stock merger with Dominion Energy — a deal that would create the largest regulated electric utility in the United States[^news-nee-merger-20260904]. And in the same week, NextEra Energy Resources committed to a $13 billion natural gas complex in Fayette County, Pennsylvania[^news-nee-gasplant-20260901]. Marcus, the cashflow take.
Marcus: Notice what NextEra is paying with. It’s an all-stock merger[^news-nee-merger-20260904] — they’re issuing paper, not writing a check, and then separately committing $13 billion of cash to a gas plant[^news-nee-gasplant-20260901]. That combination tells you something the press release doesn’t. The memo already carries $82 billion of net debt on this name[^memo-nee-netdebt-20260904]. NextEra is mid-build, free cash flow is running negative, and the multiple isn’t the right frame here — what prices this company is whether the regulated rate base absorbs the spend. Stock is the currency you use when your own cash flow can’t fund the plan and you’d rather not add to the debt load. I’d read the all-stock structure as the honest signal in that announcement.
Ava: On page 9 of the memo, the story is the opposite. Chevron is committing more than $7 billion over five years to more than double its Venezuelan production[^news-cvx-venezuela7b-20260902], after getting assigned additional Orinoco Belt acreage[^news-cvx-venezuela-20260903]. Mike Wirth’s framing, and I want you to sit with the phrasing: patience pays off, after two decades of staying while competitors exited following the nationalizations[^news-cvx-wirth-20260902]. Twenty years of waiting out an expropriation, filed under patience. Piper Sandler took its target to a street-high $243 on the back of oil prices[^news-cvx-piper-20260903].
Marcus: Chevron is buying volume because price isn’t going to do the work. The memo has Chevron at 16x free cash flow on $26 billion of trailing free cash flow[^memo-cvx-evfcf-20260904][^memo-cvx-fcf-20260904] — which is not a demanding number for an integrated. But the forward revenue line in the same dataset has Chevron shrinking roughly 9% over the next twelve months[^data-cvx-ntmgrowth-20260904]. That’s the context for Venezuela. When the top line is going backwards on price and mix, you either accept the decline or you go buy barrels. Wirth went and bought barrels, in the one basin he never left. What I’d watch is whether that capital shows up as production or as a write-down, because the history of that basin is not kind.
Ava: Midstream, same week. ONEOK is buying Brazos Midstream’s Permian gathering and processing assets for $4.4 billion in cash[^news-oke-brazos-20260901]. And here’s the part worth slowing down for — Apollo is putting up $9 billion, as a minority equity investment in ONEOK, to fund it[^news-oke-apollo-20260830]. That is not a loan. Somebody else is buying a piece of the company so the company can buy the assets.
Marcus: That structure is doing real work, and it’s the most interesting capital decision of the week. The memo carries ONEOK at $33.6 billion of net debt against $3.4 billion of free cash flow[^memo-oke-netdebt-20260904][^memo-oke-fcf-20260904]. Call it 10x net debt to free cash flow. At that level of leverage, the bond market is not where you go to fund a Permian acquisition, so ONEOK sold a piece of itself to a private-markets buyer instead. It’s a rational trade and it dilutes the people who already own it. Both of those are true at the same time, and the release only tells you the first one.
Ava: And then there’s Cheniere, which is the cleanest of the four and got the least attention. Corpus Christi Stage Three reached substantial completion, the 5,000th cargo went out of the Gulf Coast, and management took the 2026 EBITDA guide up to a range of $7.9-8.4 billion, from $7.25-7.75 billion[^news-lng-ccl3-20260831][^news-lng-ebitda-20260831]. Equinor took its first cargo under the long-term agreements the same week[^news-lng-equinor-20260903]. Cheniere is on page 9 as well, at 17x free cash flow[^memo-lng-evfcf-20260904]. No merger, no arbitration claim, no private-equity partner. They finished the thing they said they’d finish and raised the number.
Ava: Four companies, four balance sheets, four completely different answers to the same question. One issued stock, one bought barrels, one sold a piece of itself, and one just built the plant. And not a single one of them reported earnings this week.
Deep dive — Broadcom vs. Snowflake
Ava: Now the part that should bother you. Wednesday, September 2. Two companies on page 2 of the Cash Flow Memo both reported, both beat, and the market treated them like they were in different industries on different continents.
Ava: Broadcom: third-quarter adjusted earnings of $3.32 a share, ahead of expectations[^news-avgo-q3eps-20260902]. Then the fourth-quarter revenue guide came in at $34.8 billion against a consensus of $35.03 billion[^news-avgo-q4guide-20260902]. Snowflake, same day: quarterly revenue of $1.5 billion, and management raised the full-year product revenue guide to $6.07 billion[^news-snow-q2-20260902]. Shares jumped on adoption of the AI coding tools, with more than 60% of accounts now using the AI offerings[^news-snow-aiadoption-20260902]. One of those stocks got sold. The other one ripped. Marcus, the cashflow take.
Marcus: Broadcom is the surprising one, because it beat and got sold anyway. Going into that print, the memo had Broadcom at 53x free cash flow, on $34 billion of it[^memo-avgo-evfcf-20260904][^memo-avgo-fcf-20260904]. The page was drafted two days before he reported, so that is a clean pre-print anchor, and it says the market was already paying a full price. Then the guide came in roughly $200 million light[^news-avgo-q4guide-20260902]. On a quarter that size, that is inside the rounding error of anybody’s model. It got treated as information anyway. That’s what a multiple that full does to you. It removes your tolerance for a guide that isn’t perfect. We re-anchor when the fiscal third-quarter filing lands.
Ava: $200 million. On $35 billion.
Marcus: Now put Snowflake next to it, because the pair is the whole point. The memo had Snowflake at 100x free cash flow going into the same Wednesday[^memo-snow-evfcf-20260904]. And the forward revenue line has Snowflake growing about 29% over the next twelve months, against Broadcom’s 65%[^data-snow-ntmgrowth-20260904][^data-avgo-ntmgrowth-20260904]. Twice the multiple, less than half the growth. I’d hold that read carefully rather than hard, because a raised guide is a real fact, and the sell-side point about accelerating against hard comparisons is a fair one[^news-snow-analyst-20260903]. But whatever is setting the gap between these two names, it isn’t the growth rates. We re-anchor on both when the filings land.
Ava: So what is setting it?
Marcus: Direction of surprise, not level of value. Both managements said the same thing this week — demand is fine, timing and supply are the constraint. The market paid 100x for the version of that sentence that came with an upward revision[^memo-snow-evfcf-20260904], and marked down the version that came with a rounding error. That’s a statement about how the market is pricing revisions right now, and I’d weight it as the more durable read of the two prints.
Ava: Which is a polite way of saying the tape is trading the delta and ignoring the denominator. Same Wednesday. Same page of the memo. Same story about AI demand. And the company generating $34 billion of free cash flow[^memo-avgo-fcf-20260904] is the one that got punished. If you want the one sentence to carry out of this segment, it’s that anybody telling you the market repriced AI this week is describing a mood, not a mechanism. It repriced two guidance revisions. That’s a much smaller thing, and it’s much easier to be wrong about.
Rapid-fire
Ava: Lightning round, and it’s a leadership round.
Ava: Apple. John Ternus officially took over as chief executive from Tim Cook, ending a 15-year tenure[^news-aapl-ceo-20260901]. Ternus’s first internal memo talked up the product pipeline and a, quote, phenomenal iPhone launch on September 9[^news-aapl-iphone-20260901]. Phenomenal. Eight days into the job, describing a product he inherited. The memo has Apple at 34x free cash flow on $136 billion of it, with $76 billion of net cash on the balance sheet[^memo-aapl-evfcf-20260904][^memo-aapl-fcf-20260904][^memo-aapl-netcash-20260904]. Whatever Ternus wants to do, he has the means. Less fun: a UK class action is seeking £2 billion, about $2.7 billion, over App Tracking Transparency[^news-aapl-uklawsuit-20260903].
Ava: Albemarle went outside. Ragnar Udd, from BHP, becomes president and chief executive effective February 1, 2027, with Kent Masters moving to executive chairman[^news-alb-ceo-20260903]. Same week, Albemarle is preparing to reopen the Kings Mountain lithium mine in Cleveland County after finishing dewatering[^news-alb-kingsmountain-20260903]. A mining executive to run a lithium company that’s reopening a mine. The board was not subtle. The stock fell about 2% on the news[^news-alb-stock-20260903]. And ConocoPhillips seated both a new chief executive and a new chief financial officer on September 1[^news-cop-leadership-20260902].
Ava: PayPal. Stripe and Advent walked away from a $53 billion bid to take the company private[^news-pypl-stripeadvent-20260828]. Days later, another 251 jobs cut at the San Jose headquarters, part of a multiyear restructuring under new chief executive Enrique Lores[^news-pypl-layoffs-20260904]. The memo has PayPal at about 10x free cash flow[^memo-pypl-evfcf-20260904] — the cheapest name anywhere near this show tonight. Somebody was willing to pay $53 billion. Nobody is willing to pay it now. Same company, same cash flow, three weeks apart.
Ava: Also in the cutting business: Uber is eliminating 3,300 roles, 10% of its workforce, to strip out management layers[^news-uber-layoffs-20260902]. The specifics are unusually precise for a layoff announcement — a 20% reduction in the number of employees sitting seven or more levels below the chief executive, and half as many teams with only one or two direct reports[^news-uber-layoffs-structure-20260902]. That is not a cost cut dressed up as a reorganization. That is a reorganization that happens to cost less.
Ava: Forward week. Oracle reports Thursday[^earn-orcl]. Lennar the following Wednesday[^earn-len]. FedEx Thursday after that, and FedEx just agreed to sell its Supply Chain logistics arm to CMA CGM for $1.4 billion, so the print will have a story attached[^earn-fdx][^news-fdx-cmacgm-20260903]. Three reports in two weeks. That’s the whole calendar.
Close
Ava: That’s the show. A quiet earnings week is when you find out what management actually believes — and this week the boards, the balance sheets, and the buyers all said it out loud. Wall Street’s consensus this week: Broadcom’s guide was a disappointment and Snowflake’s was a triumph. One of those companies generates $34 billion of free cash flow[^memo-avgo-fcf-20260904]. It is not the one that went up.
Ava: Next week the calendar finally reopens with Oracle on Thursday[^earn-orcl]. On Wednesday’s show, Hunt walked the memo to page 3 and pointed out that Nvidia’s run-rate free cash flow now sits about $50 billion ahead of Apple’s — $186 billion against $136 billion[^ep-e2636]. Hold that number next to what you just heard about who the market rewarded this week. Hunt, Jason, and Mike are back Wednesday on episode 2637.
Ava: Download the memo at telltales.us, and send us a note through the Substack — we read them.
Ava: The show is produced entirely with AI tools, and both voices you’re hearing are AI-generated.
Disclaimer
Ava: This podcast is intended for informational purposes only. You should always do your own work to determine if an investment is suitable for you. The views expressed on this podcast are the host alone and do not constitute an offer to sell or a recommendation to purchase, or a solicitation of an offer to buy any security, nor a recommendation for any investment product or service. While certain information contained herein has been obtained from sources believed to be reliable, neither the host nor any of their employers or their affiliates have independently verified this information, and its accuracy and completeness cannot be guaranteed. Accordingly, no representation or warranty, express or implied, is made as to, and no reliance should be placed on, the fairness, accuracy, timeliness, or completeness of this information. The host and all employers and their affiliated persons assume no liability for this information and no obligation to update the information or analysis contained herein in the future, and may or may not hold positions in the securities mentioned.
Sources
* Albemarle Corp. names outsider as new Charlotte CEO in 2027. (2026, September 3). Charlotte Observer. https://www.charlotteobserver.com/news/business/article317113061.html
* Albemarle names BHP executive Ragnar Udd as next CEO. (2026, September 3). Investing.com. https://www.investing.com/news/company-news/albemarle-names-bhp-executive-ragnar-udd-as-next-ceo-93CH-4887717
* Apple Ternus touts product pipeline, phenomenal Sept. 9 iPhone launch in memo. (2026, September 1). Bloomberg. https://www.bloomberg.com/news/articles/2026-09-01/apple-ternus-touts-product-pipeline-phenomenal-sept-9-iphone-launch-in-memo
* Broadcom (AVGO) Q3 earnings report 2026. (2026, September 2). CNBC. https://www.cnbc.com/2026/09/02/broadcom-avgo-q3-earnings-report-2026.html
* Broadcom guides to big numbers but investors wanted more. Why we’re OK to hold for now. (2026, September 2). CNBC. https://www.cnbc.com/2026/09/02/broadcom-guides-to-big-numbers-but-investors-wanted-more-why-were-ok-to-hold-for-now.html
* Cheniere announces substantial completion of CCL Stage 3 project, production and export of 5,000th LNG cargo. (2026, August 31). Shipping Herald. https://www.shippingherald.com/cheniere-announces-substantial-completion-of-ccl-stage-3-project-production-and-export-of-5000th-lng-cargo/
* Cheniere Energy: Corpus Christi expansion strengthens its LNG growth outlook. (2026, August 31). Yahoo Finance. https://finance.yahoo.com/energy/articles/cheniere-energy-corpus-christi-expansion-234831212.html
* Chevron CEO says patience pays off in giant Venezuela oil deal. (2026, September 2). Bloomberg. https://www.bloomberg.com/news/articles/2026-09-02/chevron-ceo-says-patience-pays-off-in-giant-venezuela-oil-deal
* Chevron gets street-high PT at Piper Sandler on booming oil prices. (2026, September 3). Investing.com. https://www.investing.com/news/stock-market-news/chevron-gets-streethigh-pt-at-piper-sandler-on-booming-oil-prices-4887973
* Chevron to expand in Venezuela operations, doubling production through $7 billion investment. (2026, September 2). CNBC. https://www.cnbc.com/2026/09/02/chevron-venezuela-operations.html
* CMA CGM to acquire FedEx third-party logistics arm for $1.4 billion. (2026, September 3). EuropeSays. https://www.europesays.com/france/73958/
* ConocoPhillips names new CEO and CFO in leadership transition. (2026, September 2). BIC Magazine. https://www.bicmagazine.com/resources/people-going-places/conocophillips-names-new-ceo-and-cfo-in-leadership-transition/
* Equinor takes first LNG cargo from Cheniere. (2026, September 3). LNG Prime. https://lngprime.com/americas/equinor-takes-first-lng-cargo-from-cheniere/196203/
* John Ternus replaces Tim Cook as Apple CEO. (2026, September 1). The New York Times. https://www.nytimes.com/2026/09/01/technology/apple-tim-cook-john-ternus.html
* NextEra-Dominion $66.8bn merger clears shareholder votes. (2026, September 4). Power Technology. https://www.power-technology.com/news/nextera-dominion-merger-shareholder-approval/
* NextEra Energy Resources to build $13B natural gas plant. (2026, September 1). Pittsburgh Business Times. https://www.bizjournals.com/pittsburgh/news/2026/09/01/nextera-energy-resources-gas-plant-13-bil.html
* Oneok to buy natural gas assets with $9 billion Apollo stake. (2026, August 30). Bloomberg. https://www.bloomberg.com/news/articles/2026-08-30/oneok-to-buy-natural-gas-assets-with-9-billion-apollo-stake
* ONEOK to grow Midland midstream footprint with Brazos acquisition. (2026, September 1). Yahoo Finance. https://finance.yahoo.com/energy/articles/oneok-grow-midland-midstream-footprint-181837691.html
* PayPal adds another round of layoffs. (2026, September 4). Los Angeles Times. https://www.latimes.com/business/story/2026-09-04/paypal-adds-another-round-of-layoffs
* Snowflake (SNOW) Q2 earnings report 2027. (2026, September 2). CNBC. https://www.cnbc.com/2026/09/02/snowflake-snow-q2-earnings-report-2027.html
* Snowflake is ripping after its earnings report. Two analysts say there’s more upside to come. (2026, September 3). CNBC. https://www.cnbc.com/2026/09/03/snowflake-is-ripping-two-analysts-say-theres-more-upside-to-come.html
* Snowflake raises sales outlook, shares jump on AI coding tool adoption. (2026, September 2). Bloomberg. https://www.bloomberg.com/news/articles/2026-09-02/snowflake-jumps-on-revenue-outlook-uptake-of-ai-assistant
* Stripe and Advent end PayPal pursuit. (2026, August 28). Axios. https://www.axios.com/2026/08/28/stripe-advent-end-paypal-pursuit
* Uber to cut 3,300 corporate jobs in management overhaul. (2026, September 2). The Guardian. https://www.theguardian.com/technology/2026/sep/02/uber-to-cut-3-300-corporate-jobs-in-management-overhaul
* Uber to cut 3,300 jobs in company overhaul to reduce management layers. (2026, September 2). Bloomberg. https://www.bloomberg.com/news/articles/2026-09-02/uber-to-cut-3-300-jobs-in-company-overhaul-to-reduce-management-layers
* UK group tries to get $2.7 billion out of Apple over App Tracking Transparency. (2026, September 3). AppleInsider. https://appleinsider.com/articles/26/09/03/apple-sued-over-app-tracking-transparency-in-27-billion-class-action
* US oil giant Chevron confirms it will expand operations in Venezuela. (2026, September 3). AP News. https://apnews.com/article/venezuela-chevron-trump-oil-4d77fa9a9d53fc8693a242fbbb07fcef
* Why is Albemarle stock sliding today? (2026, September 3). Investing.com. https://www.investing.com/news/stock-market-news/why-is-albemarle-stock-sliding-today-93CH-4888185
Internal data
Internal data is provided on a best efforts basis.
From the publisher's feed