Maybe it’s not high-tech that’s the answer to improving the world, maybe it’s actually… “deep tech.” Pablos Holman ran a deep tech lab for investors, including Bill Gates and Nathan Myhrvold, and built spaceships for Jeff Bezos’ Deep Blue. He’s written a book this year called “Deep Future: Creating Technology That Matters.” BBTW editor Peter Green sat down with Holman for a chat.

Peter Green: We talk a lot about how tech is going to save the world, and we’ve seen software and now AI effect massive change and transformation on our world — and make its founders and funders very rich. But what is “deep tech” and why does it matter?

Pablos Holman: With software, we’re just going places we’ve already been before. You’re going to move some puzzle pieces around, but they’re the same puzzle pieces, the same businesses. Starbucks $SBUX is going to have to tear out all their IT and put in new IT that has AI in it, but it’s still the same service they’re providing. I’m talking about what we call deep tech. Using advanced technology and science breakthroughs to reinvent things that humans do. Old industries. It’s about how do we make mining 10 times cleaner? How do we make shipping 10 times cheaper? How do we make airplanes 10 times cleaner and faster and cheaper? Things like that.

Software is what we’ve been doing for 25 years or more. Now we have to go there with AI, but we’re just going places we’ve already been before. Software is not going to solve the processed foods that are growing fat on your liver. We’re going to have to come up with new technologies that help us to solve those kinds of problems. That’s what I mean when I say deep tech.

How is the deep tech revolution going so far?

We have a whole raft of breakthroughs that have come out of scientific research in the last few decades. But they get dropped on the floor because all of the capital in the world that’s there for doing new things, venture capital, is only being used for making software. And investors and capital allocators, when they see hardware coming, their brain says, ‘that looks hard.’ That’s why it’s called hardware.

But 30-40 years ago, Silicon Valley invented the PC and the smartphone. What’s stopping them from reinventing the hardware?

One is all the easy stuff’s been done. All the big opportunities have been soaked up by somebody. So now they have to refocus their attention on where are the bigger opportunities? Well, the bigger opportunities are everything that software ignores and that software can’t fix.

The new inventions, the new technologies, the insights that make it possible to make something 10 times better, that’s new technology. Those are deep tech breakthroughs. And they come from all different kinds of places. They come from labs and garages, and they come from basements with nerds who are tinkering. Once they have that insight, you ask, does this make anything humans do 10 times better? And if the answer is yes, then you start looking at, all right, how do we bring that to life?

So how do you bring that to life, when all the capital is focused elsewhere?

I co-opt the machinery of startups, of venture capital, because that’s the machinery that’s good at doing new things. I try to find these things, and then I invest in them early to get them out of a lab and into a startup. A startup is the package that the world knows what to do with. Then we can attract capital. The startup machinery is amazing because it’s a kind of milestone-based success thing. That’s the good part about Silicon Valley. They’re just aiming it in the wrong direction.

What are some of the projects you’ve got going in deep tech?

We have a nuclear reactor that goes in a borehole. It’s buried a mile deep. Even if the thing were to have a catastrophic meltdown, there’d be no radioactivity at the surface. It’s also super cheap. What you put on the surface is a turbine and generator: 54-inch diameter, 15-megawatt electric. That’s enough power for a skyscraper or a data center, and if it’s not enough, we drill another hole next to it, and deploy arrays of these things. We have a company in California. They’ve figured out how to automate apparel manufacturing. Instead of using stitches, they use glue on a seam. And the glue is now stronger than sewn seams. It’s stretchy. Now the robots can assemble the clothes. They can produce on demand, and the bonding for the seams is reversible: You heat them up, they fall apart, and you can recycle the fiber. Another great example: Cargo ships that can sail themselves. We have electric backup. With zero emissions, no fuel, and no crew. So we make fleets of smaller ships, and they go directly to where the cargo is needed without trans-shipping at a major port. I try to learn about all the big problems in the world. I’m kind of collecting problems, and big industries are basically a euphemism for big problems. So to me, that’s where it starts.

This interview has been condensed and edited.

—Peter S. Green

Big Businesses Mentioned This Week

$NVDA ( ▼ 1.88% )  $PSKY ( ▼ 2.65% )  $ORCL ( ▼ 5.11% )  $CMCSA ( ▼ 3.79% )  $UPS ( ▼ 1.79% )  $AMZN ( ▼ 2.23% )  $RIVN ( ▼ 4.12% )  $TAP ( ▲ 0.51% )  $GM ( ▲ 0.36% )  $BRK.A ( ▲ 1.1% )  $BAM ( ▼ 1.3% )  $CCJ ( ▼ 1.85% )  $GOOG ( ▲ 3.35% ) 

The usual suspects

• How high is up? For Nvidia’s $NVDA share price, the answer is all the way up. On Thursday, Nvidia became the first private company worth $5 trillion, with its shares closing at 207.16 on Wednesday, up nearly 50% in the past 12 months. That’s because the U.S. economy is making a massive bet on AI. So forget the shoulda, wouldda, couldda. Will Nvidia keep rising? “It is justified only if margins and profits continue on the current trajectory or even get better,” David Kotok, co-founder of Cumberland Advisors, told the Wall Street Journal. Nvidia is now trading at about 33 times its projected earnings for the 2027 fiscal year, while the S&P average is 24x. That said, Tesla is trading at 243 times earnings. Morningstar analyst Brian Colello says Nvidia will keep growing. “A 40%-plus growth year is [likely] in the cards for fiscal 2027,” he wrote on Wednesday.
• The cuts just keep on coming… or do they? It’s Powell Power time: The Fed’s Open Market Committee cut its benchmark overnight interest rate by another quarter of a percentage point, to 3.75% to 4%. But don’t expect that to show up anytime soon in your mortgage or credit card bill. That’s just the rate at which the Fed lends money to banks overnight. Mortgages and other consumer (and business) loans are shaped more by market forces than anything else. And those forces are not suggesting there’s any hurry to lower mortgage rates. The standard 30-year mortgage was at 6.19% before the cut, and on Thursday stood at 6.14%. That’s in part because home sales are rising again and mortgage demand is also rising as a result. Mortgage rates were down just a tenth of a percent, to 6.39% from 6.49% earlier this month, while the latest numbers on existing home sales showed a 4.1% jump from a year earlier, largely, says a CNBC survey of realtors, because home prices are dropping. Wednesday’s cut doesn’t mean there will be another at the Fed’s December meeting. For one thing, with the federal government shut down, the Fed is flying blind when it comes to real inflation and job numbers. And then there’s the fact that two of the 12 members were against the cut (one wanted no cut, and Trump appointee Stephen Miran wanted a half-point cut). Powell said, “There were strongly differing views about how to proceed in December. A further reduction in the policy rate at the December meeting is not a foregone conclusion. Far from it.”
• Cuts At Paramount: Well, that was quick: David Ellison, the new head of Paramount $PSKY and son of Oracle $ORCL founder Larry, has just made his first cuts since acquiring the fabled Hollywood studio and its assets, including CBS. More than 2,000 people are getting pink slips, including about 1,000 in the U.S. They include CBS Saturday Morning co-hosts Michelle Miller and Dana Jacobson. CBS Evening News anchor John Dickerson said he’s walking by the end of the year, and Paramount’s movie and TV unit is losing “Yellowstone” and “Landman” producer Taylor Sheridan, who’s off to Comcast’s $CMCSA NBCUniversal, where studio boss Donna Langley offered him a film and TV deal worth $1 billion over eight years. Hollywood news site The Wrap said Ellison “started butting into” Sheridan’s creative process, a longtime Hollywood no-no.
• Making the cuts: And more job cuts are coming: UPS $UPS has cut its half-million-member workforce by 48,000 people over the past year, the company said Tuesday, pointing to its share price, which is down about 27% in the past year. Tariffs have cut the number of packages sent to the U.S. from China by 30% in the third quarter, and Amazon $AMZN is shipping more of its own packages. Amazon cut 14,000 managerial jobs. Target’s cutting 1,800 of its 22,000 corporate roles, and pink slips are going to white-collar workers at Rivian $RIVN, Molson Coors $TAP, Booz Allen Hamilton, and General Motors $GM. Behind the trend: Shareholder and board pressure to use AI to boost profits, and general uncertainty about the future of the U.S. economy.
• Has Buffett had all he can eat? The departure of America’s favorite investor, 94-year-old Warren Buffett, from his holding company Berkshire Hathaway $BRK.A has some Wall Street analysts concerned the company may not be as good a bet without him. Analysts at Keefe, Bruyette & Woods downgraded Berkshire to “underperform,” saying lower car insurance margins, tariffs, falling interest rates, fewer clean energy tax credits, and Buffett’s own planned departure are hurting profits. KBW lowered its target price for Class A shares to $700,000 from $740,000. Berkshire shares have fallen 3.2% in the past five days, closing Wednesday at $713,036.

AI energy news

• More Nukes? AI needs power, but building new power plants is expensive and speculative, so where private industry fears to tread, the Trump Administration has gone in: On Tuesday, Energy Secretary Chris Wright announced an $80 billion deal with the owners of nuclear powerplant constructor Westinghouse, Brookfield Asset Management $BAM and Cameco $CCJ, to build an unspecified number of reactors to help fuel the AI boom. It’s not clear how the $80 billion in taxpayer money will be spent, and the cost of the last two reactors built by Westinghouse, in Georgia, cost $35 billion, about double initial estimates. They were also finished seven years late.
• Who’s gonna pay for all that power? You and me and anyone else who wants to invest. OpenAI this week completed its conversion to a for-profit company, and is now preparing to go public, raising $60 billion at a $1 trillion valuation next year, according to press reports.
• The heat is on OpenAI: Rival AI firm Anthropic says it’s reached a deal with Google $GOOG to buy up to 1 million of Google’s custom-designed Tensor Processing Units, adding a gigawatt of compute capacity by 2026. That’s a lot smaller than OpenAI’s planned 33 gW Stargate data center, but Anthropic’s picked the slow and steady lane, and its algorithms work across chips made by multiple designers, allowing for less expensive and more efficient computing.

Trumplandia

• It’s never a quiet day in Trumplandia. Remember when Joe Biden was president and world trade hummed along fairly smoothly? As Canada’s Toronto Blue Jays trounced the L.A. Dodgers in the opening game of the World Series last weekend, Ontario Premier Doug Ford ran an ad on U.S. TV showing Ronald Regan speaking out against tariffs. Trump was outraged, and said he’s cutting off trade talks with Canada. Neither Ford nor Canadian PM Mark Carney seemed particularly upset. Meanwhile Trump’s three-day visit to Japan ended with an ill-defined pledge from Japanese leaders to invest some $550 billion, about one-eighth of its annual GDP, in the U.S. “No specific projects have been finalized at this point,” Finance Minister Satsuki Katayama said Tuesday. In South Korea, Trump tried to extract a $350 billion investment, or about 20% of the country’s annual economic output. But after being gifted a replica of an ancient Korean royal crown, he touted an as-yet-unsigned and undocumented agreement to invest the $350 billion at the rate of $20 billion a year. Any more, the Koreans said, and it might blow up their currency.
• Then came the big reveal: China, where Trump rolled back all his tariffs in exchange for China dropping the tariffs and rare-earth export ban it had imposed in response. China said it would buy U.S. soybeans again, and made vague promises to end the export of chemicals used to make fentanyl. The big business question is whether Trump will let Nvidia sell its Blackwell AI chip to China. Access to those AI wafers could let China get a permanent jump on the U.S. “The defining fight of the 21st century will be who controls artificial intelligence,” said Democratic Sen. Chris Coons. “It would be a tragic mistake for President Trump, in order to get some soybean orders out of China, to sell them these critical cutting-edge A.I. chips.”
• Shutside Story - The longer the government shutdown goes on, the more it costs, and the non-partisan Congressional Budget Office, working without a paycheck for the past four weeks, says the U.S. economy will never recover between $7 billion and $14 billion in lost productivity, between missing paychecks for 750,000 federal workers and interrupting food benefits for low-income Americans. “Depending on its length, the government shutdown will reduce annualized real GDP growth in that quarter by 1.0 to 2.0 percentage points,” the report adds.
• Apprentice Fed Chair Edition: it’s down to the Final Five, Treasury Sec. Scott Bessent said Monday, naming the finalists to replace Fed Chair Jerome Powell, when his term expires in May. The five are current Fed Governors Christopher Waller and Michelle Bowman, BlackRock executive Rick Rieder, and two guys named Kevin: National Economic Council Director Kevin Hassett and former Fed Governor Kevin Warsh. Bessent has been interviewing the candidates himself and said he’d present Trump with a super-short list after Thanksgiving.

Elon’s World

• Alienation: Elon Musk’s controversial entry into being more outspoken about politics cost Tesla between 1 and 1.26 million car sales in two and a half years, according to a new study for the Bureau of Economic Research by four Yale professors. Tracking county-level data on car sales and partisan voting trends, they found that Musk’s increasingly extreme political views simply drove away potential Tesla buyers. He “antagonized his most loyal customer base, for, as we show, Democrats are far more likely than Republicans to purchase a Tesla,” they wrote.

Get Big Business This Week in your inbox every week—and read it before everybody else! Sign up today.

TKTK