Capital Flows and Asset Markets

This is a free preview of a paid episode. To hear more, visit www.russell-clark.com

I am thinking of going to New York for 10-11 June. If you are an allocator, and would like to meet, then drop me a line.

Below is really a short thesis on AI - why its has blown up so aggressively, and how it could fall apart. Generally speaking, bullish ideas have relatively low value, but short ideas are very valuable indeed. For that reason, this is a paid post only.

I like Goldman Sachs research. They tend to produce big bits of research with enough raw data for you to make your own decisions. If you wanted a single image to explain the market explosion higher, then these next three graphs would probably be it. In the first graph, it shows that GS thinks “agent” demand for tokens is going to grow demand 13 times more than current token demand by 2030.

My guess is this GS Chart is an extrapolation from the sharp pick up in usage we have seen so far this year. This chart is taken from Openrouter.ai, and only shows volumes going through OpenRouter.

The below graph is showing an inflection in token pricing to be profitable. If correct, this should imply that the AI boom could be self financing, which would be bullish. Intriguingly, the OpenRouter data shows a dip in usage as prices rose - perhaps a sign that there is price sensitivity for AI users. I don’t have enough data to be sure of that conclusion, but something to watch.

Despite this inflection in pricing, Goldman still sees pricing fall. If there is price sensitivity, then falling prices would be necessary for GS forecast volumes to be attained. So the rise in compute pricing is something to watch.

But do these numbers make sense? In 2030, GS estimate monthly consumption of 120,000 trillion tokens at a price of at USD 0.18 per 1m tokens, this would imply monthly spend on tokens would be USD22bn a month, or USD260 billion a year. For context, the world consumes around 100m barrels of oil a day, which at USD 100 a barrel, implies USD 10bn a day, and USD3.65 trillion a year. This implies AI spend would be around 8% of oil spend. I struggle with those numbers, as the amount of investment announce by the hyperscalers and AI labs feel like they are looking for a business that generates far more revenue than that. Nvidia, which produces the chips that basically makes AI possible is forecast to have USD 500bn of revenue in 2028. Spending USD700bn annually for an industry that will generate only USD260bn in revenue does not make sense.

Nvidia sales is basically an inverse of hyperscaler spending.

I think the key question in this analysis is whether prices for AI will keep falling or not. The AI boom is creating software inflation. Or in other words, rising cost pressures for users.

Again from the GS report, they estimate that a AI agent is much cheaper for coding, but already at parity for a call centre and data agent. When I talk to my friend who is building out an agent to replace call centre, the costs for AI seem well below call centre agents - so not sure if this data point is correct.

It is possible that start ups have been subsidized by the cloud giants to drive demand and adoption - so the costs that start ups see if below the actual costs. I think this is what this chart below is trying to say - that in essence token usage and API costs can be very different.

I think what I am seeing is that the AI boom is now pushing cost pressures through the AI and software sector. As the GS model shows, the general assumption is for ever lower prices - offset by higher volumes. But if the GS numbers on call centres and data entry operators are correct, rising prices should put a dent into adoption rates. That being said, I would not think hyperscalers cut spending on lower adoption - I think they cut capex spend if competitors like OpenAI start running out of money, and then cut capex. From the data I have seen - I think that is 50/50 proposition now. But that is a post for another time.

This is a public episode. If you'd like to discuss this with other subscribers or get access to bonus episodes, visit www.russell-clark.com/subscribe

There was a time, not that long ago, when the cheapest provider would be the corporate winner. Remember the days when online shopping was always the cheapest option? Or when streaming was much cheaper than buying DVDs? This was a period when government enforced competition, which has sadly disappeared. These days, particular in tech world, its how much capital you raise that is the most important factor in determining a winner. A good recent example was Netflix, who during the “streaming wars” massively increased its balance sheet which drove up the cost of content.

Disney tried to compete for a while, but ultimately Netflix won.

And once it won, content prices fell again, and content creators got squeezed. In AI, we are seeing a similar dynamic play out. Masayoshi Son, the founder of SoftBank, virtually invented blitzscaling with an early investment into Alibaba, where he encourage Jack Ma to build quicker, or his capital would go to competitors. In the AI race, we have the hyperscalers, and the AI labs, OpenAI and Anthropic. One of the more intriguing aspects in the current AI boom has been the weakness in some of the AI names CDS. Oracle CDS has been weak.

As has Softbank Group.

Softbank is really more of a concentrated investment firm, with a number of high profile investments with a mixed track record. Alibaba was a great success, WeWork not so much. In recent years, is has been busy turning itself into an AI focused investment vehicle.

As of today, ARM and OpenAI are its two biggest assets.

Softbank was looking to raise capital by taking out a loan backed by its share holding in OpenAI, but saw some push back from lenders.

Softbank also have about USD 20bn of margin loans supported by its holding in ARM. ARM is one of the more expensive tech stocks out there - trading at 45 times EV/Sales.

SoftBank, Oracle, OpenAI some other partners have attempted to blitzscale the AI market with its Stargate project.

Softbank is increasing looking like the “soft flank” in the OpenAI v Anthropic battle for AI supremacy. Softbank has USD 130bn of debt outstanding (not including margin loans against ARM and OpenAI). Most is in USD, but there is a fair bit of Japanese debt as well.

In line with JGBs, the yield on this debt has risen substantially.

Not hard to see why - operating cash flow is poor, but capex is high.

One of the reasons I have been cautious on markets is that I saw capital becoming scarce - and we are now seeing that in the AI space - as these huge corporates keep tapping the market for more and more money. The question is when do bond markets break? When do markets stop asking how much do you want to borrow, and start asking when are you going to pay me back. We may be getting closer than you think.

This is a public episode. If you'd like to discuss this with other subscribers or get access to bonus episodes, visit www.russell-clark.com/subscribe

This is a free preview of a paid episode. To hear more, visit www.russell-clark.com

Showing my age, I am paraphrasing from Monty Python’s “The Meaning Of Life”. For years after seeing this movie, when ever I tried to get someone to order dessert after a big meal, I would put on a French accent, and say “It’s only wafer thin”.

Speaking of wafers, paid subscribers to the Substack and Slice will know that I quite like wafer stocks. Like most semiconductor related names, they are up a lot this year. Unlike most semiconductor names, they still look cheap. I detailed the buy case for wafers in February here.

If you cannot be bothered to go through the old presentation, I repost the most important part below. Basically ASML said in its January presentation that the new NVIDIA chips were going to need more silicon wafers.

This is a free preview of a paid episode. To hear more, visit www.russell-clark.com

I have preferred gold to bitcoin for awhile. Theoretically, they should be the same - scarce assets that should rise as investors worry about debasement. The debasement trade is easy to understand. The US is forecast to run 7% of GDP deficits for as far as the eye can see.

And yet the market is only going to pay you 4.4% a year to lend to the US government for 10 years. Somewhere closer to 10% feels right to me.

For the past 4 years or so, since the Russian invasion of Ukraine, and the subsequent freezing of Russian foreign reserves, I have preferred gold to long dated treasuries - GLD/TLT. It has been good, but during the ongoing war with Iran it has fallen some 14%.

In contrast while bitcoin has been very volatile over the last 5 years, it has rallied 28% from recent lows. Does it make sense to switch from gold to bitcoin?

One of the markets theories I was taught when I was young was that when oil prices are up over 50% year on year, the S&P 500 does poorly. Oil is certainly up over 50% year on year.

But rather than being a bearish signal, markets seem to think its a bullish signal.

I had also learnt to be cautious when gold starts to outperform the S&P500 - a sign that capital no longer felt safe in growth assets. Gold WAS outperforming the S&P 500, but not so much now.

Why did moves in commodity markets used to be bearish - but not now? In 2022, surging oil price did lead to weakness in US equities. What has changed? One guess is that normally, when oil prices and gold prices are surging, food inflation is also surging. Below is US Food CPI YoY. This is a lagging indicator, but you can see when food prices rise - problems for equities begin.

Above is a lagging indicator, but we have a more forward looking one. The CRB Foodstuff index. Here we can see spikes in food prices tend to lead to problems in markets.

For the US, I have found the wheat price is a good one to watch (its called the bread line for a reason). It is up from lows, but not much.

One argument you could make is that rising oil prices pushes up Urea price, and food prices with it. Maybe. But potash prices are still pretty low.

Another big counter argument is that the wholesale price of pork in China has collapsed, which means Chinese grain imports are also likely to collapse. Or in other words, global food inflation looks unlikely.

I have looked at agricultural markets for over 20 years - and they are by far the toughest markets to understand. Usually with commodities, supply is easy to model, but demand can fluctuate. With ags, demand is usually pretty constant, but supply can change rapidly. The big question is whether food prices are about to surge here or not? Or will that happen with a lag, as higher energy prices cause food production to fall in a year or two time? Personally, I can not generate a strong view on whether food inflation will surge or not from here, but it does seem to be the missing link between gold, oil and equities.

This is a public episode. If you'd like to discuss this with other subscribers or get access to bonus episodes, visit www.russell-clark.com/subscribe

This is a free preview of a paid episode. To hear more, visit www.russell-clark.com

To be perfectly honest with you, I should have lost money this month. If I look at the big macro trades I follow, they all pretty much went against me. The most extreme was GLD/SPX, which finished the month at lows. Big up months for me have been when this spikes, like in April of last year, or January of this year. That it was down 10% this month, should have been a problem for me.

When we look at GLD/TLT it also gave me nothing this month - with both GLD and TLT falling the same amount.

I also prefer Japan to the S&P 500, which on a relative basis was back at lows. This was also compounded by the fact that Japanese banks just traded in line with Japan this month, so no alpha there.

On the short side, private equity names, also bounced hard (as proxied by PSP US, the Invesco Global Equity Listed Private Equity ETF).

Despite ALL of that, it was a pretty decent month. How?

I have been engaged in a number of side projects for the last few months. One of them is writing a book about myself. I have written about eight chapters, but I am struggling to finish it off - there is just so much happening in the world today. So I have decided to serialise the book - that way, I am setting a deadline for myself to get my act together. Below is a taste….

OUTSIDER TRADER

They were the last big investor, the last hope, and the last chance to keep the dream alive. I was running out of time. The partners had given me a few years to work things out, but I was a loss-making venture; I was running on borrowed time. I had known it would be hard taking over from John, but I had never contemplated failure before. Not really. Now it was staring me in the face. And I was calm.

It is live here - OUTSIDER TRADER. I have also added a podcast for those who prefer to listen than read.

Also as previously announced, I have been working with a new website Slice. Slice is still a relatively new website - but it does offer competition to substack. They have improved both the front facing and back facing aspects of their website over the last few months, so I am making a renewed effort with this website.

With Slice, I think it works better with the “short selling” part of my business. Short selling naturally requires more trading, and by definition can not just short and forget (unlike long investing where buy and forget works). I also know that more trading oriented of my substack subscribers are very keen on short selling (as a rule, people tend to pay more for short selling as it so difficult to do). For that reason, I am skewing my Slice page to more short focused ideas.

Another feature I like is the portfolio part of the website. Again I am going to skew this to more short ideas.

Slice is more expensive than substack, and does not offer a free option. But if you are interested in trying, I am offering 50% of for the first 6 months - use BRUMBY50 as code. Any problems, drop me a line. With Slice, you get more a stream of consciousness, and less of the considered thought you get here. Find it here.

Enjoy!

This is a public episode. If you'd like to discuss this with other subscribers or get access to bonus episodes, visit www.russell-clark.com/subscribe

One of the questions I am often asked is surely the US government or the Fed will step in to stop yields rising too much. And usually I answer this question - with yes that true but that is why you need to own gold, as a hedge against exactly that. And that is a nice, simple answer, but could also be wrong. I could see another reason why long dated US yields rise, one that would closely echo the 1970s.

One of the more weird features of the current inflationary period we live in has been the total lack of improvement in the US government fiscal position. IMF is forecasting big deficits as fare as the eye can see. The contrast with Japan or the UK is striking.

And the yield premium over JGBs keeps falling.

But there is another, political, way to think about US treasury yields. Back in the 1960s and 1970s, yields were constantly going higher.

But unlike now, debt to GDP ratios were falling in the 1960s and 1970s. That is yields kept going up, even as the “creditworthiness” of the US was improving, while US yields today stay stable, despite the US government ability to pay declining.

As we have seen in recent years, the US government has used trade flows (via tariffs) as political weapon, financial flows (via sanctions) as a political weapon, and military dependence as a political weapon. We are also beginning to see oil and natural gas flows as a political weapon. So, logically, we should also see bonds and bond yields as a political weapon. In simple terms, owners of capital demand higher yields from the Democrats. Generally speaking, we tended to see bigger falls in yields in the 10 Year Treasury with Republican presidents, and yield rises under Democrats. Under President Trump, 10 year yields have generally behaved well over his entire term.

As mentioned in a previous post, there is a conservative political strategy called “Starve The Beast” which entails cutting government revenue, which then ensure that Democratic administrations must cut spending - a policy that worked well with both Clinton, and Obama - but one that Biden conspicuously ignored. Starving the beast strategy would probably accurately reflect the second Trump Administration - large scales tax cuts, plus the gutting of the IRS would be in line with this. The Democrats recognising that they were getting played, have joined in the tax cutting frenzy.

Since the 1980s, until Biden, the Democrats feared the bond market - and have broadly enacted Washington Consensus policies. But as we have seen with Biden, and now with Trump - there seems to be no fear. Or perhaps the Democrats have their own starve the beast strategy - spend aggressively, and let the Republicans be the ones to cut spending.

The UK is an intriguingly example. In 2022, Liz Truss announced an aggressive budget, and yields surged, and she was ejected from government. Since then, more conservative budgeting has happened, and yet yields have continued to climb. Is this because Labour is in power? Is it because the Greens and Reform are leading the polls? I don’t know, but I do not think the US political environment is too different - both the Democrats and the Republicans look fiscally irresponsible.

I guess what I am saying is that, markets could give up on US treasuries if President Trump announces a fiscally aggressive policy, or, should the Democrats starting doing well politically, bond markets should begin to act skittishly to try and convince them to enact austerity. Either way I cut it, short TLT still looks good politically.

This is a public episode. If you'd like to discuss this with other subscribers or get access to bonus episodes, visit www.russell-clark.com/subscribe

Yesterday, I gave a presentation to ANU Alumni of UK. It was just a quick presentation on AI, which I am now posting here.

This is a public episode. If you'd like to discuss this with other subscribers or get access to bonus episodes, visit www.russell-clark.com/subscribe