Capital Flows and Asset Markets

The big news is that Nasdaq fell 4% on Friday, which takes it all the way back to where it was a month ago.

The AI trade has become huge. And it is noticeable that equities have diverged from HYG. When HYG is going down, then equities tend to follow.

But credit spreads remain tight.

The real driver is the market reassessing the likely interest rate path of the Federal Reserve. US 2 year rates sold off on strong data.

So my main view is that a negative shock on bond yields led to a sell off in momentum stocks.

The big problem is that if I look at memory spot pricing, its hard to see any big change in the market. That is AI demand is still pushing prices up. The proposed equity raises seems to encourage that view.

We have seen a big increase in announced IPOs in 2026.

Could this mark a top? Unlike 2000, buybacks will still be greater than issuance.

As a share of nominal GDP, it is a large number, but not outside of historical ranges, but certainly at levels that suggest caution.

This issue here is that AI is now the main driver of growth and inflation. Could the Fed really end the AI boom by raising rates? My guess, given the AI/Cloud cartel that now exists, interest rate policy would be almost totally ineffective. JGB yields for me still reflect government policy that is inflationary rather than deflationary. A real change would be government policy to encourage competition and lower prices - something I do not expect from a Trump Presidency.

The price action of JGBs in the 2000s was starkly different.

My best guess is that the market is worried that new head of the Federal Reserve, Kevin Warsh, could be a Paul Volcker in disguise. With consumer expectation of inflation running at 6%, to my mind short rates in the US should be closer to 8 or 9%.

Politically, I find the Fed being uber hawkish very hard to believe. To me it looks like a position clearing sell off - but not a top. If I saw evidence that AI adoption was slowing, then I think we could read a top.

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

The way I work is I try and think about the world we live in, and what are the most likely long term outcomes. As of today, I still think a pro-labour world which drives higher inflation, and an eventually dearth of capital. My favourite way to play this has been GLD/TLT. This has been a great trade - but to paraphrase Monty Python, is seems to be having a bit of a lie down after a long squawk.

I don’t really read much other financial research. I do read Chris Wood, and I also quite like IMF reports. But one person I do pay attention to is Jeff deGraaf from RenMac. Why I like Jeff is that he only looks at charts, and then just publishes the most interesting ones. This chart showing an inflection in industrial metals versus precious metals.

He follows it up with this chart on general industrial metals breaking out..

And then this one on aluminium in particular. Aluminium in particular is related to war in the middle east as this has reduced supply.

I have been surprised by the relative weakness in gold since the war in Iran started, so maybe switching gold to industrial metals makes more sense? I do remember that copper/gold ratio and the US 10 year yield were highly correlated for a long time - but that this relationship broke down in recent years.

Looking at the above chart, you could argue gold is too expensive relative to copper, given the move in bond yields. For me, it just says that gold became a much more attractive reserve asset from 2022, when Russian foreign reserves were frozen. So gold has been much better than copper in recent years. Should I switch? My read of markets is that Central Banks remain far too dovish. And politically, cost of living hurts politicians far more than a bullish stock market helps them. So at some point, central banks will be forced by markets to raise rates (the long end has a huge sell off), or central banks raise short term rates to a high level to prove their inflation fighting skills. See circled part below.

I can not help but think, in either a long bond sell off, or a central bank induced slowdown, gold will outperform copper. I think what I am saying, is that in a free market world, treasury yields and copper/gold would indeed move based on economic activity. In this new pro-labour world of tariffs, and industrial policy, yields reflects lack of faith in central banks and governments, and gold also reflects lack of faith in central banks and governments. How will copper do with 10% on 30 year treasury? I don’t know. Maybe okay? But I think gold could keep going higher.

I guess what I am saying, is that backward looking models make industrial metals look attractive. But political analysis, which I think is more important these days still favours gold.

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

I have been generally very bullish on fertility rates rising again. I saw three big changes - first was rising wages, the second was improved reproductive technology, and thirdly was the rise of “work from home” culture, which made raising children easier. The one country that might be the first to show an inflection higher in birth rates was Denmark, which has greatly liberalised its access to reproductive technology. Sadly, month birth data has not really supported this view. Laws were liberalised in 2015, which saw a birth spike, but post Covid birth rates have fallen again.

I also liked Denmark, as Scandinavian nations tend to be quite progressive, with strong maternity rights. Despite this, Danish fertility rates remain low.

My brother forwarded to me a post on Facebook under the account name The Visible Hand which I found interesting. Basically, Visible Hand is arguing that fertility rates are declining because children are no longer a necessity, but a luxury. Whereas a hundred years ago, you needed kids as free labour and potential pension pot, now these needs are met through other means. The comparison he made was to horses. We used to need horses for everything, but as mechanised transport took off, horse population has collapsed, and owning a horse became a luxury/prestige item.

And we can compare this to say the numbers of chickens that exist today. Obviously we still need chickens for sustenance, and so their numbers continue to increase. For anyone familiar with chicken breeding will be familiar with the extreme use of technology to optimise chicken production.

It is an interesting argument, and it does raise the question of free-will. Horse and chicken populations are closely controlled by humans. And after the Malthusian fears of over-population in the 1960s and 1970s, humans made a conscious effort to reduce population growth, most notoriously with China’s one-child policy. But for the last ten years or so, politics has switched from reducing population growth, to promoting it. But as the Danish data shows, with still little to show for it.

As anyone with children knows, raising kids is expensive. And getting more so. Below is an inflation adjusted cost of raising a child from birth to 18. If we divided these numbers by 18, we get that annual cost in 2026 dollars in 1960 was USD 6,000 a child. In 2026, you are looking at USD 20,000. Or in other words, for the money you pay for one child today, you could have raised 3 children back in 1960.

As someone who grew up in the 1970s and 1980s, I can tell you that some of the cost savings came from a blatant disregard of modern health and safety norms, and I don’t think we are going back to a time when 6 years old are allowed to walk home from school unsupervised. The other big change has been the rising cost of housing. Rents have been rising as a share of income for decades.

One of the reason I am still bullish on fertility rates is that women increasingly are freezing their eggs, and then using IVF later in life, presumably when their incomes have risen. In the UK, we have seen a share rise births from frozen embryos, and a decline in births from fresh embryos.

The bullish argument on fertility is that with rising female participation in the work force, the age that couples started families were pushed back, and this naturally reduced fertility rates. But now that IVF is “solving” that problem, fertility rates should rise back above replacement rates, but with a lag.

The bearish argument is that capitalism required more labour in the post World War II period, so humans responded to the low cost of raising children, and relative high wages by producing more humans. And now, as automation and AI reduce the relative wages that labour can command, we are reducing supply, and turning child raising into something akin to raising thoroughbred horses.

When I write both arguments down, both are reasonably convincing. Even from my personal experience. I would liked to have had more children, but after having our second child at nearly 40, one more was a bit daunting. On the other hand, the two children we have are far better educated, travelled and trained than I was at that age. They are also more expensive than I ever was.

I guess what I like about both arguments, is that they both point to a world that is getting “better”. If its the first argument, technology is solving the issue of women having children at an older age - which I like. If it is the second, falling population growth is due to substantially more investment into children, which is creating a “super-generation” of humans. Right now, I would say the evidence is evenly split. As a fund manager, I would argue that the markets are not priced for rising fertility rates, which is what makes it more interesting - but this type of thinking is why most people hate finance people.

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

GLD/TLT actually lost 1.7% in May, as gold remains lacklustre, and TLT which looked like it was ready to break down, staged a rally. The US 30 year treasury briefly broke 5%, but came back again.

Coming into the month, it was beginning to look like credit markets were backing away from endlessly financing AI spend at both Softbank and Oracle, both firms closely related to OpenAI. Oracle CDS was perhaps the clearest example of these concerns.

I had lunch with an old client a few months ago. One thing he said to me was the GLD/TLT was my only idea. A little harsh, but maybe is is the only idea I have been REALLY committed to for the last four years or so. And why not? It WAS a great idea. But since a blow off peak in January it has been going sideways.

The really problematic part at the moment is gold. For the first time since 2024, its threatening to break below its 200MDA.

If I had to put a reason to its weakness, it would be almost exclusively with the movement in short term interest rates in the US. Since the War in Iran, markets have been pricing in a more hawkish Federal Reserve (and changing governor as well).

The 30 year has sold off a little bit, which is why GLD/TLT has held up, but we have not had a decisive move lower.

Now in a pro-labour world, I expect the long dated bonds to LEAD the Fed fund rate higher, as it did in the 1970s.

If that line of thinking holds, then I expect the 2 year treasury yield to move higher from here - back to 4.5%, where there 10 year is now, and possibly 5%. Given that the Atlanta Fed GDPNow forecast if 4%, this does not seem excessive.

Looking at the 2-30 spread on US treasuries, it seems possible that 2 year yields could rise another 1% before forcing the back end 30s higher. The 0.7% higher move in 2 years, has seen gold drop nearly 20%. Could another move of a similar magnitude see gold drop 20% again? That would take gold back to USD3,800 an oz, or where it was late last year. The problem would come if the yield curve flattens, and TLT does not fall. In rough number, that would imply a fall back toward 4 on GLD/TLT.

That is the risk if I just looking at market movements this year. Fed Fund rate expectations rise, gold weakens, and yield curve flattens. Against that, US 30 year yields are already trading at one of the lowest premiums to JGBs ever.

In absolute terms, its only just about 1%.

In percentage terms, the gap has probably never been lower. Typically treasuries have been 200% of JGB yields.

IF the gap between JGBs and treasuries were to revert to historical norms, that implies a US 30 year yield of between 6% and 8%.

So this leaves GLD/TLT in an interesting place. If the Fed tightens, TLT could rally and gold fall - which would be bad for a already sideways moving trade. If the old relationship between JGBs and treasuries reasserts itself, then TLT should fall between 20 and 40% (based on TLT having a duration of about 20).

Normally, I do not like 50-50 bets like this. Maybe JGBs yield trade through US treasury yields? It seems odd - but odd things happen alot in this market now. The big kicker for me is that there is still USD 3.5 trillion of official treasury bond holdings out there. Given the freezing of Russian treasury holding - who is buying this debt?

The largest holder is Japan, which needs to sell treasuries to support its currency, and the second largest holder is the UK, which is probably hedge funds doing basis trade, and the third largest is China, which is a seller.

I think it is possible that GLD, and other precious metal do nothing, especially true given the massive retail participation they drew during their run up. But, GLD/TLT still looks good - but mainly the short TLT side.

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

The original thinking behind the Pro-Labour trade was that politicians everywhere were being forced to address imbalances in the economy. The main one was sky high property prices. After the GFC, another property crash was off the political cards more or less. This left politicians with only one choice, raise wages while keeping property prices flat in nominal terms. This issue, as I saw it, was that rising wages would have an upward bias on property prices, so yields would have to be quite high to keep property in check. In the UK, 30 year bonds yields have indeed risen substantially.

And London house prices have stagnated.

And UK wages have surged.

Broadly speaking, this has generally been repeated across the western world. I also assumed private equity would struggle in this world of higher wages and higher interest rates. The Invesco Global Private Equity fund has confirmed that view.

So far so good. I had assumed that higher yields would also put a dampener on equity valuations, like we saw in 2022. On this matter, I could not be more wrong. Using price to sales as a valuation metric, stocks have gotten more expensive, not cheaper.

If you use an equal weight index, you could maybe argue there has been a slight de-rating, but it is pretty flimsy.

The place to be has been semiconductors. If dominates both the S&P 500 and momentum indices.

Weirdly, AI almost now makes a direct connection between wages and semiconductor prices. The higher wages go, the more corporates are incentivised to use compute, rather than labour. Whereas oil was the substitute for labour (or oil prices moved with rising wages) in the 1970s, semis now seem the substitutes for labour in 2020s. Oil rose 1800% during this era, as the market worked out the “right” price for oil.

Semiconductors have done even better than that.

If the analogy between oil and semis hold, then the market breaks only when politicians decided that rising unemployment to break inflation is a political winner. I don’t know when that is.

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

In a recent note, I was looking at Softbank and Oracle, and thought they looked the weakest link in the AI financing chain. In simple terms, all the other AI players have access to the cash to compete, while Softbank was relying on margin loans on its ARM and OpenAI holdings. Looking at debt markets, CDS were also signalling weakness.

The law case that Elon Musk had against OpenAI was dismissed, and Softbank shares have surged to new all time highs. I do find it odd that credit and equity markets are saying different things.

Share price surges like this can happen when there is a lot of short interest. There are plenty of examples of this, but Avis Budget Group is a recent example. A slightly messy chart, but the blue line in the upper chart is share price, and the the middle chart shows that short interest built in 2025, leading to a mega short squeeze this year, with stock up 700% at one point.

There is no short interest in Softbank, at least reported short interest. In fact you can see short interest in Softbank was obliterated in the move from 1500 to 7000 in 2025. Of my short selling rules is never short stocks with high short interest, mainly to avoid squeezes like we saw in CAR US. There was no sign of squeeze potential in Softbank.

One reason could be that OpenAI is looking to launch an IPO, and this will greatly add liquidity for Softbank and Oracle, OpenAI partners in Stargate LLC. The problem with that view is that Oracle has barely reacted to news of the potential Open AI IPO..

Thinking about it, there are a few options for what is driving Softbank - the first and most obvious is retail enthusiasm, which given the move in most AI plays this year is understandable.

The second, which is more centred in finance, is that Softbank has taken margin loans out on its ARM and OpenAI stakes. In both cases, the size of the loans could not be easily hedged via ARM or OpenAI directly. Lenders to Softbank may have chosen to directly hedge via Softbank options. This should lead to implied volatility moving above historic volatility . We sort of see that with Softbank. This means Softbank ends up doing the opposite of what I expected - where a doom loop of falling valuations means equity comes under more pressure, the reverse is happening. Rising valuations lead to a rising Softbank share price.

On a completely different thought track, I also thought that OpenAI might be in trouble due to Anthropic taking market share, and/or retail subscribers not accepting price increases. I am subscribed to £20 a month ChatGPT function - which I use with work. It is good at making images, so have cancelled my Shutterstock account, and good at research, and I particularly like how it filters out Google ads from search results. But for writing, and organising, stuff, I felt it was incrementally better, rather than ground breaking. That changed this weekend, when I was working on my golf swing in my back yard, and uploaded a video to ChatGPT to see what is thought. I am still getting to grips with golf - and I am told by friends that I am “armsy”… which is exactly what ChatGPT told me. I was genuinely surprised at how quickly it was able to analyse my swing and give tips. The big one was getting my shoulders to follow through.

After a few drills it had improved my swing measurably. A golf lesson cost £50 - and ChatGPT had done a decent job. Made me think, there is definitely going to be some pricing power for these LLMs.

There were a lot of reasons, I liked the look Softbank being a weakest link - and it may still play out that way. But against that, there should be short interest from options that I cannot track easily, retail meme style investing, which you do not want to short against, and it has hit a new all time high. I have been bearishly minded for the last couple of year, with the breakout of gold versus the S&P 500 - which has normally a good time to be at least cautious. However the last few month has seen half of gold’s outperformance get reversed AND break through the 200MDA.

I never try and fight the market - its a fools errand. And markets were sending mixed messages until very recently. But I am a bit nervous about the signal now. On top of that, I have a new found respect of the capabilities of LLMs (I use LLMs for many things, but video analysis is the first one that I have really been amazed at). I think we would need to see more financial problems at OpenAI before getting really bearish. A price war might do it - but if anything the market is getting prepared for price hikes I think. The macro graph that has really broken out is SPX/TLT - which is at new highs, after having shown some weakness in 2025 and early 2026.

When do sovereign yields rise enough to bother markets? The market WILL tell us - but at the moment, it is saying not now. And within that analysis, betting on renewed Softbank weakness looks tough.

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 when a prospectus was a dime a dozen. As an analyst in the 2000s, my main job was to read them and then go the management lunch, and work out if they were a buy or not. I love a good prospectus - its the only time management has to be clear and up front about risks. But with the rise of private equity and venture capital, IPOs have become much rarer. But today we get to go through a truly interesting prospectus. Above picture is taken directly from the prospectus. You can have a read yourself here. For such a momentous IPO, basically the key feature is that the IPO is needed because the capex into AI is larger than the cash that can be generated from its excellent rocket launch and connectivity businesses.

Not included in the above is that the IPO will allow SpaceX to purchase Cursor USD 60bn. There is quite a hefty termination fee, so obviously Elon is very keen on this deal. Cursor is an AI powered coding company, and was something that Grok was missing when compared to the other big AI companies. It also provides a ready made corporate client list.

I am not going to give you a detailed financial analysis, AI and plenty of other Substacks can and will do that for you. But I will give you my read. Its pretty transparent that you are getting two excellent businesses, the launch business and Starlink, and one speculative venture. The Space business only lifted 2,000 metric tonnes into space, or about 5 International Space Stations (420 tonnes). If a moon base is to be built, this number will have to go up exponentially, and it has a huge cost advantage. Starlink has 10m customers now - and with a ARPR of USD 66 per month is not that expensive. Not hard to foresee 100m subscribers in this business.

AI is a cash sink, but it is meant to be. But its pretty explicit that the AI business is the business that could really surprise to the upside. The theory of space based data centres is very interesting.

So far as expected. But I have two different takes on the SpaceX IPO. First of all, it confirms in many ways, the rising cost of capital theory. The rising cost of capital for governments is beginning to force some changes in capital markets. There is no way not to see equity valuations as expensive relative to government bond yields.

But the credit market has been increasing harsh on companies borrowing for AI. Softbank and Oracle CDS has been showing that.

So we are seeing the return of equity markets as the source of capital. For years the share count for the S&P 500 was falling, but now we have seen a turn as equity gets issued. We are not at dot com levels of issuance yet - but with the IPO pipeline opening up, this should climb higher, especially if government bond yields keep rising.

The second observation I would make on SpaceX IPO, is that Elon Musk really hates Sam Altman. I am using the word HATE, about Sam Altman. I suspect Elon Musk has a driving urge to be number 1, and a overdeveloped sense of vengeance on those that have done him wrong. For me, the SpaceX prospectus make it clear that Elon wants to destroy OpenAI.

So what has he done to make me think that? Not only has agreed to purchase Cursor to have a more competitive product, the prospectus also notes that the entire Colossus facility has been lease out to Anthropic at a cost of USD1.25bn a month.

He also has been relentlessly pushing Grok in twitter - which obviously allows him to compete with ChatGPT on the consumer side, if not competitively yet. All of this raises an interesting questions about the potential OpenAI IPO. With Microsoft moving away from OpenAI, and Elon Musk supporting Anthropic, OpenAI will need to raise a very substantial amount of capital from the market to compete. If Microsoft also looked to sell some of its stake - would this supply overwhelm demand?

I guess what I am saying is that capital has become a competitive weapon - and it still looks like OpenAI is at a disadvantage. Softbank has bounced on the OpenAI IPO announcement.

But its CDS remains strangely unmoved. Credit markets tend to be more “show me the money” type of markets (not always - but can be).

This makes sense to me, as SpaceX IPO is signalling even more capital is going to be invested in this area. As long as growth rates remain good, equities should be okay I guess, but the rising cost of capital is causing equity issuance to pick up. This is a change from the last 10 years or so. Is OpenAI still in trouble? Well if Oracle catches a bid, then we can assume the IPO will solve the capital raising issues. Time will tell.

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

Three years ago I wrote a note on How to Read The Economist. A lot of financial people do not take the Economist seriously - there is the famous contra indicator of the cover being the exact wrong thing to do. I do not disagree with this. But as I say above, and in my previously published note, the key to the Economist is to read it backwards.

The first thing you find at the back is the Obituary. I find reading the Obituary is great motivation to try and do great things with my life. In fact one of my goals is to have my Obituary published in the Economist. This does mean I would have to fake my death to know if I have achieved this or not. This raise a few issues, but I will deal with that when I think I have done enough with my life to get a write up.

So when I picked up the paper copy of the Economist on my way to the toilet, rather than read about Indonesia at a Crossroads (again), I started at the back. The Obituary of Raghu Rai, and his photographs of the Bhopal tragedy humbled me. I pondered if that is the standard, then I will never make it. Raghu Rai was not only an excellent photographer but a brave and moral man. I must do better.

After contemplating my shortcomings for awhile, I turned the page from left to right. This took me to the back of the Culture section, which was reviewing the book Project Maven by Katrina Manson on AI and American warfare.

I knew that AI was being used with drones, sometimes to fly autonomously, but from what I had read, Russia was able to keep pace with Ukraine here. But as I read the article, what came clear, is that ability to crunch huge sets of data points was becoming ever more important. The other thing that it makes clear, is that AI starts off making a lot of mistakes, but then learns rapidly.

This seems to be made clear with Ukraine now seemingly turning the corner in the War with Russia. Drones and AI now seem to be much more effective than at the start of the war. This below is taken from another Economist article.

Since the AI boom started I have wondered why American corporations have been willing to accept the huge price increases, particularly when Chinese AI seems to be so much cheaper. I really thought the DeepSeek moment last year would be a motivator for US corporations to drive down prices. Instead pricing has continued to surge. In recent weeks, I have been thinking that the hyperscalers are pushing up prices to try and bankrupt the competition.

But if AI is seen as an extension of the “Department of War” then the above pricing dynamics make much more sense. When the size of the navy was important to America, then building far more ships than anyone else was the most important thing. Prior to Pearl Harbor, the US probably correctly perceived that a huge navy was not really necessary. After Pearl Harbor, a navy that was many times larger than Japan’s was seen as necessary.

A similar dynamic played out with nuclear weapons. Once USSR had nuclear weapons, too much was not enough.

And the US has a good model for bringing on supply - it guarantees payments and profits to encourage private sector investment and innovation. If anyone watched the movie Aviator about Howard Hughes, he was criticised for building a plane that was never flown, but this was perfectly inline with how the US does things. AI as a military trade makes much more sense than a rational economic exercise. But the risk/reward for AI then moves to a political calculation. More on that another time.

Also, if the above analysis is correct, and AI is learning, we should expect US drone strikes to become more and more effective against Iran, just as Ukraine has become more effective against Russia. That would be a real game changer and bearish for oil. It would also effectively imply that Taiwan was also now impregnable, and could only be reunified through peaceful means. Strangely, if that did become the case, it would be bearish for AI trade as rational investing would take over again. Interesting times indeed - and another lesson in why I read the Economist backwards.

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

After a blow off move in January, GLT/TLT has been going sideways since then.

As mentioned numerous times, I am more bearish TLT than bullish GLD, but I needed to have GLD just in case the Fed did YCC or something similarly foolish. And in 2024 and 2025, gold has been much the better trade. As is the case with all these relative value trades, the market fell in love with the side that is working - gold and other precious metals. To be fair, they still look to be in an upturn, but back in January every retail trader and fund manager were super long gold, silver and miners.

Gold looks ok, but much more problematically, is has given up a huge amount of relative performance to the S&P 500. That is, holding gold let you outperform the S&P 500 since the beginning of 2025, but you are now down on gold relative to the S&P 500 this year. This year, gold as a bearish trade is not really working.

So should I get rid of gold? Probably not. Gold is reacting to higher interest rates as you would expect. The real mystery is corporate debt. Even with oil up, rates selling off, and private credit not looking great, corporate debt spread remain at lows.

The corporate debt market has learnt to price in a Fed put. And as we have seen, corporates are now borrowing against asset values for capex. That is the Fed put is inflationary. Kevin Warsh and Scott Bessent have spoken against the Fed trying to price everything. If TLT is pricing in a more “free market” Fed - why are equities not? S&P 500 is looking like the mispriced asset here. Fun times.

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