In my old fund, I used to hire summer interns. Like everything I do in my life, I looked at the results, and then try and improve the outcomes. When we first started, we asked for CVs first and then did interviews. This process sometimes worked, and sometimes didn’t. The interns that seemed great on paper and interviews, were just not great investors. One summer I changed the process, and I asked potential interns to write 200 words on an investment idea. I also asked them to submit the idea with just the mobile phone number, so we had no idea of sex or ethnicity of the candidates. Inverting the process improved the intern outcome substantially. We wanted people who understood investing, not people good at writing CVs or doing interviews.

One of these interns from the enhancde process has gone on to quite a successful career in fund management, and whenever he starts his own fund would put some money with him (which is about the highest praise I can give). He recommended I read Leopold Aschenbrenner’s essay, Situational Awareness, which became the basis of this hedge fund. I can see why he raised so much money. It is surprisingly clear. The first graph is really the key one, and basically says the same thing as what Anthropic is saying in its prospectus.

This is a straight line graph, but it is a log graph. And the intelligence scale on the right is a bit arbitrary, but we can easily correct that (using ChatGPT of course - my new intern). In simple terms, as you add compute, you add intelligence. But it is non-linear axis, it is log scale. Adding 10 times the compute generates 10 point IQ increase. Note Chat GPT-4 is standardised at 1 here. Today’s Chat-GP6 used 45 times more compute - so would be at 45, so we falling behind a little on his chart above. The key idea is for a linear increase in intelligence, you need a order of magnitude increase in compute. This graph from ChatGPT is not exactly right - like an intern it still makes mistakes, but its close enough for our needs.

Here is the full estimate of the compute increase.

Humans are bad at exponential stuff. We are just not hard wired that way. But looking at this graph, I can see why the big tech companies are investing so heavily. But you may be asking, what exactly does more intelligent mean? IQ is really an arbitrary measure I once went to a Mensa meeting (long story) and was not impressed by what I saw. IQ and life success do not correlate, at least not at Mensa anyway.

In this case, IQ for AI means it gets cheaper to do the same task, or easier. What we humans call smart. So using a state of the art chat bot, actually is cheaper to do the same work. I find this confusing as well There more expensive the training, the cheaper it does the work. There is a motorsport analogy here. Better engineering is always good, but more money is always better.

To be fair to Aschenbrenner, he also points out where things could surprise. Efficiency has been doubling every 8 months, which is essence is the bear case. There will be a bear market if efficiency improves so quickly, that current investments look diabolical. But he also points out, we could have already picked the low hanging fruit in efficiency gains.

The exponential nature of AI does lead to some wild graphs. On the trend he predicted, AI would need 100% of USA power generation in 2030. This is probably the most bearish graph in the essay. Does the rest of the US give up electricity in the search for super intelligence? I think not.

But despite that Ashenbrenner also sees the USD 100bn and even USD 1,000bn clusters coming. No doubt these numbers are what made Elon Musk think about space based data centres.

Looking at his prediction he made in 2024 for 2026 and 2028, he is not far off.

I would also say his predictions on TSMC capacity look about right.

To be fair, this is one of those very good bits of research you read from time to time, and they are so wild you think they are crazy to start with, and over time you appreciate the prescience. I have my own experience of this. I did two degrees in Australia, Asian Studies and Economics. One of my friends, did the same two degrees, and ended up at a mining company in 2000, at the very bottom of the market. He told me a story about how management asked the research department to forecast Chinese steel demand over the next 20 years. They looked at Japan, Korea and Taiwan, and assumed a similar growth pattern. They pretty much predicted China would be half of world steel demand, and would need to import 6 to 7 times the then current amount of seaborne iron ore the world produced at the time. Management sent the report back, and told them to put some more reasonable numbers in. In the end management were wrong, and the economists were right.

Why I mention this, is that humans have a habit of rejecting big numbers - so we need to keep that in mind with bearish AI narratives. But iron ore miners became shorts when the rate of increase in Chinese iron ore imports slowed around 2011, and supply took off. What is different here is Aschenbrenner is saying that order of magnitude growth will continue (with caveats about power and compute limitations). Knowing when Chinese demand for iron ore would slow was actually easy - you could use the same Japan and Korean analogies. With AI - we are flying blind. But at least I know what sort of data set that Elon Musk and Mark Zuckerberg are looking at when they make their USD 100bn investments.

Subscribe now

While I do believe you should never start with the idea that “equities are wrong”, when I see the performance of AI hardware against software companies this year I do laugh. Was it really only June when all software stocks were toast, and you HAD to be in hardware stocks? This trade has roundtripped.

I did find this amusing, until suddenly this month, the bond markets started to think the travails of Coreweave and other AI stocks was a deflationary event. Japanese 30 year yields fell nearly 40bp this month, and back to trend line.

Even more disturbing is GLD/TLT moved lower. Or yields fell, and gold fell. This is not the plan.

So could AI be deflationary? Well I will answer that question in a second, but I just want to point out how 100% inflationary politics remains. Lula won in Brazil by offering free handouts to voters. The French are rioting over underfunding of schools, and poorly conceived War in Iran has metasized into a messy regional war. Donald Trump offers USD 5,000 for voters in the mid-terms, and the Democrats are pondering tax cuts for middle and lower classes. Oil and natural gas prices are rising, as are semiconductor and copper prices. But despite all this, the travails of Coreweave make investors think deflation is more likely? Okay…..

So lets talk about inflation and AI. A bit hard to get deflationary as DRAM prices punch to new highs.

Also Coreweave is not struggling because its core business is getting squeezed, its is struggling because it wants to invest even more. This is fundamentally different from the failure of Lehman, or Bear Sterns, or even a Worldcom back in the day. All of those companies pricing for their assets fall. We are not seeing that with Coreweave. What we are seeing is existing capital suppliers worried that new capital raise is going to dilute them (my read anyway). We saw something similar with Silicon Valley Bank, which failed for financing reasons not asset reasons. This is completely in-line with my view of a market that is moving to a dearth of capital, from excess capital.

I had seen the AI boom as something like a capital war, or an arms race, where everyone is trying to bankrupt everyone else. This meant there was a risk that someone might “get rational” and the whole show ends. But I did read the Anthropic prospectus, and was really taken with their adherence to “scaling laws”. What is a scaling law? Basically, the more compute the better the system. So if you want to have a great AI, you need to throw a tonne of money at it - and the more the better. I have heard this before, but was sceptical - and particularly when Deepseek was launched. This seemed to prove the opposite, but since DeepSeek, the hyperscalers have ramped up capex. That is the Anthropic scaling law prediction seems to have been borne out.

In other words, bigger is always better, and clever is not that much better. This is very different to how most industries operate. Typically, once a product has been invented and developed, later entrants can replicate at lower cost. Its the whole points of patents and intellectual property to stop this happening. But the implication here, is ONLY the biggest spender can have the best AI. We can see that Anthropic has convinced the other hyperscalers of this.

So when Deepseek came out, this was an attack on the scaling laws, as it was put together much cheaper than ChatGPT or OpenAI models - but Anthropic in essence saying, who ever spends the most WILL have the best model. It implies that the biggest spenders CANNOT be undercut by cheaper versions, as they will not have spent enough. This means it is a capital war, not a price war, but not driven by ego, or competition, but by the nature of the LLM model. This does seem wild, but Mark Zuckerberg, Elon Musk and the entire IT industry believes it, so who am I to argue. This does imply a winner takes all industry ultimately, so the capex makes sense from that perspective. This is in line with my original thoughts. The issue the market was concerned with was how much longer would it have to fund AI. But Anthropic’s quarterly data seems to imply that it can now generate more revenue than it spends on compute - which is incredibly bullish. It implies is has the revenue to keep investing to have the best model, and become potentially self financing. Of course its raising funds to invest even more heavily - but as proof of model its very bullish.

I CAN make a bear case of Anthropic, but it is tenuous. I see loads of new AI models and LLMs being released all the time. And this training requires a large amount of compute. It could be possible that the revenue growth for Anthropic and OpenAI is totally or largely reliant on other model makers. The bear case would be, as Anthropic and probably OpenAI come to dominate, eventually all the other model makers give up, and suddenly compute demand falls off a cliff. This could be possible - certainly I seem to have a lot of different AI models I can use now - ChatGPT (which I pay for), Grok, Co-Pilot and even a Bloomberg version now. This is a risk - but one I find hard to see. Instead, we see Meta launch Muse, and the next thing we see if Meta wants to issue USD 100bn in equity. For this bear case to hold water requires the suddenly pulling of capital for LLM training and model making, something has has nothing do with Coreweave.

What I see in DRAM, Coreweave, corporate bond markets, copper and oil, is an age of supply constraint. What this really needs is for governments to cut spending aggressively. I see no political chance of that happening BEFORE a debt crisis. France is talking about it now, but the crisis needs to happen first. More likely is spending and taxation goes up - and my guess wealth taxes are coming. Why wealth taxes? That is where the money is.

Does the AI trade recover? Maybe. Do precious metals go higher? Maybe. Do yields go higher? And Maybe. But I do know inflation is going higher - and how governments react to it, will determine what these various assets do. All I know is which of these assets will do relatively better in a rising inflation environment.

Subscribe now

Many years ago I met a short only fund manager at a conference. He was a highly strung character, who constantly checked his phone. Someone asked what do you do when someone acquires a company you are short. It was a good question I thought - you would be surprised how often this happens, often by a company you are long off. His comment - which has stayed with me all these years - is that when this happens, it is “an axe to the forehead - an occupational hazard”.

Now I don’t believe in relative trading in a single industry, for exactly the reason above. I see industries as either thriving or dying. Which is why I think neo-clouds make very tough shorts. Why? Neo-clouds looks like potential axes to foreheads.

Amazon and Elon Musk have confirmed demand is strong. If we take Coreweave - their main customer is Microsoft. Now they could let Coreweave go bust and then come in and buy good assets at knockdown prices. But what happens if a competitor comes in early to bid. xAI is one. Private equity is another - especially if they can reissue debt at lower yields. For Microsoft, a cloud leader, would allowing competitors to build up capacity be wise? I think not. In fact, I thought we would come into markets today, with some resolution. Maybe the 12% interday swing is a sign.

I have said to pretty much anyone who asks me - I am neither bullish or bearish on equities. What I see is revenue growth offset by rising costs and rising interest rates/yields. Coreweave is an example. For the longest time - it seemed that rising yields had only temporary effect on asset markets. We did have the sell off in 2022. We have also seen various “yield led” blow-ups. Silicon Valley Bank blew up not on bad loans - but lack of deposits. And now we see NeoClouds blowing up - not on falling revenue (rental rates are up!) but on large unfunded capex plans. To avoid “an axe to the forehead” you want the whole industry to be in trouble - and I don't see it. Not yet anyway. The time to get mega bearish would be if Microsoft bonds were falling They have rallied this month.

Why are markets so tricky? The pro-labour trade makes you want to own cyclical assets over defensive assets - in complete reversal of what has been correct for the past 40 years. GLD/TLT is an expression of that. And this month, it has been an axe to the forehead in a way. Suddenly bonds have rallied, but gold has not done anything.

But the biggest axe to the forehead has been Brazil. A surprising good result for Bolsonaro in the first round saw Brazilian assets leap higher. But when Lula actually won, assets have fallen below where they were trading before the first round result. EWZ US - is a USD ETF for Brazil. Axe to forehead indeed.

But the axes have been flying this year. Remember when software companies were going to default, and you needed all your money in HALO stocks? That trade has roundtripped.

From what I hear, the whole active management community is struggling this month. And for good reason. US bonds were literally telling you the cycle would continue, before suddenly deciding actually its deflation again.

I have had plenty of axe to forehead moments in my career. Sometimes its because there is a market surprise, and everyone is positioned the same way, and they trip over themselves to liquidate. But sometimes its because something more fundamental has changed. My analysis, for what its worth, is that it is the first kind. I think calling an end to the AI trade, when Elon Musk and Amazon are saying demand is surging and pricing is improving is very risky. Just as it was it was risky to claim that Brazil was definitely moving right wing. Ultimately, everything I look at suggests we are still in a inflationary left wing move. Its just markets decided everyone needed an axe to the forehead.

Subscribe now

I have been looking at markets, and wondering why something like GLD/TLT looks like the SOX. GLT/TLT below.

And here is the SOX.

I know that the macro guys are super bearish. And they are bearish because of credit. We have French CDS.

And we have Coreweave CDS

So the macro view, and why markets are getting pummelled is that the markets are telling France to cut spending or raise taxes, and the market are telling Coreweave to raise equity and cut capex. This signal is so clear, we assume it must happen. So markets look at this signal, and assume deflation/depression. The final nail in the coffin is comparing AI to other build outs.

And this then becomes a no brainer short. AI makes no logical sense, so sell sell sell. And from a macro point of view, it is VERY hard to argue with.

But from a political point of view, I think the railway or dot com comparison is wrong. What we have seen in the Ukraine-Russia war is that AI and Starlink are strategic assets. Modern war, and modern power is build on satellites and compute. I think the better comparison is the build out of nuclear warheads after World War II. At that point, it was not question of how many warheads do we need? But whether we had more than the other guy. In fact Reagan explicitly increased spending to successful bankrupt the USSR.

What markets are saying, in my mind, is the AI spend, and government spend is so much, we would like you to stop. So we are going to raise the cost of capital. But what I see, looking at politics, and reading SpaceX and Meta quarterly results, they are like, we don’t really care what you think - we are going to keep spending. If you like, you can think of AI as defence spending. From that perspective, there is way more spending that could be done. Defence spending used to be 10% of GDP.

From a political point of view, arguing for a AI bust, and deflationary outlook seems mad to me. Now I could be wrong, and this month is saying I am wrong, but my favourite lead indicator, JGBs has tended to move earlier than treasuries for both inflation (yellow circles) and deflation (red circles). It was showing no signs of concern.

Of course bond markets can be stupid, so I also look at Japanese banks, which are extremely sensitive to inflation/deflation pulses. Japanese banks rolling over on a relative basis tends to mark a strong deflation cycle. You could argue maybe there was a topping signal, but hard to see a rollover.

In essence, we are entering a “war period” which means growth is good, and wasteful investment abounds, and capital gets destroyed. This reasoning also explains why copper has not been bothered by AI shenanigans.

If you accept this political reason - then what is going to happen to Coreweave? Well either the US government steps in, or someone buys it. The US still needs more compute, just like it needed that extra nuke. And what about France? Well as the only truly independent nuclear arsenal in Western Europe, I am pretty sure a deal will get cut at some point. I must say I find the political argument far more convincing than the macro argument.

Subscribe now

Everywhere I look in markets is pricing the end of the AI trade. It is even turning up in sovereign bonds, with yields now falling, after rising for most of the year. I find it a bit odd, as every physical commodity price I look at is still going up. The issue here is that you have to be sure that equities are accurately pricing the future. SK Hynix has typically been pretty responsive to spot DRAM prices. Not for the last three months.

Normally, starting from a point of “equities are wrong” is a recipe for disaster. But I can’t help but come back to some weird activity. One area in the AI space really confounds me.

Read more

I have written about AI and Coreweave a lot in the last few days. And I have received a lot of replies, almost universally bearish, and from a lot of managers and investors that I respect. And to be honest, if gives me pause. So I thought I should give some more detail on how I think. I am not a short term trader. When I go bearish on something, I expect it to fall for quite some time. I look at both demand and supply, and basically figure out when profits in an industry are going to go to zero. So for me a perfect bear case is when demand looks to be falling just as supply is coming on line.

A good example has been iron ore in 2011, or oil in 2013, or restaurant franchises in recent years. My problem with being bearish AI, is that there is no clear sign that demand is falling. If anything, the opposite is true. The other bear argument is that there is excess capacity. Again, if this was true, pricing would be weakening, and I cannot find that. One area, that does look interesting is that financing is getting scarce. And I can get fully on board with that idea. The problem of course is that the underlying collateral has value, so it makes for a tricky short.

Why do I mention this? Well short selling is dangerous. And it is particularly dangerous when consensus. For years, the market was bearish on Tesla. Social media was full of all the financial shenanigans that Elon Musk got up to, and how he needed capital. What happened? Tesla became the most shorted stock in the S&P 500 before rising 3000%.

What happened to all those Tesla bears? Well hard to believe, but most of them write Substacks now.

Many people were also bearish on crypto and bitcoin, for very sensible reasons. And we all know what happened there.

So when I think something might be a short - I like it to have three qualities.

1 - none to little short interest.

2 - almost no bearish consensus.

3 - a committed share holder base.

4 - share price weakness in suppliers or customers

And I avoid the following

1 - a high cost to borrow

2 - “never short a dream”

3 - key customers or suppliers shares prices are surging.

Now just because Coreweave does not meet my standards as a short, does not mean its not a short. But what I can tell you for free in my career, is that if it is not a short (that is if it goes up instead of down) it could go up a LONG way. That is the risk reward is poor. I have no position in Coreweave, I am just intrigued by it.

Subscribe now

A lot of the price action in markets after the problems with Coreweave have been deflationary, which sort of make sense. GLD/TLT, which is my big inflation trade is looking problematic.

Read more

(Apologies for this post - it is slightly more stream of consciousness that I like, but it is my thought process. Will follow up with more clear logic)

When something does not make sense to me, my subconscious keeps nagging away at me. Generally speaking I trust my subconscious - its gets things right. If I wake up in the middle of the night, worried about a position, then I know its probably too big or just wrong. My best positions, both long and short are always when I am so excited, I just need to get the position on NOW. Historically speaking, the price action at Coreweave would normally motivate me to get heavily short Nvidia and the whole AI trade. And yet my subconscious stops me. This note is trying to work out why.

Historically speaking, the set up look a slam dunk short on Nvidia. For example, many years ago I wondered why William Brother a pipeline company, whose main client was Chesapeake was doing so well, while Chesapeake was crashing. After taking a good look, I discovered that the deal Chesapeake had with William Brother was a fixed quarterly payment, no matter how much gas they piped. When the deal was signed, it was a win-win. Williams could show its investors steady revenue, and Chesapeake could cut costs by pumping as much gas as possible. The problem was that Chesapeake and its competitors where piping so much gas, prices were collapsing, and something had to give. I shorted William Brother confident there was limited upside, and big downside. I was correct. However, working out why the market was so divergent did take a couple of weeks digging.

I am getting a similar confusion when I look at Coreweave, which I know is telling me something, but I just don’t know what. This always bugs me. First of all, Coreweave stock has been weak, but it is back to levels seen in July. It has been clear the market is struggling to price this business.

But on the other hand, its CDS are at new highs of 1128 - and a level that implies serious concerns over its debt. This is most unusual. I would expect a stock price to be collapsing with the CDS this high. In fact, Coreweave trades at 6 times book.

Is Coreweave reacting to problems with its customers? Well here is a little tricky, because its main customers also have data centres, but use neoclouds when they need extra capacity. Microsoft share price is close to highs. So that does not seem the problem. I wondered if neoclouds were getting squeezed by the hyperscalers, but pricing seems to be rising for neo clouds and hyperscalers.

And if GPU rental prices are rising, its pretty safe to say GPU prices are stable too. So that’s not the problem. I think the problem comes for Coreweave outlook. They have been investing aggressively this year, with Capex in the first two quarters of near USD 15bn.

But in their most recent outlook statement, saw up to USD13.5bn for Q3, and up to USD39bn for all of 2026.

So Coreweave CDS seems to be reacting purely to capex plans.

And now I know why Coreweave bugs me. Its share prices is falling because it has huge capex plans - which is good for Nvidia and all the other hardware stocks. Try as I might, I cannot find any data that shows falling prices in GPUs or AI costs. That is, unlike the William Brother/Chesapeake example above, there is no sign of a price war, or over supply. In fact PPI for semiconductors says the total opposite, as does DRAM spot prices.

More logically, the market is probably pricing in an equity issuance. With a current market cap of USD 36bn, and guided capex of USD 36bn for 2026, and Barron’s suggested US 50bn for 2027. This level of capex would undoubtedly make current equity holders and debt holders nervous. I have had my concerns of AI capex before, and then along came the SpaceX IPO, which delivered Elon Musk the capital he needed to be build xAI. That is equity investors seem happy to fund at least Elon’s capex plans.

This is what I would expect to see from the “capital war” that is being waged in AI. But this “capital war” is happening in a time that pools of capital are being run down. The question is when do we run out of capital? I am pretty sure, we can say that the credit markets are close to being tapped out. The question is whether the equity markets are tapped out or not. Now this is where it gets “weird”. Elon Musk has no problem tapping equity markets for capital, but the rest of the tech stack does. Nvidia is a good example. In the most recent quarter, its account receivables went up as it changed payment terms. Despite this, it should be noted that Nvidia also bought back share this quarter, to make USD 50bn of share repurchases this year.

And now we get to the heart of the matter. I have puzzled over this chart for some time. The cost of capital rises, and yet equities get more expensive.

Or a better way to understand it, is that US equity valuations are built on an understanding that share buybacks will continue no matter what relative valuations look like. Apple, which has conspicuously not entered the AI race, and essentially returns all cashflow to share holders, has just touched a new all time high. When Apple started its share buybacks, it was a no brainer trade. Is it still a no brainer trade today? Does it even matter? As long as share buybacks push the share price up, its MUST be considered value creating.

What I am trying to say, is that modern US capitalism has become extremely focused on share holder return, with a heavy focus on share buybacks. This has led to a greater focus on raising capital via debt markets than via equity. If we look at shares outstanding on the S&P 500, it still bears no resemblance to the massive issuances we saw in 2000.

What I am seeing is the culmination of two trends in US business. One if the blitzscaling model - that of massive capex to gain market share, and the other of creating share holder value via accessing the debt markets rather then equity markets. Both are now under pressure. First the blitzscaling model does not work if everyone does it. The idea of blitzscaling is to give you a dominant position, which allows you to set and raise prices. This is not happening yet. And the second as Coreweave shows, when everyone is blitzscaling, the debt markets are correct to get nervous.

What does that mean for markets? Well its hard to see neocloud’s cutting capex. If you cut capex you admit you will be a subscale producer, and you are dead anyway. Oracle share prices has been under pressure for a year now, and no capex cut has been forthcoming. In some ways, US business represents the political position we are in now. Everyone HAS to spend, as they cannot afford to lose. The true meaning of Coreweave is that yields go much higher - for everyone.

Subscribe now

My investment process is built around looking at how other investors have made mistakes from time to time, and then building a process to avoid that mistake, and if possible to profit it from it (which is big driver of my shorting process). In my first half of my career, I was heavily macro, as I found currencies were the main driver of investor mistakes. I did my best to quantify this process by formalising the use of NIIP and BIS cross country data. I think I was one of the first investors to do this. But during Covid and in 2021, I discovered that politics was a far more effective in predicting market moves. And a move to populist politics made me suggest GLD/TLT (long gold/short treasuries) was a good trade.

But when I was researching politics as a driver of markets, I learnt a very awkward fact. Politics can change 5 to 10 years BEFORE markets react. That is the gap between politics changing and markets reacting can be quite long (a lifetime in modern fund management). I will start with most recent examples and work backwards.

CHINESE TECH - 2021

The Chinese tech index, Hang Seng Tech index has been in a prolonged bear market since 2021. For most US investors, they may have been exposed to this via a BABA US or KWEB US long position.

Chinese politics started to turn against its internet giants in 2019, initially limiting gaming time for children. But in November 2020, Chinese regulators cancelled the ANT financial IPO. Due to the raging bull market in tech, most investors ignored these salvos, and the tech index continued to rise. But there was a steady drip of new of regulatory backlash. In this case, the changing politics was becoming obvious in 2020, but the market rallied another 50% before finally cracking.

BREXIT/TRUMP - 2016

The Brexit referendum and the first Trump election win were clear signs of a political realignment in the West, a realignment that continues to this day. At its heart were traditional left wing voters being up for grabs through a combination of social conservativism and fiscal liberalism. When Trump was first elected, the bond markets recognised the inflationary nature of his politics, which saw yields sell off sharply, but going fully bearish Treasuries then would have crushed you come 2019/2020.

As mentioned numerous times, I found the JGB markets better lead on US treasury than anything else. In 2020, JGB yields sold off well before the US, and have continued to do so. This implies that Japanese investors are much more sensitive to US politics than American investors. I think there is a good reason for this. See the section below on Japan.

Japan - 1990

For almost three decades the Nikkei was a byword for economic mismanagement and stagnation. Only a few years ago did it finally breach the high set in 1989.

But for me, the Nikkei is almost 100% political driven. From a virtual standing start, Japan built a world conquering semiconductor industry in the 1980s.

The rise of Japan created severe misgiving in the US, and led to a number of political agreements. The key one for me is the 1986, US-Japan semiconductor agreement. The main part of this agreement was that Japan agreed not to take any more market share from the Americans. The other, and much more famous agreement, was the Plaza accord, where the Japanese agreed to let the Yen appreciate. Down in this chart mean Yen appreciation.

You would think an agreement to kneecap your most important industry and to allow a huge rise in your currency would be bad for markets, and you would have been right. But the Nikkei doubled from 1986 to 1990, before breaking lower. In this case there was a four year gap between changing politics and market recognition.

LEAVING GOLD STANDARD - 1972

This is the wildest analysis of all. The 1970s are known as the stagflation era, where high inflation and interest rates acted as a break on stock markets and growth. For many macro investors, leaving the US leaving the gold standard was THE cause of inflation. But politically I would argue the opposite. When a country devalues, it is reducing the wages of its workers in the global market place. So while it may be temporarily inflationary locally, its overall effect is deflationary.

I know a lot of people are going to struggle with the idea of devaluing as being deflationary. But if you look at the Great Depression, it was “beggar thy neighbour devaluations” that kept prices in check globally. The Great Depression ended when the US agreed to a strong currency, and from there we entered an inflationary era, but also an era of higher bond yields. What this means in practice, was that Nixon taking the US off the gold standard, was for the benefit of corporates, not workers. However the top in US yields did not happen for another 8 years. Of course for other countries, yields peaked when the US left the gold standard. The all time high in Swiss ten year bonds was in 1974.

CONCLUSION

I think the current political set up, which heavily favour billionaires and stock markets looks unsustainable. Taxes will have to rise, and most likely they will rise on wealth, rather than wages would be my read. But just because the politics has changed, markets can keep moving the opposite direction for sometime, even nearly a decade. Politically, this makes me favour gold over stocks. And in 2025, this started to come true, but in 2026 substantially reversed.

Markets have reversed this year, but my read of politics remains the same. But as the above shows, sometimes it takes times for politics to fully leak to markets.

Subscribe now

I had barely sent out an update on GLD/TLT when market action has come to test me. If you have been living under a rock, the AI trade has swooned as the bonds of Coreweave has been sold aggressively. They are now yielding 15%, or perhaps more importantly, as a big discount to par - signalling fears of default.

I took a good look at Coreweave, and came up with two totally different outcomes. In one we enter a deflationary spiral in the tech space, with huge financial repercussions. Coreweave goes bust, and sets off a doom loop in the AI trade. In the other, inflation continues unchecked, and Coreweave is a buy here. GLD/TLT should point the way here, but as it happens, technically GLD/TLT is right in no man’s land at the moment.

So let me explain the logic of these two outcomes, and which I think is more likely.

Read more

Years ago when I came up with GLD/TLT, I thought it was a great trade - but super simplistic. So simplistic, I thought people would laugh at me. But I have been pleasantly surprised at how many people have been influenced by this trade. I have even seen people use it in their own research. Now this is all great and dandy, but I had another investment catchphrase back in the day - “long bonds/short equities” which worked great, until it didn’t. But this is a risk for all managers - they become wed to an idea that is no longer valid. So even though GLD/TLT has been good to me, I am acutely aware that one day I will need to do a “Reverse Lawrence Stroll (aka Kaiser Soze)”, which is I will need to cut my progeny off, and move to greener pastures. But is that day today?

Read more

Back in 2025, I shorted Nvidia on the back of the DeepSeek LLM model - but fortunately covered at a profit. It seemed very logical to me that the appearance of a much cheaper LLM model should be deflationary across the whole AI stack. But this analysis was wrong.

Today, we are seeing another sell off in tech, on the back of announcement that Chinese chip company CXMT has made a breakthrough in memory chips.

To be honest, Deepseek methods seemed to me to be very replicable, and so very likely to change the industry, while CXMT claims are very hard to verify, and also much easier to be kept out of the western AI chain. In other words, if DeepSeek was not a problem, then I struggle to think that CXMT will be either. That being said, its a good time to run through my AI checklist to see if its time to either get out of the way, buy the dip or that greatest feeling of all, time to short!

Read more

There is a widely held belief that the cover of the Economist is a contra indicator. The Economist itself also covers this question in an article. The quantitative analysis concludes no. For me, I feel like the cover is too narrow. I feel you need to read the main lead article to see what their logic is to really get an idea of it contra indicator ability. For me, reading the article, it feels generally good news I think, as least in regards to short TLT.

Read more

( I sat down today to try and work my way through the AI conundrum - below is a stream of consciousness style analysis - I suspect some with think me insane, and some will think me genius. Time will tell.)

The weird thing about the AI boom is how familiar it seems to me. Back in the day, I made my name shorting mining and emerging markets, based on a very simple observation. I was looking at Chinese production of steel rising almost continuously from 2000.

But I noticed from 2008 onwards, Chinese steel companies share prices were moving in the opposite direction to steel production. Basically, over production in China was driving margins down

China rebar (type of steel) prices are published daily, and when they started to fall in 2011, I knew that the entire industrial commodity (iron ore, coking coal, copper, nickel) trade was at risk.

This is capitalism at is finest, and best (in my view). What is so fascinating about AI, is that in China, they seem to have embarked on a similar price war model with tech. Every time I look at AI, Chinese AI pricing is much cheaper. See DeepSeek and MiMo below.

And we see it in the performance of the Hang Seng Tech Index - which basically is so uncorrelated to either the Nasdaq or the Semiconductor index, you wonder if its really tech and not actually steel related!

So like steel before it, Chinese tech industry competes via price. This is deflationary, and we see this in Chinese bond yields. Remember this is China, the other tech superpower, and its bond yields are low, and its stock market moribund.

I will not put up graphs of the US stock market or bonds yields, everyone knows they are the complete opposite of China. How can we explain the difference? I think where as Chinese corporates are competing on price, US corporates are competing on capital. Who can invest the most, rather than who is the cheapest. This has lead to a capex boom.

As I tell everyone, I think this is rational behaviour. If I saw Elon Musk entering my industry, I would also spend everything I could, because the other option would be eventual market share loss and decline, as we have seen with legacy car makers who failed to invest heavily in EVs. The problem for the legacy hyperscalers is that SpaceX has a USD 2 trillion market cap, for its current USD 55bn in revenue, which Musk can use to raise capital. He apparently wants to raise USD40bn. This is in addition to USD 20bn raised in corporate debt markets. For the hyperscalers, with cashflow, they generally do not issue equity (buybacks are more common), so they have also become heavily reliant on the bond market.

And here is the big question. Do rising tech spreads indicate credit concerns? Or that the capital borrowed is so great, that they need to offer higher prices to get investors to lend to them?

If I wanted to, I could easily paint this as tech rerun of the Chinese steel overcapacity story. And I am very tempted to do so, as many of financial guys on Substack has done, because its great marketing. But for me, I have a big problem. There is no sign that capital markets are buckling. What we have seen is equity do well even as high yield does poorly (macro trading has been poor of late due to these breakdowns).

What I find fascinating as I think through this, is that corporate bond investors are basically now creating the inflation that is eating their principal. What I mean is that inflation is now being created by corporate bond issuance - not government bond issuance. Normally credit spread widening is a deflationary event. 1991 - Savings and Loans, 1998 - Asian Financial Crisis, 2007 - GFC, 2015 - China Deval, 2020 - Covid.

But as noted above, we are seeing spread widening on RECORD issuance of tech related bonds. This issuance is going straight into data centres and Nvidia chips, which drives growth, and inflation. The spread widening is INFLATIONARY. Hence, when investors buy tech bonds, they are creating inflation, which also hurts treasury yields.

Most financial chat has been about how Oracle CDS blow out marks the end of the AI trade (so far so wrong).

If AI is driving inflation, then we can rely on our tried and tested deflation lead indicator, the 10 year JGB. Its track record in picking deflationary events remains unrivalled. Usually with credit concerns, as seen with HYG or Oracle CDS, you would expect JGBs to rally (ie yields fall) - but not this time, and this above analysis explains why.

Amusingly, if seems if governments want lower bond yields, its looks like they would need to regulate the tech companies, as China has done. The weird thing is that this would be politically popular with voters, but difficult in practice given the amount of money in US politics these days.

Subscribe now

Without question the worst investors over the last ten years have been macro investors. The one trade I probably talk about the most on this substack has been long Yen, which I now call the widowmaker trade. There was a time when Yen would rally when US stocks were weak - but not anymore.

Another foolish trade from the macro community is that higher Japanese and US bond yields would have some sort of negative effect on US stock valuations. Very wrong indeed.

Macro investors used to insist that large fiscal deficits would lead to a weak currency. Brazil has a 10% budget deficit, which outside of Covid is the highest deficit since the 1990s.

And yet the Brazilian Real remains near levels first seen in 2002.

The macro community have been leaning heavily on comparisons with past investment booms to drive a bearish view on AI. This chart from NBER is indicative.

And yet this fear of overinvestment that stalks the market, all I see is pricing in the compute space still rising.

As are DRAM prices.

Some of the other more amusing macro calls has been bullishness on Chinese tech. And yet, this somehow also does not work.

The reason I am pointing this all out, is that I too used to use macro thinking for investing. But, it has not been a winning strategy for a long time now. The best reason I can come up for it not working is that governments now operate much more openly to secure desired economic outcomes. Both the Chinese and West are acting to keep Chinese tech down, so that does not work. The Americans have only just showed some interest in strengthening the Yen, after years of benign neglect. And the market is obviously saying that fiscal spend is bad for bonds, but good for equities. Macro is good, but when it does not align with politics, its a waste of time.

Subscribe now

A few months ago I wrote an article on whether declining population rates was following the pattern of the horse population in Europe a hundred years ago. The key graphic is below. The main metric I focused on was the decline in draft horses - which was caused by the rise in motorised transport.

What I really disliked about this comparison was that it was at odds with my general view that humans are programmed to reproduce, while domesticated horse populations are controlled directly by humans. That is horses have no free will, while humans do, and should want to reproduce. But one thing that has been stirring in my mind about this graphic, is the growth in thoroughbreds at the expense of draft horses. When I say thoroughbreds, I mean horses used for sport and leisure. Lets us assume that the global human population is going through a similar shift, where the returns to physical labour is declining, so humans need to focus more on “sport and leisure”.

So I have been thinking about what a transition from a human as draft horse era, to a human as thoroughbred era might look like? A small caveat, I don’t like thinking in these terms, but as an intellectual exercise it does have some explanatory power, so here goes. In the draft horse era, I imagine breeding was rather haphazard, where stables would breed draft horses, and sell the ones they did not need. In the thoroughbred era, breeding is far more select, with only the top studs allowed to reproduce at all. You could argue that we have seen a similar behaviour in the West. The rise of young men who have not had sex (incels) has been striking.

Of course no human wants to be a “draft horse” who gets no action, so a natural human reaction would be to spend, and particularly to spend on their progeny to move them from draft horse to thoroughbred status. Certainly college tuition CPI supports such a view.

As a parent that has one son doing high level motorsport (www.jarrettclarkracing.com) and another at a football academy ( blak33959 on Tiktok) I can tell you the time and money commitment I see from parents to get their children to be an elite athlete is phenomenal. Of course, the number of professional athletes is capped, but what I see is that often people can take professional training and turn that into a social media career. The number of digital creators is rising rapidly, not a perfect measure - but an influencer is definitely more a thoroughbred than a draft horse.

Or to put in another way, goods producing employment (draft horse equivalent) peaked in the 1970s.

One of the reasons I am (was?) hostile to this line of thought is that it implies that I and my cohort (I was born in 1974) were never meant to be “thoroughbreds” but “draft horses”. While I don’t like that thought process, when I reflect on my fellow students in primary school and high school, aspirations were not that high. In fact, for me it only occurred to me that a career in finance might be the way to go only happened in 1998. And to be fair, I did well at school with almost no effort (being naturally gifted at maths was more than enough back then). This is not the case anymore - not at the schools that I am familiar with in London in any case.

Anyway, what can we take from this and apply to fertility rates? Well the shift from draft horse to thoroughbred culture means that women will be far more picky in the choosing partners, something we see in the data already. But what we also see in the horse data at some point, the population of horses can rise again, once the draft horse population has shrunk. In the case of horses, mechanised transport took off in the 1920s, and horses live for about 25 years. This means it took about two generations to get the draft horse population down. With humans, around 1970 is when it looks like machinery started to replace humans in a meaningful way. Assuming a 70 year life span (maybe too conservative) we are still in the acute phase of population adjustment.

One other final thought is that modern financial markets have also undergone a draft horse to thoroughbred transformation. VC investing is often about supporting a talented individual - for example giving a start up USD 2m at USD50m valuation to motivate them to see what they can do. If you had invested purely on a belief that Elon Musk would succeed and ignored all traditional financial analysis you have done very well for yourself.

What does all this mean? It explains a big generational divide. Youngsters are thoroughbreds used to be pampered to achieve maximum performance, while older generations are accustomed to toil and inconvenience. But it is also a fundamentally bullish story. Replacing horses with mechanised transport was GDP creating. Replacing human draft horses with thoroughbreds should be similarly GDP creating. But just like every obsolescent model - its hard to believe the world has moved on without me.

Subscribe now

September was a long month - but a satisfying month. It proved, to my mind at least, that political analysis is superior to old school analysis. First and foremost in this was the merit of GLD/TLT. I think many “investors” and “macro commentators” find it much easier to go all in on debasement trades. But politically I found it much more difficult. There is always the chance that politics or circumstances forced central banks and governments to run much tighter money than expected. When gold was surging last year and early this year, I did say I felt it was time for the short TLT side of this trade to exert itself. And over the last few months it has.

Gold and other precious metals have and remain lacklustre. But that’s the whole point of GLD/TLT - you can only jump and shout about debasement IF yields are not rising. I know logical and rational thought is at a premium these days (Donald Trump and Elon Musk rallying against the establishment strikes me as odd, when one is the most powerful politician on the planet, and the other is the richest man on the planet - if they are not establishment - then who is? If establishment is holding them back, then how are they so powerful and rich? Like I said logical thought is at a premium these days).

Read more

In case you have been living under a rock, French bonds have been under pressure. Typically, sovereign bonds cannot default, but with European bonds there is a risk that you are forced to leave Eurozone, so with European bonds, CDS can actually work as an indicator. French CDS have soared in recent weeks.

France has a lot of debt - but so do most countries. But it does have a budget deficit that is very high by European standards. 3% is the supposed limit.

One of the weird things about the “French bond crisis” is that until this month, Euro was strengthening against the Swiss Franc.

Since the beginning of debt problems in 2007, you have tended to want to be long Swiss Franc vs Euro - but as of today, CHF is no stronger versus Euro than in 2024. Very odd behaviour indeed. Why?

One reason may be the French political reaction. I was stunned to see that Marine Le Pen has suggested a debt brake be added to the constitution.

I find this very bullish for the French bonds if implemented (big if), but also seemingly send a big signal that Le Pen is not seeking to leave the Euro. This is also “deflationary”, and has been reflected in the stock market.

What I find so interesting about the current set up, is that the macro signals are very bearish Euro, while the political signals would be bullish. Eurozone debt levels as a whole are very manageable.

In both Italy, and now potentially France, the populist leaders have signalled a preference for fiscal consolidation. The biggest headwind for the Euro has been very expensive energy prices. TTF Gas is back at USD 24, a big premium to the US prices, and so a competitive disadvantage.

But while higher energy prices led to collapsing activity in Germany in 2022 (most reliant on cheap energy), in 2026, there has been no noticeable decline in PMIs in Germany.

All of this reminds me of something that I learnt when first starting to managing money. Almost all US statistical releases get revised downwards, and almost all European statistical releases get revised upwards. Here, we are focused on the price of French bonds, but not on the political response or even the surprising strength of European economic data.

Old school macro models will probably be bearish Euro here, while I think the politics is bullish. For the last few years, political analysis has won over macro analysis. Lets see if that happens again.

Subscribe now

I just got back from a couple of days in Switzerland, where I spent most of my time talking about markets. I also met quite a few Substack subscribers. Here are my thoughts.

SUBSTACK

Substack tells me that my average article gets 8,000 views. I thought this was exaggerated as the usual 10 to 15 people are commenting on it, and a typical article now only generates maybe 5 or 10 new subscribers. I suspected that number of people that actually read my substack was lower than what I was told. This trip has made me think the opposite. Almost everyone I met in Switzerland did read the Substack. And given that I only have 500 subs in Switzerland (or so I am told) - I think the Substack reach is more substantial that I thought. And thank you for everyone who turned up for a beer and a coffee. It was great meeting you all.

TLT

In case you have been living under a rock, TLT got smashed in September. This was nice, as I had suggested earlier this year, it was time for TLT to do more work in the GLD/TLT trade, and so it has come to pass.

The question to be answered now, is the TLT trade done? I think there are two ways to approach this question. Lets call its the mathematical, logic driven approach, and the political approach. We will start with the mathematical approach. 30 year Treasury yields have risen from 1% to 5.6% - a substantial move. But in 1970s and 1980s, they were well above 10%. So yields could go higher.

TLT has a duration of 15.

And a current yield of 5% (this is different to the current yield in the bond market, because TLT owns bonds with different coupon rates).

Duration tells you how much capital values fall from a rise in yields. So if yields rose 1%, the TLT would fall 15%. But it pays a yield of 5% - so you only get 10% total return from a short. By definition, we will get less bang for a buck with a bearish view on TLT. Back of envelope calculation makes me think at 8%, TLT will no longer be an effective short as higher coupons reduce the duration and the coupons will overwhelm the fall in capital values. If anyone has a better calculation, please get in touch. One thing to remember with TLT, it has no pull to par, so rising yield environments will see PERMANENT capital impairment. So TLT is still good - not as good as it was, but good. It also pleases me to see the share outstanding in TLT rise again.

So mathematically, TLT is still good to go. But its politically that really gets me going. This time next year, we will start thinking about who will be running for the White House in 2028. According to betting markets, AOC is the most likely Democrat candidate. Even if she does not win the primaries, the presence of AOC should drag all candidates to the hard left.

What is most likely then? I think bond yields go higher, and potentially USD lower. Unless I see a sudden shift right in US politics, a short TLT position still makes sense.

CHF

I was hoping my trip to Switzerland would give me clarity on what to do with the Swiss Franc. I want to be bullish on the Swiss Franc. When compared to the Great British Pound, over almost any time frame the Swiss Franc has been a much better currency.

If the main currency of the world was Pounds, owning Swiss Francs would be a no brainer. But sadly the default option is US dollars. Against the US dollar the Swiss Franc still has an appreciating bias, but not as reliable. It tends to have big moves like 1970s or 2000s, and then spends a long time in a range, before then breaking higher again. The Swiss Franc peaked against the US dollar 0.72 in 2011, and has been in a range since then.

Complicating the view on Swiss Franc is the long term Japanese Yen v Swiss Franc Chart. There is no question that the Yen is cheap versus the Swiss Franc, or that Japan is cheap now. It is hard not to look at this chart and think perhaps I should be long Yen and short Swiss Franc.

If that was the tactical set up, what was Swiss prices like? Well Switzerland IS expensive, but in line with London on some things, and cheaper in some ways too. The only place I felt that prices were total out of whack was Geneva airport - but airport prices don’t count. A beer at the train station as around 9CHF - so 8GBP. Expensive - but London prices. But Switzerland is a financial centre - it should be expensive.

So what to do with Swiss Franc? Well it reminds me of gold. Everyone is keen play debasement trades (long Swiss, long precious metal, long crypto) and I understand that. The issue is that if US bonds yields are rising, then the debasement trades loose water - and that is what we see. US 2 years up - gold, Swiss Franc down.

So its a political call - and if you feel confused, you should. The political sounds out of the US are very confusing. President Trump wants lower rates, and Bessent is the House on Yen and does bond buybacks, but the Fed is suddenly hawkish. Ultimately, all I can say, is there is a reason I said GLD/TLT and not long GLD or short TLT. These themes need each other to hedge out political risk, which is what last month showed. Given what I said about US primaries starting next year, long CHF still makes sense.

Subscribe now

I will be at the NZZ bar from 4.30 to 6pm at Zurich HB tomorrow, and Café Tartines - Champel in Geneva from 8am to 9am on Wednesday if you would like to say hello.

I have tended to make fun of “macro” analysts on this Substack - mainly because they are such easy targets. One of their consistent problems is that they apply economic or accounting logic to sovereign bond markets. Through the mega bull market in bonds, were bearish all the way up (or as yields went lower). Short JGB trade was called the widowmaker because so many of these analysts convinced hedge funds there was no way 10 year JGB yields could fall below 2%. They or course went to negative yields for a while.

Now, the “macro” analysts are not bearish on US treasuries, even though the 10 year treasury yield is through 5.2%

So why are the macro “analysts” not that bearish. Their argument, in simple terms, is that the US cannot afford the interest payments, and hence will be forced into “financial repression”. In other words, government will control or force yields lower, and hence you should not be bearish bonds - but bullish “repression trades” like gold.

I don’t deny that financial repression is a very real possibility. But I think this “macro” analysis misses the real purpose of bond markets. And what is this purpose I hear you ask? I am almost certain that sovereign bond markets exist to discipline left wing politicians. This has been clear in emerging markets for decades. When Lula was first elected President in 2002, the Brazilian Real tanked, and bond yields soared until he committed to even tighter fiscal targets. 1 year swap rates went from 30% to 60%

The Brazilian Real lost 50% of its value.

What has been the political transformation that has led to bond markets being weak? Well, the Washington Consensus, of free markets and balance budgets began to fray back in 2000. If you think we had Ronald Reagan, George HW Bush, Bill Clinton were all relatively serious politicians. But at a certain point, the Republicans could no longer win elections by appealing to business minded Americans, and started to reach out to “other communities”. George W Bush, was noticeably different in style and substance to previous Republican presidents. Both John McCain and Mitt Romney were throwbacks to older Republican politicians, and both lost to Obama. You should also note that John McCain nominated Sarah Palin as Vice-President - again reaching out to “other communities”. It took President Trump to break the mould. While the traditional pitch for Republican was to cut taxes and entitlements, Trump has ditched the unpopular part of that offering, and now cuts taxes and increases entitlements. This leaves bond markets in a tricky place. They know the policy is inflationary, but they also know that a debt crisis that forces out Trump would mean a likely replacement by a more left wing politician? What to do? I think the obvious answer, and one that is part of the “starve the beast” thinking of the Republicans, is to wait until left wing politicians are in charge, AND then have a debt crisis. Here is the Polymarket odds of the Democrats winning the Texas Senate seat (Texas!). Remind you of anything?

Around the world, the traditional right wing parties are under siege, and are either transforming themselves, or at threat of being wiped out. And hence bond markets are behaving as they should. Yields reflect the sensibilities of policy making or the ability of governments to resist populist movements. In that interpretation, only Switzerland and China can seemingly resist. Politically, this makes sense to me.

As a final note, I had a chat with George Noble about being on his podcast in a few weeks. And he mentioned something that I had forgotten. I said to George, I like to always have two sides of a trade, just to hedge against political changes. But in my experience, investors always look at my ideas, and stick with the ones that are working. GLD/TLT is a perfect example of this. GLD/TLT has been volatile this year, but it is up for the year.

But in 2024 and 2025, it was the GLD part of the trade that worked. This year, gold is down year to date. I suspect many people just stuck with the GLD side of this trade - certainly that’s what I see in the “macro” community on Substack.

This is why I downgraded “macro” analysis in my approach - it is too black and white - and does not allow for politics. For me, there is a yield where US treasuries becomes attractive again. That yield is when it forces all sides of politics to be sensible. We are not at that yield yet. GLD/TLT still looks good to me.

Subscribe now