The new Fed chair has told markets plainly today that he intends to break inflation, that the Fed missed the job for five years, and will now fix it. The easy way is to believe artificial intelligence will do the job. He may be right about where this ends. What leaked this week just shows how unsustainable the current price of AI actually is.

OpenAI’s revenue is growing about as fast as it can. It is nowhere near covering its costs. Research and development alone – the cost of training the next model – exceeds what the company earns. Nine hundred million people use ChatGPT each week; only a fraction pay, and serving the rest costs money. The product is remarkable. It is also sold below cost for now. That is the real source of the disinflation everyone is celebrating without really seeing it yet, the “structurally disinflationary” miracle the Fed is counting on: a transfer from shareholders to users, funded by losses.

The largest loss-maker of the cycle has chosen this moment to go public.

Activity on the supply side – above all, large IPOs priced at full value – is a more reliable signal of a top than anything on the demand side. Insiders sell to the public when appetite is high, not when it is low (‘feed the ducks while they’re quacking’, as they say). And the signal asks nothing of my view on valuation: I do not need OpenAI’s mark to be wrong. When the marginal loss-maker moves its funding burden off a closed circle of believers and onto the public market, that decision is information, whatever the equity turns out to be worth.

And it arrives in anything but calm. There is a war, a truce that may not hold, and a Strait of Hormuz – through which roughly a fifth of the world’s oil passes – that sits one headline from closing again (assuming it’s currently open).

It is all one position: long the miracle, short the cost of capital, calm in the middle of a war, and now inviting the public to write the cheque. Retail investors are buying calm, and someone is selling it.

Click here.

In early February 2018, an exchange-traded note called XIV was worth about $1.9 billion. Within a single trading session, it lost 96 percent of its value and was wound down. The name was VIX spelled backward, in case anyone missed the joke: for most of a decade, it had done one simple thing, which was to sell calm – to pay you for assuming tomorrow would look like today. For two years, it returned 585 percent, and the danger printed on its label kept being wrong in exactly the profitable direction. The people who owned it were not all fools; many knew precisely what it did. That is the part I keep returning to. The dangerous thing was not ignorance – it was success. Being right is how a position grows, how a habit becomes a strategy, and, for a seller of calm, how the trade eventually learns to kill you.

Low volatility is not the absence of risk. It is a risk that has been moved somewhere, so it no longer has to be counted.

XIV was only the cleanest version of a much larger trade that keeps changing costumes. Tim Lee, Jamie Lee, and Kevin Coldiron gave it a useful name in their 2019 book, The Rise of Carry: carry is the monetization of stability – you receive a steady income for taking the other side of someone else’s wish for safety. The trader selling a put, the bank lending cheaply, the company borrowing to buy back its own stock – different costumes, one family resemblance. Each is paid to assume disruption will not arrive soon; each is, broadly, short volatility. The mechanism is seductive because it works. Capital crowds in, prices rise, measured volatility falls, the models permit more leverage, and prices rise again. The calm is not merely the backdrop to the trade – it is partly produced by it. From the inside none of this feels reckless. It feels like discipline; the returns are real, and after a while, the trade stops looking like a trade at all. It becomes the way sensible people behave.

Carry is also a habit the whole system has learned. Central banks do not blow bubbles in the cartoon sense; they respond, for understandable reasons, to serious stress by cutting rates and putting a floor under prices. Do it once, and it is a rescue. Do it for thirty years, and it becomes a promise nobody had to write down. Investors learn that if disorder grows large enough, the authorities arrive – gains in the calm stay private, losses in the crisis are socialized. When carry unwound in March 2020, help came within days. When the yen carry trade began to snap in the summer of 2024, and the VIX briefly traded above 60, a single reassuring signal from Tokyo turned the market within days. The lesson lands every time: disorder will not be permitted to persist. And each time it lands, the next position is built a little larger, because the thing that made it truly dangerous – a fall allowed to keep falling – has been quietly removed from the model.

• • •

So where does this logic show most clearly today? I think it is the AI build-out, but I want to be careful with the claim. AI is not itself a carry trade. I use it every day; the demand is real, the revenue is real. By late May, Anthropic was reportedly running at roughly a $47 billion annual run rate, against about $9 billion at the end of 2025; OpenAI is reported to be near $24 billion. The growth is extraordinary, and people are paying gladly. That is the engine of the trade, and the engine is genuinely impressive. The part that worries me is the financing edge – where long-dated physical infrastructure is funded on the assumption that demand, capital, power, chips, and model economics all arrive on time. To produce those tokens, the industry is committing to spend on a scale with few peacetime peers: McKinsey puts the global data-center build near $7 trillion by 2030, with the largest American hyperscalers spending on the order of $700 billion this year alone.

The obvious retort is: so what? Microsoft, Google, Amazon, and Meta carry among the strongest balance sheets in corporate history. True, and important – which is why the fragility is not at the center. It lives at the edge.

The first edge is the labs. OpenAI’s compute commitments have been reported near $1 trillion of deals this year – its own chief executive has framed them at about $1.4 trillion over eight years, a figure since guided down toward roughly $600 billion of spend by 2030 – set against revenue still in the low tens of billions and a cash burn expected to become enormous. That may be the right bet. But it is a bet that the future arrives on schedule, and the future does not have to fail for the financing to hurt. It only has to arrive late.

The second edge is the structure. The early build came largely from retained earnings; that phase appears to be ending, and more spending now runs through debt, leases, special-purpose vehicles, and joint ventures – what the Bank for International Settlements has described as shadow borrowing, raised against chips and buildings while kept off the parent’s main balance sheet. The lenders on the other side are often private-credit funds and insurers: precisely the institutions that like steady coupons in a low-spread world, and that are asked to hold the risk that looks boring until it suddenly is not.

The third edge is circularity and time. Nvidia has agreed to invest up to $100 billion in OpenAI, which will spend vast sums on Nvidia chips; Nvidia backs CoreWeave, which buys Nvidia chips and rents capacity back to OpenAI. Money leaves one balance sheet and returns as another company’s revenue – in a real boom a virtuous circle, in a carry regime a system manufacturing proof for itself. And the GPUs at the center may be depreciated over five or six years, while the debt behind the buildings assumes a useful life of twenty or thirty. Each of these assumptions may be reasonable. Together they begin to look like a sale of volatility. It rhymes with 2007 – not as prophecy, but because the structure is familiar: clever vehicles, large obligations, off-balance-sheet comfort, and serious people persuading themselves that risk has been engineered into something safe.

• • •

The strongest objection is that I am using the wrong lens. On that reading the build-out is not carry at all but the early infrastructure of a general-purpose technology, and infrastructure always looks wasteful while it is being built. The railways bankrupted many of their first investors and then carried a century of growth; electrification and late-1990s fiber demanded capital far ahead of visible revenue. That objection is serious, and it may be right. But AI can be transformative and still destroy capital for whoever finances it first – the internet was real in 1999, the fiber was useful, and many of the companies were not. The question is not whether AI matters. It is who owns the stranded capital if it matters, but the margins, timing, or bargaining power land somewhere other than expected.

I also have to be honest about myself. By temperament, I am a buyer of protection, drawn to the side of the trade that bleeds quietly and waits for the day the dam breaks. That instinct has been right before, and being right is the most dangerous thing that can happen to it: a man vindicated once can spend a decade paying a premium for the encore. So I have to ask whether carry is the right analysis, or merely the lens a congenital bear reached for this week. I would mock that in someone else; I should not excuse it in myself.

Here is the test I have settled on. If this is infrastructure, the calm should survive a shock. If it is carry, the calm depends on the shock not lasting. So I am not watching the level of the Nasdaq, AI revenues, or capital spending in isolation. I am watching whether the calm still repairs itself.

That means I owe the reader a number, because without one, this is a mood dressed as analysis. The early warning is probably not a crash but a deal that does not clear: the first piece of AI paper – an asset-backed tranche, an SPV refinancing, a hyperscaler bond, a data-centre loan – that is pulled, repriced hard, or left to roll and cannot; or the first volatility spike that takes more than a week to be bought back, where the system that healed 2018 and 2024 in days suddenly needs longer. I would call the break confirmed if the VIX holds above 30 for ten trading sessions or more while investment-grade credit spreads widen alongside it. Volatility alone heals; volatility and credit together, refusing to mend, are a different thing. And here is the part the sellers of calm rarely publish – the level at which I am wrong: if the big AI financings clear, if OpenAI and Anthropic reach public markets without disturbing credit, if spreads sit near their lows through the end of the year and every volatility spike is bought back within a week, then the regime is intact, this was infrastructure rather than carry, and I should fold the position and say so plainly.

The sellers of XIV believed the calm was simply the world as it was. The people writing the cheques for data centers may believe something similar – with better arguments and far more at stake. They may also be right. I am not standing above them; I have made their mistake in the opposite direction more than once. The only useful discipline I know is to write down, in advance, the price at which I would admit it.

Calm is not free. We have just become very good at forgetting where the bill is sitting.

PS: With thanks to my big friend Ahmed Husain – whose newsletter, , is one of my favorite weekly reads. He always recommended great books to me, and The Rise of Carry was no exception.

Earlier this week, I published The Supply Side of a Top: the hypothesis that this market’s top, if and when it comes, might be made less by an exhaustion of demand than by a wave of equity supply arriving into a system losing its capacity to absorb it. It is just a hypothesis – markets have digested mega-listings before without marking anything. Which is exactly why Friday matters: it is the first clean, real-time test.

What the deal anatomy adds

The valuation dispute is not about discount rates. At ~94x 2025 revenue, Starlink is the only profitable segment, and the S-1 itself projects the AI division to burn ~$10 billion in 2026. The gap between Morningstar’s $780 billion and the $1.75 trillion ask is therefore a disagreement about whether the AI segment is an asset or a liability: Goldman, the lead underwriter, models explosive AI revenue growth in its roadshow materials; Morningstar’s analyst writes that the acquisition “poses a material threat of value destruction.” That is not fine-tuning – it’s almost as if two different businesses are being priced.

The xAI merger is the governance event the prospectus cannot dress up. It was a related-party transaction in which the same person negotiated both sides – a deal that Morningstar explicitly flags as not conducted at arm’s length – and was folded into an entity where Musk retains roughly 82% of the votes. AkademikerPension went further than abstaining: a full blocklist, including passive products. A fund that blocks a top-ten index constituent is paying for its view in guaranteed tracking error.

The retail allocation is a structural tell. Up to 30% is reserved for retail – triple the norm – including a dedicated European tranche across seven countries. Institutions negotiate on valuation; retail subscribes to narrative. When $75 billion of paper needs to be placed at 94x sales, you go where the bid does not ask questions.

One fairness note: this is not an insider distribution. The offering is 100% primary – no existing shareholder sells a share at the IPO. The bearish reading is therefore not “smart money exiting.” It is $75 billion of entirely new paper in a single session, into a float of only 3–4% of the company.

The passive bid is real, and it cuts both ways. No S&P 500 entry for at least a year – the index provider declined to bend its rules – but Nasdaq-100 fast-track inclusion roughly fifteen trading days after listing and confirmed early MSCI inclusion. Forced buying into a float this small is genuine technical support – and a textbook setup for a strong debut that tells us nothing about value. The thesis is not tested in week one; it is tested when the mechanical bid fades and the organic market has to absorb the $75 billion of new paper – and price a $1.75 trillion company – on its own.

The markers I am setting now

What proves me wrong? Between now and the end of Q3:

1. Absorption, not direction. With index buying queued, a strong debut is the base case and proves nothing. What matters is what the rest of the tape does while $75 billion is digested: breadth, equal-weight versus cap-weight performance, and funding conditions.

2. The cohort effect. SpaceX is the first wave. OpenAI and Anthropic are racing toward listings, and each successive deal arrives into a market that has already absorbed the previous one.

What would prove me wrong

The thesis fails if the market absorbs SpaceX, then OpenAI, then Anthropic, without deterioration in breadth, credit spreads or liquidity conditions – if participation broadens rather than narrows and the later deals price at or above target. That would mean absorption capacity is far larger than I estimate: that money-market cash and foreign demand can simultaneously can swallow historic equity and Treasury inflows. Possible.

Friday’s open will be loud and largely meaningless. The signal sits in the digestion, not the debut.

Not investment advice. I hold no position in SPCX and do not intend to participate in the offering.

On 4 June 2026, SpaceX began its roadshow: about 4% of the company – 555,555,555 shares at $135 – for roughly $75 billion, so an implied valuation of $1.75 trillion. The largest IPO ever, more than double Saudi Aramco’s 2019 listing, the previous record. It lists on the Nasdaq this Friday, 12 June, under the ticker SPCX.

More is coming. Anthropic filed confidentially on 1 June at $965 billion, days after raising $65 billion in private capital. OpenAI is apparently readying its own filing, valued at $852 billion, eyeing an IPO above $1 trillion. Three companies, around $3.5 trillion of combined value, lining up in a matter of weeks. Last year, OpenAI burned through at least $9 billion and Anthropic around $5.6 billion.

SpaceX looks different – a profitable launch business, a cash-generative Starlink. Not anymore. In February, it merged with Musk’s xAI, and xAI’s $6.36 billion operating loss dragged the combined company to a $4.94 billion net loss for 2025, on $18.67 billion of revenue. The latest quarter alone: $4.28 billion. The rocket company now carries a cash incinerator inside it.

The 96% not being sold stay with insiders, behind a lock-up the bankers have fitted with release valves – the right to sell in stages within weeks, not the customary six months. This is not generosity. It builds a tradeable float fast enough to pull SpaceX into the Nasdaq 100 as soon as 15 trading days after listing – at which point every fund tracking the index must buy, at whatever price the float has reached. The insiders sell. All the investors holding the index through ETFs, mutual funds, etc., buy, as the index is not an abstraction: it is the pension fund that must hold what the benchmark holds, and the saver who never chose this trade will own it all the same.

Morningstar ran the cash flows on a business trading at more than 90 times revenue and arrived at $780 billion, 55% below what the bankers are asking. They did not soften it: significantly overvalued. Of that $780 billion, launch and Starlink account for some $611 billion; the entire AI operation – the xAI that turned SpaceX loss-making – they value at roughly $170 billion, flagging its flagship orbital-data-center ambition as scientifically and economically unproven. They named the structure a related-party merger not struck at arm’s length. Musk left with about 85% of the vote through a dual-class share. And in the same note, they explain why the stock will likely rise anyway – a thin float, almost every bank on the planet underwriting it, and that forced path into the index.

Supply like this does not arrive at the bottom.

Between them, the three are seeking something like $200 billion from the public – $75 billion of it firm, at SpaceX; the rest is an estimate, since OpenAI and Anthropic haven’t set their sizes. The entire US IPO market raised $45 billion in all of 2025. Goldman Sachs expects roughly $160 billion in proceeds this year – four times last year’s figure.

David Solomon, who runs the bank underwriting much of it, stood at the Economic Club of New York and did not reach for caution. There is, he said, “more greed than fear” – plenty of liquidity, an exuberance he expects to run a while yet. He meant it as reassurance: the market can take the supply, the way it had just swallowed Alphabet’s $80 billion raise without strain.

He is probably right that it can. That is not the issue. The issue is that the man whose bank is manufacturing the supply is also the one calling the market greedy – and choosing this moment to bring trillions of it to market. He is not forecasting a fall. He is placing an offering. And an offering this size can only be placed when buyers will take almost anything.

These companies are raising capital because they burn cash by the billion and must do. The appetite is simply what lets them do it at these prices – capital-hungry sellers meeting a market that will fund anything.

Whether this is a top, no one can say yet. But it is what the supply side of one looks like: the need to raise has met the willingness to fund, and both are near their maximum.

And equity is not the only issuance into this appetite. The largest seller this year is the government.

In the second half of 2026, the US Treasury must raise around $1.3 trillion in new money – not the far larger sum it rolls over as old debt matures, which is mechanical, but fresh borrowing, sold on top, to fund a deficit running near $2 trillion a year. Add the cash the three IPOs need – the float they sell, a fraction of those headline valuations – and the genuinely new supply landing in a single half-year is on the order of $1.5 trillion. Close to 5% of GDP, roughly 2.5% of the entire S&P 500. All of it paper someone will buy (a lot of us included, directly or indirectly…).

There is, as Solomon said, the liquidity– some $8 trillion sits in money-market funds, several times the year’s new supply. And yet the long end of the curve is already flashing. The 30-year yield has pushed back above 5% and does not trade that way when buyers are hungry for paper, let alone with the heaviest issuance still to come.

This is the second sense in which it is the supply side of a possible top. Equities are the first stream; sovereign debt the second; the insider stock the lockups will release, the third – all competing for the same finite pool of savings. Solomon is right that the liquidity is there. Perhaps it absorbs everything the way it has for most of a decade, and a year from now, this reads as yet another time the skeptics were early.1

Here is what probably should bother us anyway. The justification for owning the most expensive assets in the market keeps changing

We were told these were worth a fortune because money was free. At a zero-discount rate, a dollar of profit two decades out is worth almost a dollar today – and technology, whose earnings sit furthest in the future, gained most from the arithmetic. It was coherent. But Fed funds now sit at 3.5–3.75%, 10-year yields are at 4.55%, and the arithmetic has gone into reverse. The prices have not. Where they once rested on free money, they now rest on the AI revolution we’re in, et al.

This is motivated reasoning at the scale of a market: the conclusion comes first, the reason is framed to fit, then quietly swapped when it stops working. US equities are worth about 230% of GDP – the gauge Buffett once called the best single measure of where valuations stand. At the dotcom peak, it reached about 145%. We are far above that now, at the highest reading on record. When everything is expensive, price stops being information or, at least, such a relevant signal. What’s left to go on is the story – whichever one is currently in the media.

On Friday, the tape cracked. The Nasdaq 100 fell 4.77% in a single session – its worst day in more than a year. The trigger was a jobs report strong enough to make the market fear the Fed would probably need to hike rates, but it landed on a tape already cracking. The sell-off was led by semiconductors – the chip index was down some 10% on the week, after Broadcom declined to raise its AI guidance and a skeptical AI note circulated. The rate shock didn’t start the fire. It poured accelerant on AI names whose conviction was already thinning. The index closed at 28,958, below its 21-day average for the first time since early April.

The warning was already in the indicators. Through May, as the index ground to new highs, momentum refused to follow – RSI and MACD both made lower highs while price made higher ones. The textbook negative divergence: a market rising on thinning conviction. Friday turned the divergence into a break, with MACD crossing below its signal. And midterm technical indicators – relative strength sits at 48, not even close to oversold (below 30); the stochastic has only just rolled over from the top of its range. The indicators have a long way to fall before anything says the selling is done.

None of which, on its own, is a top. A 4.77% day after a 25% run in two months is a pullback until proven otherwise. The 50-day average sits barely 3% below Friday’s close; the 200-day is still about 12% lower, and a bull market reverting that far toward a rising average is doing something it does quite often

The honest case against all this is not weak – it has been winning for years. Start with the liquidity: an ocean of it, $8 trillion in money-market funds alone. Friday’s fall came from a sudden fear of higher rates, and I doubt that fear is warranted. An oil shock and a Middle Eastern war are, on a lag, disinflationary – the kind of supply shock a central bank looks through rather than leans against and can do little about in any case. I don’t think the Fed hikes. I think it holds and probably begins cutting before the year is out. If that’s right, the rate scare might have been noise, and the cash on the sidelines does what it has done for fifteen years – buy the thing. Betting otherwise has been the great wealth-destroyer of the era.

The liquidity case rests on an assumption that deserves daylight: that these businesses grow into their valuations. In the first quarter of 2026, OpenAI lost $1.22 for every dollar of revenue – an adjusted operating margin of minus 122%, even after stripping out stock compensation. At that level, growth is not the cure; it is the cost. Every additional dollar of sales comes with more than a dollar of loss. The bull asks you to fund the growth and trust that scale inverts the economics – on schedule, in a market where half a dozen well-capitalized rivals are driving the price of intelligence towards zero. Maybe it does. What you’re funding, for now, is a forecast.

The largest IPO in history cannot tell you the top date. It can tell you that the sellers already have theirs.

1. A story told against himself

Some years ago, I sat in a conference room in New York and listened to Kevin Warsh.

He was the most impressive man in the room, and it was not close. Quick, disarming, fluent in the particular way that makes you put your pen down and just listen. Near the end he told a story against himself, the way clever men do when they have decided they would rather be liked than feared. He had been in a car, he said, coming back from some conference, Ben Bernanke on one side and Janet Yellen on the other. It was the early life of quantitative easing – QE1, perhaps QE2 – the years the Federal Reserve was buying bonds with money it had willed into existence, and the whole profession was still arguing about whether the thing would work. Bernanke turned to him and asked, more or less: you know what we’ll do if this doesn’t work, don’t you? We’ll do more. And then, before Warsh could answer: and you know what we’ll do if it does work? We do more.

The room laughed. I laughed.

I have carried that story for the better part of eight years, and somewhere along the way I stopped finding it funny.

2. The hawk in the chair

He was confirmed in mid-May, by the narrowest vote in the Fed’s history, and took the chair days later. He had always been a hawk – that was never in question. The surprise was that Trump chose him knowing it, an avowed hard-money man installed by a president who wanted softer money, in the apparent hope that the office would turn the hawk into a dove. The market is not so sure it will. Before the war, the curve was priced for cuts; now it has at least one hike in it, in some places two, and the man who wanted lower rates has, for the moment, the opposite.

And inflation has run above the Fed’s 2% target for more than five years now and is rising again on energy.

What makes the line worth keeping is not that a central banker said it – they say many things – but where he said it: in a car, among peers, off the record, in the one setting where a man in that job stops performing and tells you what he actually believes. Notice the structure. The outcome is irrelevant. Works, fails – the response is identical. That is not indecision. It is the opposite: a policy with only one available direction, admitted in private precisely because it could never be admitted in public.

Monetary policy, at the scale of a sovereign that has borrowed more than it can ever repay in honest money, does not flow both ways. It is a river. It runs in one direction, toward the sea, and no act of will sends it back uphill to the spring.

A hawk can dam the river for a season. One did more than a season: Volcker turned the water back, and it stayed back for the better part of a generation. Volcker could win because the debt was small – about thirty cents on the dollar of income. Double-digit rates hurt, but the country could carry them. Run the same rates over a debt larger than a full year’s income and you don’t break inflation; you break the borrower: the interest bill alone starts eating the budget. That is why this is not 1981 in a new suit. Warsh can hold rates above inflation and shrink the balance sheet, and for a while it holds – the people positioned for his hawkishness will look right. But holding the line and being able to afford it are two different things.

A dam is not a reversal.

the water is still only going one way, and it is rising behind the wall. The question was never whether a disciplined chair can hold for a quarter or a year. It is what happens the day holding the line collides with a debt that cannot be serviced at the rates the line requires. We know what happens. A man told me, in a car, a long time ago.

A river runs one way.

3. Why are the markets at record highs?

Which brings me to the question I am asked more than any other right now. How can the headlines be this bad and the market be at a record high? This week alone: a war throttling the artery through which 20% of the world’s oil moves; real wages compressing at once across the United States, Britain and the euro zone; a cost of living that bites a little harder each month. And against all of it, the S&P 500 closing the week around 7,580 and the Nasdaq-100 north of 30,000 – at records, or close enough to touch them – as though reading an entirely different newspaper.

Bull markets climb a wall of worry, as the saying goes. It is true, and it explains the smaller version of the puzzle. It does not explain this one.

The instinct is to call one of them a lie. Either the market is delusional, sleepwalking toward a fall – or the misery is overstated and things are secretly fine.

The honest answer is that the market is not lying – but it is not telling the whole truth either. Part of the rise is real: earnings are growing, and if the machines deliver what their believers promise, the market has earned some of what it shows. But part of it is the ruler, not the thing measured – the index is being counted in a currency that is quietly being made smaller, and a number that only ever goes up can hide a value sliding underneath it. Which raises the question no one at a record close wants to ask: why would a government choose to shrink the measure?

4. Three doors out of the debt

Because the alternatives are worse for nearly everyone in the short run – and worst of all for the people who have to stand for election. A country buried in debt has exactly three exits, and only one is survivable for someone who faces voters. It can default – but default crystallises the loss now, on named victims, conveniently before the next vote. It can grow its way out through real productivity – but demographics and physics are not cooperating, and growth of that order takes a generation nobody has. Or it can inflate: hold interest rates below the rate of prices, year after year, and let the real weight of the debt evaporate while the nominal numbers all stay reassuringly intact.

Economists have a polite name for the third door. They call it financial repression – the postwar machinery, documented by Carmen Reinhart among others, that liquidated the vast debts of the 1940s not by paying them down but by pinning rates beneath inflation for three decades until the burden simply melted. It worked because almost no one noticed. As Keynes put it a century ago, a continuing inflation lets a government confiscate, secretly and unobserved, a large part of the wealth of its citizens – by a process not one man in a million is able to diagnose.

The line everyone quotes is the one about no man in a million seeing it. But the people reading this can see it perfectly well. That was never the trick. The trick is that seeing it changes nothing – there is no ballot against it, no morning it arrives, no name on the loss. You can read the tide exactly and still get wet.

5. The painless version

Here is the version I half-believe on my better days. The Fed holds rates a little below inflation and keeps them there. Nominal growth runs just ahead of the interest bill, so the debt shrinks as a share of the economy with no vote, no tax, no austerity. Wages rise two or three percent above prices – enough to feel like a raise, not enough to march over. The debt mountains of France, the US and Japan erode grain by grain, with no single day you could point to. No crisis, no reckoning: just money worth a little less each year, and a world that grows out of its debts in name while never repaying them in fact.

What that means is simple, and it depends entirely on what you own. Hold cash or government bonds and you are the one paying – taxed at the rate of inflation, with no line item and no name on the bill. Hold what cannot be printed – equities, property, gold, a business – and you are roughly kept whole, because the price rises in the shrinking unit. That is most of the answer to the record high: the index is not calling the economy strong, it is being repriced in a smaller currency.

There is a kinder version, and I won’t pretend it away. If AI delivers half of what its believers promise, the second door – real growth, the one I called shut – opens again, and the debt is genuinely outgrown instead of inflated away. From a record close the two look identical; both print the same number. But I wouldn’t bet the house on the kind door. Growth needs time, which no government facing voters ever has; inflation can be had this year, on no one’s signature. And the machines may cut the other way – if they are as deflationary as they are powerful, a falling price level over a debt this size is not relief, it is the trap closing, and it forces the press to run harder still. Only one of these is real, and only one leaves a victim. In the good version the victim is scarcity; in the bad version it is the saver. You will know which door you walked through by who paid for it.

And it might happen, the painless way. But notice who is missing from the photograph.

The retiree holding cash and high-grade bonds, whose “safe” assets are taxed at the inflation rate without a single line item ever appearing on a statement. The wage-earner whose raise trails prices by 18 months, every year, forever – a tax paid in arrears that never quite catches up. The saver who did the responsible thing. The foreign creditor who lent real goods and is repaid in softer money. The debasement is not free. It is merely unseen, because the bill is presented in the denominator, and no one ever reads the denominator.

5. France has the debt and not the key

Not everyone gets to choose this exit. To inflate a debt away you have to own the press – and France doesn’t; the euro is run from Frankfurt, not Paris. So here is a country with every reason to inflate and no key to the machine: public debt past 115% of output – over 320% once private debt is added, more than three times national income – with roughly half the government’s share held abroad. France is not bankrupt, and it is not Argentina; it is rich, productive, administratively capable – and politically trapped. That combination is more interesting than a simple crisis.

The irony is sharp: the first modern hyperinflation was French, the assignats of the 1790s, paper printed against seized land until it was worth nothing. France gave up the press because it remembered where the press leads. Now it has the debt and not the key. Its only move is to make Frankfurt run the press for it – and a shared central bank cannot debase for one member alone. A euro cheapened for France is cheapened for everyone who holds euros, the frugal alongside the profligate.

France has the debt and not the key.

Which is where Portugal comes in, where I am writing this. Lectured all through the 2010s, it ran surpluses and brought public debt from around 135% at the pandemic peak down toward 90% today – partly discipline, partly the same tide, the debt shrinking against an output inflation had swollen. It didn’t escape the river so much as ride it – but it had far less that would ever need printing away. And here the two stories collide: if France forces Frankfurt’s hand, the euro debased to lighten French debt is the same euro Portugal saved in. Whether prudence is rewarded or quietly expropriated to rescue the profligate, I don’t know. The euro was built without an answer to that. We may be about to find out if it has one.

The slow is where you live.

6. The slow is where you live

If the most consequential financial fact of your life is the slow shrinking of the unit you count in… Nassim Taleb’s argument about the news is not that it lies. It is that it over-samples. The more often you check, the worse your ratio of signal to noise, because the things that move on a daily timescale – the strike, the summit, the print, the panic – are almost all noise, while the things that matter move too slowly ever to clear the bar for news. A debasement is the purest case: no single day’s move is ever large enough to report, so it never leads a bulletin, and yet it is the most important number in your life. And the noise it hides behind is not neutral. The news is negative by construction, because fear travels and calm does not. So, a whole population can come to believe the world is ending, reading it in a thousand alarming headlines, while the market quietly counts something they are not – and the gap between the dread and the record close turns out not to be a paradox at all. It is a diet. And the slow is where you live.

I think often now about that room in New York, and about how completely Warsh had me. He was brilliant, and I wanted him to be right, wanted the competence to be real, wanted the story to be the self-deprecating joke he served it as. It was not a joke. It was a forecast – and the man who delivered it, smiling, is now the one being paid to make sure it never comes true.

They will do more. Perhaps not this year, perhaps not until something none of us can yet name forces the question. Warsh may hold the line longer than anyone expects, and the discipline may look, for a season, like a choice. But the river he described in that car is larger than any single man’s resolve – his included.

The headlines will keep screaming that the world is ending. The market will keep printing records. Both are reading the same country, in two different units – and the work, the only work that matters now, is learning to read the one nobody prints.

This is the personal editorial view of Nuno Amado Mendes, CFA. It is opinion, not investment advice, and not a research report.It is opinion, not investment advice. It reflects a view that may be wrong and may change without notice. Images generated by AI.

For twenty-three years, he refused. Then, on the twentieth of May, the largest IPO in the history of the public markets landed on the SEC’s desk.

Elon Musk had spent the better part of two decades explaining, on the record, why he would never take SpaceX public. Public markets force short-termism, he said. They reward the quarter. They don’t understand businesses whose horizons measure in decades. He repeated this through every funding round – 2013, 2015, 2017, 2020. SpaceX raised what it needed through a chain of private secondaries at ever-rising valuations, climbing from tens of billions to hundreds of billions, then past $400 billion. There was no shortage of capital. There was a stated principle. Then he filed.

The structure of the filing answers part of the question. A dual-class share architecture designed to preserve Musk's control over every meaningful corporate decision, regardless of how much equity is sold to outsiders. Where Tesla relied on its founder stake and its hold over the board to keep it in charge through every dilutive round – control that was contingent enough to be contested in Delaware – SpaceX is hardwiring the control into the share class itself. He keeps the control. The IPO happens anyway.

So what changed isn’t the company. It’s the market.

Musk is not the only one whose behaviour has shifted. Warren Buffett is sitting on the largest cash position in the history of Berkshire Hathaway. Jeff Bezos is selling Amazon stock at a pace without precedent in the company’s public life.

When the best-informed actors all move at once, in the same direction, after years of moving differently, you notice.

II. The Filing Itself

The filing makes SpaceX the largest IPO in human history. The valuation range – one point seven five to two trillion dollars, with a rise of approximately seventy-five billion – exceeds Saudi Aramco’s December 2019 listing on both axes. Aramco raised 25 billion at a $1.7 trillion valuation. The new filing nearly triples the raise. Aramco was, until last week, the reference point. It is no longer.

For Musk personally, the timing aligns with a peak. He’s filing at peak personal wealth, into peak market liquidity. His stake in SpaceX alone is worth several hundred billion dollars at the filing range. Combined with Tesla, xAI, and the privately held Boring Company and Neuralink, his net worth has never been higher.

The mechanics deserve attention because they explain how he will get paid.

In a private secondary, Musk could sell a few billion dollars per round to a finite pool of late-stage investors. These rounds were priced at progressively higher levels – that is how SpaceX climbed from tens of billions to four hundred – but each round was constrained by buyers’ appetite at that valuation and by the dollar amount the pool could absorb without dampening sentiment for the next round.

A public listing changes the absorption math entirely. Once trading begins, the world’s pool of retail and institutional capital – mutual funds, pension funds, index inclusion, retail flows – becomes available simultaneously. For a controlling shareholder, this is the difference between selling into a finite room and selling into an open market. The dual-class structure ensures that the seller remains the controlling shareholder anyway.

There’s a pattern in mega-IPOs that’s hard to wave away.

Saudi Aramco was listed in December 2019. The oil market peaked early the following year before COVID detonated the structural collapse. Aramco itself, while sovereign-backed, has spent the years since its IPO debut trading well below its IPO price.

Glencore was listed in May 2011. The commodity supercycle, which had been the backdrop to the largest mining and trading conglomerate going public, peaked within weeks. The next decade was unkind to commodities.

Blackstone was listed in June 2007. The world's largest private equity firm chose to monetize its general partnership stakes in public markets weeks before the credit cycle began to crack. Blackstone, the firm, did exceptionally well in the years that followed. Blackstone the stock, for shareholders who bought at IPO, took more than seven years to recover its issue price.

The pattern is not infallible. Plenty of mid-sized IPOs occur in mid-cycle conditions and produce normal returns. But the largest filings of any cycle – the ones that absorb a meaningful fraction of all equity-issuance demand for that month or quarter – tend to land at the moments when the absorption capacity of the system is at its outer bound. That is when the controlling shareholder rationally chooses to issue.

The filing is not, on inspection, a capital-raising story. SpaceX has never struggled to raise capital. Its private secondaries have been oversubscribed at every level for a decade. The argument that this IPO funds the next phase of investment – Starlink expansion, Starship cadence, deeper AI infrastructure – survives only until you notice that the company has had access to all the capital it could deploy for years, and chose, until last week, not to broaden it.

What changes with the IPO is not the company’s ability to raise capital. It is the controlling shareholder’s ability to monetize its position. The IPO is the cash-out vehicle. The dual-class structure is how it is done without giving up the company.

III. The other “suspects”

Musk’s filing is the most visible of a set of behavioural shifts that have been accumulating over several quarters. Each of the actors involved manages capital at a scale where their decisions are the news.

Warren Buffett. Berkshire Hathaway closed the first quarter of 2026 with three hundred and ninety-seven billion dollars of cash and treasury bill equivalents. That is the largest absolute cash position in the company’s history. It is also, as a fraction of total assets, the most defensive position Buffett has run since Berkshire reached its current scale. Three years ago, the same cash pile sat at roughly 120 billion. The progression – not the absolute number, but the trajectory – is what matters: from one twenty to two hundred to three hundred to three ninety-seven, in a span when the S&P 500 nearly doubled. He has been a net seller of equities in nineteen of the last twenty quarters. He has not, in the same period, made a major acquisition. The position is not a forecast. It is a statement about the absence of a fairly priced opportunity.

Jeff Bezos. Filings show Bezos sold approximately $13 billion in Amazon stock in the 18 months ending in April. That is the largest aggregate insider sale by the founder in the company’s public life. The sales were not under 10b5-1 plans set during lower-price periods; several were discretionary and executed at progressively higher prices. The cash has been redirected to the Earth Fund, Blue Origin, and asset classes outside public equities. He continues to hold a substantial stake in Amazon, and his selling is not a vote against the company. It is a vote about what to do with capital extracted from Amazon at current levels.

Mark Zuckerberg. Zuckerberg sold over a billion dollars of Meta stock in the first quarter of 2026 alone, the largest quarterly insider sale in the company’s history. The selling has accelerated through 2025 and into 2026, tracking the share-price ascent rather than fading against it.

Aggregate insiders. The sell-to-buy ratio across S&P 500 insider transactions has been at multi-quarter highs since the start of the year, last seen at comparable levels during the early stages of the 2021 retail euphoria. The data point itself is unreliable in isolation. Insiders sell for personal liquidity reasons unrelated to their view of the stock. But when the rate of selling rises against rising prices, rather than falling, the noise begins to cancel out, and a signal emerges.

These are not coordinated actors. What they have in common is information asymmetry – about their company, their balance sheet, their portfolio – and a willingness to act on it.

The signal is not their collective rightness. Each of them may turn out to be wrong. Bezos has been a net seller of Amazon stock for fifteen years. The signal is that they are acting at the same time, in the same direction, after years of acting differently. That is what tells you something about the distribution of risk that experienced operators are seeing in late May 2026.

IV. But Maybe They’re Wrong

The reading I have given so far has an obvious counter, and it is worth taking seriously.

Founder selling has explanations that have nothing to do with markets. Bezos has been a net seller of Amazon for fifteen years, financing space ventures, philanthropy, and a divorce. Zuckerberg’s selling is partly to fund the Chan Zuckerberg Initiative, partly tax planning, and partly simple diversification away from a single-name concentration that no portfolio manager would tolerate in a client account. Musk’s filing, similarly, may be more about creating liquidity for a position that has grown beyond the absorption capacity of the secondary market than about a market timing call. Each of these explanations is partially true. Most likely, they are all true at the same time.

Mega-IPOs do not always mark tops. Facebook went public in May 2012 at what was then the largest tech IPO ever attempted. The stock collapsed in its first months and then rose roughly sevenfold over the decade that followed. In September 2014, Alibaba raised twenty-five billion dollars at a one hundred sixty-eight billion valuation. The Chinese equity bubble peaked the following year, but US markets continued higher for years before any significant drawdown. The pattern I cited – Aramco, Glencore, Blackstone – is real. The counter-pattern is also real.

The bull case is not crazy. Net profit margins at the largest US companies are at all-time highs. Free cash flow generation in the dominant AI franchises is unlike anything in 1999. Capex is being absorbed by firms that can actually afford it. Microsoft, Google, Meta, and Amazon together booked roughly two trillion dollars of revenue over the last twelve months. The dispersion between earnings today and earnings in 1999 is the strongest single argument against pattern-matching too tightly.

Smart money has been wrong before. The same names that sounded cautious now sounded cautious in 2022, 2023, and much of 2024, while the index rose more than 50%. Buffett ran cash through 2017, missing three years of returns. Druckenmiller was publicly bearish through portions of 2023. Conviction held by experienced operators is not the same as accuracy.

The case for paying attention does not rest on their being right. It rests on the joint probability that they are all simultaneously wrong, after years of acting differently, at a moment when the bull case requires perfection across multiple structural variables to sustain.

That is the difference between a forecast and a posture. The forecast is what they would say if asked. The posture is how they are positioned. The posture is the more reliable instrument because it costs them something to be wrong. Skin in the game, in this case.

V. The Conditions Around Them

The behavioural signals do not exist in a vacuum. The market in which Buffett raised cash, Bezos and Zuckerberg sold, and Musk filed has its own measurements – and those measurements are, independently, at extremes.

I will be brief here. The conditions are the context, not the argument.

Concentration. The top ten stocks of the S&P 500 now account for roughly forty percent of the index. During the dot-com peak, that figure reached twenty-seven percent. Between 1990 and 2015, the range was eighteen to twenty-three percent – a band so stable across two business cycles, a tech boom, and a financial crisis that it functioned as a structural feature of American capitalism. That structural feature has broken. The S&P 500 is no longer five hundred companies. It is, mechanically, an index whose movements are determined by ten names.

Top 10 S&P 500 concentration, 1990–2026

Technical extension. The Nasdaq 100 closed last Thursday 13% above its 50-day moving average. The last time it sat this far above trend was the week of the dot-com peak in March 2000. The semiconductor sector – the operational core of the AI thesis – has gone vertical: roughly 75% in 8 months, with a slope visible in only two prior episodes over the past three decades. These are not predictions of anything. They are measurements of how far recent price action has departed from any reasonable medium-term reference.

SMH (VanEck Semiconductor ETF) — daily, last eight months

Breadth. While the index hits new highs, the share of constituents trading above their two-hundred-day moving average has been falling. In mid-2024, the figure was near seventy-five percent. It is now closer to forty-eight. Healthy bull markets have most stocks rising in concert. Late-cycle markets have a smaller group of names driving the cap-weighted index higher, while underlying participation deteriorates. The same pattern was visible in 1999, in 2007, and in 2021.

S&P 500 vs % constituents above 200-DMA

Sentiment and positioning. Call option volume now represents roughly fifty-eight percent of all S&P 500 option activity, a level historically associated with speculative blow-offs rather than mid-cycle conditions. The Shiller cyclically adjusted P/E sits near 40, a reading that has been reached only twice in the index’s history – in 1999 and in late 2021. Both of those readings preceded significant drawdowns.

None of these readings, in isolation, predicts the next twelve months. Markets remained “extended” in 1998 and 1999 for a long time before the eventual reversal, and an investor positioning purely off any single one of these indicators in 1998 would have spent the next eighteen months explaining their underperformance. The conditions are necessary, not sufficient.

What they describe is the texture of the regime in which the actors in the previous section choose how to deploy capital. Concentration tells them diversification is mechanically illusory. Extension tells them the price has detached from any means in a generation. Breadth says the rally is narrow. Sentiment says the marginal buyer is unhedged.

In that environment, Buffett raising cash is not a forecast. It is the rational posture for an operator who reads the same instruments and recognizes the configuration. Bezos' selling is not a forecast. Musk's filing is not a forecast. Each of these is a position taken by an actor with information about their own balance sheet, in a market whose measurements describe the asymmetry of risk in late May 2026.

That’s the context in which the posture makes sense.

VI. The Regulator Buys Optionality

The previous sections concerned private actors. Musk, Buffett, Bezos, Zuckerberg. They are free to position themselves as they choose. The fifth actor in this story occupies a different position entirely. He runs the Federal Reserve, and his choices are not private.

Kevin Warsh was sworn in as Chair of the Federal Reserve on the twenty-second of May. He had served as a Fed Governor under Bernanke between 2006 and 2011, and spent the post-crisis years writing extensively on monetary policy and central bank drift. He is intellectually serious. His confirmation was muted, even relieved.

During his confirmation testimony before the Senate Banking Committee, Warsh signalled a technical change of considerable consequence and very little fanfare. He indicated that the Fed should give greater analytical weight to the Dallas Federal Reserve’s trimmed-mean PCE rather than to the core PCE index, which has been the Fed’s primary inflation reference for more than a decade. He described the latter as offering only a “rough swag” of underlying inflation.

The two measures are constructed differently. Core PCE excludes volatile food and energy prices and reports what remains. Trimmed mean PCE goes further: each month, it discards the most extreme price movements in either direction and averages the remainder. The result is a smoother series. It is also, in inflationary periods, a lower series because the components that contribute most to upward pressure are precisely the ones the trim removes.

Core PCE vs Trimmed Mean PCE

In March 2026, core PCE printed at 3.2%. Trimmed mean PCE printed at two point three. The gap is almost one full percentage point. Against a stated Fed target of two percent, the first measure says inflation is well above target. The second says inflation is approaching the target.

These are not different facts about the world. They are different processing rules applied to the same data. Which one a central banker chooses to emphasize determines whether the central bank’s posture is restrictive or accommodative. The choice does not change the inflation. It changes the room.

Loretta Mester, former president of the Cleveland Fed, captured the discomfort. “I have always been a fan of looking at multiple measures of inflation,” she said, “but not necessarily picking and choosing in a way that changes meeting to meeting.” Krishna Guha at Evercore added the institutional reading: “He would not be the first Fed chair to choose metrics that serve his case. But shopping around for inflation when the indicator generally used is not behaving is risky.”

There is a historical reference that the Fed’s institutional memory tries hard not to invoke. In the early 1970s, Arthur Burns accommodated political pressure from the Nixon administration by, among other manoeuvres, emphasizing measures of inflation that excluded the most volatile components on the argument that these were too noisy to inform policy. The move brought him room to keep rates lower than the headline data warranted. The result, by the time it was unwound, was a decade of inflation that required Volcker’s recession to subdue.

The current move is structurally similar. Trimmed mean PCE is, in a sense, the 2020s version of core – a statistical refinement that produces a lower headline and creates analytical room for accommodation that the unrefined data would resist. Whether it produces the Burns outcome is unknowable in real time. What is knowable is the direction of the change.

The market reaction was predictable. Equities rallied. Front-end rates priced in additional cuts. The Fed put, which had been an article of faith, was reinforced. What the market may not be pricing is that the put requires room, and the room is being engineered, not measured.

That distinction matters because the moment the central bank needs to put a credit event, an equity drawdown, or a labour market crack, it will likely be a moment when actual inflation is still above target. At that moment, the central bank that prepared the metric will be able to cut. The central bank that did not, will not. Warsh, even before taking office, had begun preparing the metric.

The four best-informed actors are now five. The fifth is positioning the institution itself.

VII. What This Means in Practical Terms

Knowing what they’re doing doesn’t mean we replicate it.

Buffett can run three hundred and ninety-seven billion in cash. He can wait five years for a deal. He has an insurance float that requires liquidity, and a portfolio scale that makes marginal acquisitions difficult. None of those constraints applies to the reader of this piece. Bezos’s selling pace, Zuckerberg’s quarterly disposals, Musk’s IPO filing – none of these translates as a script. The question is not what they are doing. The question is what their behaviour implies about the distribution of risk in the environment in which you are making your own decisions.

Two domains deserve attention. The portfolio is one. The business is the other. The previous letter focused on the first; I will be briefer here on portfolio matters and longer on corporate finance, because the corporate-finance window is where you have decisions that Buffett does not.

Portfolio. The framework is unchanged from the prior letter. Cash earns yield and offers optionality – the option to buy when markets correct. Gold as structural ballast against fiscal regimes that will not narrow. Leverage is handled with respect rather than enthusiasm. Convexity bought long, never financed with capital that one cannot afford to lose. The behaviour of the actors in this piece reinforces the discipline; it does not redesign it. If the previous letter’s framework felt right, this letter’s actors are the reason to hold the line.

Corporate finance. This is where the implications are more specific.

Refinancing windows. Investment-grade and high-yield spreads sit near multi-year lows. Credit markets are pricing corporate risk at levels typically seen near previous cycle peaks. If you have debt to refinance in the next eighteen months, the cost of doing it now, even at a slight premium to your target structure, is the price of optionality. Wait, and that optionality disappears if spreads widen a hundred basis points. The most regrettable corporate finance decisions of the cycle after a credit event are not the deals that did not get done. They are the deals that waited for spreads to compress further and then watched the window close.

M&A and capex. Equity multiples for the businesses you might want to acquire are full. Any deal whose IRR depends on today’s financing terms persisting through the duration is overstating its expected return. Two questions worth asking before signing: does the deal still work if credit spreads widen one hundred and fifty basis points before integration completes, and does it still work if equity multiples on the target’s sector compress thirty percent over the same window? If both answers are yes, proceed. If either is no, the deal is not a deal – it is a bet on regime continuation.

Capital structure. The opportunistic equity raise is cheap. If there is growth you are planning to fund in 2026 or 2027 with a future raise, raising now removes the price risk of the raise itself. The cost is dilution at a price that may, ex post, turn out to have been the highest available. That is a risk worth taking compared to the alternative of needing capital in a regime where the cost of equity has doubled.

Pipeline pacing. The M&A pipeline that looked promising three months ago is being repriced upward by sellers who read the same euphoria you do. Slow the pipeline. Tighten the diligence. Make sellers come to you, not the other way around. The asymmetry of doing a bad deal at the top exceeds the asymmetry of missing a marginal deal you could have done.

The skin-in-the-game question. Where do you have positioning that is, in effect, a forecast of regime continuation that you would not place if explicitly asked to bet on it? Operational gearing without a hedge. A capex programme that requires demand visibility through 2027. A balance sheet whose covenants tighten if margins compress. Each of these is a position, in the same sense as the actors above. Each costs you something if the regime resolves the other way. The exercise of this piece is to take inventory of those positions and decide which to maintain consciously and which to reduce, hedge, or restructure.

What the best-informed actors are doing tells you the joint probability they assign. What you do in the domains where you have control tells you whether you have heard them.

VIII. What I’m Looking At

Not buying isn’t the same as not thinking. What follows is a watchlist, not a portfolio – positions where analytical work is in progress, or where I would deploy capital if the conditions of this letter were resolved.

Energy (XLE, XOP, OIH). The sector trades at a multi-decade low relative to the index. Capex discipline at the majors looks structural, not cyclical. Oil services is the operational leverage if commodity prices firm. The setup pays you to wait through dividends and buybacks – you are not paying a multiple to hold it.

Uranium (URA, URNM). The cleanest commodity supply-demand gap of the decade. Utility contracting is accelerating, secondary supply is largely exhausted, and new mines are half a decade or more from the first pound. Volatile, regularly crowded, never boring. I size small and ride the cycle.

Platinum. Trading at one of its deepest discounts to gold in a generation. Above-ground inventories are drawing down. South African supply structurally impaired. The catalyst is not obvious. The asymmetry is – limited downside from here, leverage to any plausible demand story.

China (FXI, KWEB). The most hated trade in global equity markets. Valuations are pricing in something close to extinction, which is too strong a claim for an economy still supported by a consumer base measured in trillions. I am not calling a bottom. I am sizing for the entry price where the math of being wrong is bearable.

Put options. On the most extended parts of the index – semis, megacap tech, and broad NDX. Convexity bought when implied volatility is low is one of the few hedges that doesn’t bleed in a sideways tape. In late May 2026, the cost of insurance is cheap for the asymmetry it provides.

That is the watchlist. The model portfolio comes later.

IX. Close

There is no prophecy in any of this.

I do not know whether SpaceX prices at the top of its range. I do not know whether Warsh’s metric will buy the room he is preparing for, or whether actual inflation will reject the room and force him to choose between the institution’s credibility and the market’s expectation. I do not know whether the actors named in this piece will be vindicated or embarrassed in eighteen months.

What I have written is not a forecast. It is a reading.

Each of these actors has spent decades acquiring information about their own positioning and the markets they operate in. Each is moving, simultaneously, in a direction that costs them something to maintain. None of them is performing for an audience.

Whether they are right is one question. Whether their exposure is information is another. I think it is.

In the meantime, I’m not buying. I’m refinancing what can be refinanced. I’m tightening diligence on what’s in front of me. I’m holding the framework from the previous letter, and adding the inventory question from this one.

The work, again, is unglamorous. Hold the line. Read the data. Read the actors.

PS — The Perfect Storm

A friend frames it well: AI, quantum computing, and robotics aren’t three separate revolutions. They’re one storm, and we’re walking into it.

Each, alone, is transformative. AI compresses cognitive labour by orders of magnitude. Quantum dismantles the asymmetry between problem complexity and computational feasibility — molecular simulation, optimization, cryptography. Robotics finally closes the loop between digital intelligence and physical action.

Investment Musings is editorial opinion, not personalized investment advice. The author is not a registered investment adviser. Nothing here should be read as a recommendation to buy, sell, or hold any security.

A great many of the signals I learned to respect during fifteen years across institutional asset management – including running the investment side of a family office in London – are in the same place. They are all telling me to be careful.

This is not the comfortable kind of caution. The indicators are at extremes I have only ever seen in the historical record. The rally has been parabolic enough to humiliate anyone who tried to fade it. And the participants who lived through the last few episodes – the people who carry the memory in their bodies, not just in their charts – are the loudest voices saying something is off.

Michael Burry put it more sharply than I would have dared. “Stocks are not up or down because of jobs or consumer sentiment,” he wrote on his Substack on 8 May. “They are going straight up because they have been going straight up. On a two-letter thesis that everyone thinks they understand. Feeling like the last months of the 1999-2000 bubble.” He has backed the call with money – January 2027 puts on the SOXX semiconductor ETF, betting on a 30% decline. The Shiller CAPE that morning stood at 40.1, a reading historically associated only with the very peak of the dot-com bubble. On the same day, the S&P 500 hit a new all-time high and the University of Michigan consumer sentiment index dropped to a record low.

That two-letter thesis is AI. And what Burry is naming is specific: a market that has decoupled from the conventional input/output relationship between fundamentals and price.

The dislocation is visible in two charts.

On Friday 8 May, the S&P 500 closed at an all-time high while 5% of its members simultaneously hit 52-week lows. This has happened exactly four times in the last century. The other three: July 1929. January 1973. December 1999.

The point of the chart is concentration. A handful of names are pulling the index higher, while a meaningful portion of the same index is already in its own private bear market. This is not what healthy participation looks like. It is what an index looks like when leadership has narrowed to the point where, mechanically, the headline number has stopped telling you what the underlying market is doing.

The second chart is the recompense. The S&P 500 just delivered one of the largest six-week rallies since 1950. The historical company is striking: every other rally of comparable magnitude in the top twenty has come off a bear-market low. This is the only one on record that happened on the highs.

Put the two charts side by side. The first tells you why caution is warranted. The second tells you why being cautious has cost you 16% over the past six weeks.

This is the dislocation. And the dislocation, not either side of it, is the actual problem.

If this resonates so far, a quick “like” helps the piece travel – it’s the main signal Substack uses to recommend writing. Thank you.

The bull case, fairly stated

The bull case is not crazy. Earnings are real. Net profit margins are running at all-time highs. The largest AI beneficiaries are generating operating cash flow that bears no resemblance to that of pre-revenue companies in 1999. Microsoft, Google, Meta, and Amazon together booked roughly $1.94 trillion of revenue over the last twelve months. Capex is being absorbed by businesses that can actually afford it.

What concerns me is not the bull case itself. It is that the gap between the bull case and the bear case has rarely been this wide. And historically, gaps this wide have resolved in one direction more often than in the other.

What this week added

This past week, the regime stopped being argued and started being shown. Five separate pieces of evidence – not coordinated, not connected, but rhyming.

1. Cerebras Systems priced its IPO at $185 on Wednesday – above an already-raised range, with the order book twenty times oversubscribed. At that price, the company was valued at $56 billion fully diluted, on trailing revenue of $510 million. That is roughly 111% of revenue, with 86% customer concentration in two accounts. On Thursday, the stock popped 68% on debut, pushing the market cap past $90 billion intraday. SpaceX, OpenAI, and Anthropic are reportedly preparing to follow, with SpaceX and OpenAI alone expected to raise a combined $135 billion. The IPO market is not awake. It is on stimulants.

2. Cboe data released this week shows retail traders are now buying calls on the “Mag 10” basket – the Magnificent Seven plus AMD, Palantir, and Broadcom – at the heaviest ten-day pace since 2021. More than half of new positions opened were call-buying. “Hedgers have thrown in the towel,” Mandy Xu, head of derivatives market intelligence at Cboe, told CNBC.

3. Korean equities have rallied 200% in twelve months, the fastest of any major market on earth. Bloomberg reported on Thursday that locals are borrowing record sums to amplify their bets, daily price swings of 5% or more have become routine, and trading volumes have hit all-time highs. The Kospi is now the most volatile major stock gauge in the world.

4. A six-week-old ETF built around Asian memory chipmakers, with Samsung and SK Hynix as its core holdings, has become a mainstream US retail trade. Yahoo Finance reported on Friday, 15 May, that activity in the product is “very lopsided toward call options,” which one Cboe analyst flagged as a classic late-cycle signal in single-name and thematic trades.

5. Berkshire Hathaway’s cash pile climbed to a record $397 billion in the first quarter, more than tripling over the past three years. The most successful long-term investor in the history of public markets is sitting on more cash than at any time in his firm's history. He is not telling you anything. The position is the statement.

Five data points in five days. None of them, by itself, is conclusive. Together, they are the texture of a regime that has stopped pricing anything except momentum – or, put differently, a market priced for perfection.

Which brings me to the question that actually matters. Not “are the bears right?” – they may or may not be, and over a long enough horizon they always are, and over a short enough horizon they often look ridiculous. The real question is:

How do you finance the patience?

What this means, if you have skin in the game

I’ll tell you how I am thinking about it. Not because I think I am necessarily right – I have been wrong enough times to know that’s not the right framing. But because someone reading this might recognize their own situation in mine, it is useful to compare notes.

Cash that earns its yield. Whatever yield your base currency pays, holding cash now is not a punishment. It is the cost of optionality, paid in foregone gains. When the dislocation eventually closes – whether by price falling to meet the signal, or by the signal failing for reasons no one is currently anticipating – the operator with cash is in a position to act. The operator who got dragged into the parabolic leg because he couldn’t stand watching has, at that moment, neither cash nor conviction.

Gold as structural ballast. Not because the technicals are good, but because the asset most correlated with budget deficits is gold, and the budget deficits are not going anywhere. I would not change the allocation here for short-term price action. The position isn’t a trade. It is a position.

Leverage is the thing to be most careful about – and I don’t just mean financial leverage on a portfolio. Operational leverage in a business is exactly the same problem.

In corporate finance terms, late-cycle regimes have a particular feature most CFOs eventually learn from a deal they wish they could undo. I am thinking of the firms that refinanced their debt comfortably in early 2007 versus those that waited, assuming spreads would tighten further, and woke up in a regime where the windows were closed entirely. The easiest financing windows feel routine until they close. Credit spreads – the premium that companies pay above governments to borrow – are sitting near multi-year lows for both investment-grade and high-yield issuers. Translation: the market is barely charging anything for corporate risk right now. Equity multiples for businesses unrelated to AI are being pulled along by the index.

If you have refinancing to do in the next eighteen months, doing it now – even at a marginal premium to your target – is buying optionality you will not have at the same price if the market reprices ten percent down. If you have been considering an opportunistic equity raise to fund growth that does not yet need funding, this is when it is cheap. If you have an M&A on the table whose IRR depends on stretched financing assumptions, this is the moment to stress-test the deal against a regime where credit spreads widen 100bps in a quarter.

The pattern that consistently embarrasses corporate finance teams in the cycle after the euphoria is not the deals that did not get done. It is the deals done at the top with capital structures designed for a regime that no longer exists six months into integration.

Optionality at the corporate level is cheap right now in the same way that index puts are expensive: the market is pricing in continuation, and refusing to charge a premium for flexibility.

Short-dated convexity is expensive and bleeds. The instinct in regimes like this is to buy index puts because they feel cheap relative to the perceived risk, and then watch them melt as the index goes up another 5%. If you want optionality, buy it long-dated, buy it in places where the structural thesis is intact rather than the cyclical one – energy infrastructure, hard commodities, gold miners – and do not finance it with capital you cannot afford to write off entirely.

Financing the patience. This is how I am doing it. Cash earns its yield. Gold anchoring. Leverage that I refuse to add. Convexity that I buy only where it lasts. No prediction in any of it. Just a configuration that lets me wait without bleeding, and lets me act when waiting stops being the answer.

Close

I don’t know if this is 1999, 1973, or 1929. None of us does. What I do know is that the people I trust most – the ones who lived through 1999 and 2000, who carry the memory in a way you can’t get from a chart – are not telling me to relax. The history of those moments is unkind to participants who chased the last leg, and equally unkind to those who tried to short the top.

Are we at the exact top? I have no idea. But I am not buying anything other than some protection. And holding some CHF seems like a pretty good idea too.

The patient operator faces a different set of decisions. Most of them are decisions about what not to do. Don’t chase. Don’t lever. Don’t get cute with timing. The work of these weeks is unglamorous. Holding cash. Reading carefully. Resisting the urge to perform – for boards, for shareholders, for the colleague at the next desk who is bragging about his concentrated tech position.

Eventually, the dislocation closes. The work is to be in a position to act when it does.

P.S. I spent part of this week revisiting the two pieces I wrote in July 2022 – peak inflation, bullish gold at $1,800. The macro was right; the timing, in places, was painful. What I remember most is not the conviction. It was the days when the conviction was at its strongest, and the price action was at its worst. I remember sitting at my desk in London, reading the same data we all read, watching gold drift lower for what felt like months. I remember the conversations with friends in the industry – some who agreed but wouldn’t position, some who disagreed entirely, and the one or two who had positioned the same way and were also bleeding.

“It was never my thinking that made the big money for me. It was always my sitting. Got that? My sitting tight.” — Jesse Livermore, in Reminiscences of a Stock Operator, 1923

The Livermore line wasn’t abstract for me then. It was the practical question of whether to add, hold, or fold – on days when adding felt insane and folding felt like betrayal. I held. The macro arrived. The timing eventually caught up. But the lesson was in the holding, not in the eventual outcome.

This is one of those weeks again. Different position, same work.

If you’ve read this far, thank you – I don’t take that lightly. A like helps the piece travel; a reply (whether you agree or not) helps me write better next week.

Investment Musings is editorial opinion, not personalized investment advice. The author is not a registered investment adviser. Nothing here should be read as a recommendation to buy, sell, or hold any security.