A Feature Not A Bug

The Credit Strategist June 2026

Whether the stock market or just certain sectors are in a bubble (more likely the latter), the rally faces resistance from rising interest rates. While stocks are rising, so are long-term interest rates and if they rise too far (as I expect them to) then the rally will end. The question, of course, is what level will constitute “too far”? I think markets will start to flinch when we see 10-year Treasuries hit 5% though even that may not be enough to stop this train. It will probably take a combination of higher rates and some weakening in the AI trade to slow things down.

Last month, I featured several charts (borrowed from SocGen’s Albert Edwards) showing rising global long-term interest rates. Those rates rose further in May to the highest levels in two decades. They aren’t high enough to burst the bubble but they are moving in a troublesome direction for markets and government finances. Commentators attribute rising rates primarily to higher oil prices resulting from the Iran war. But the growth of debt not just in the U.S. but in Europe, Japan, and China. One way or the other the war will end, but debt accumulation will never end. Debt is a feature, not a bug, of the global economic system. The Ponzi finance structures of governments are essential to their existence (borrowing money without repaying it in constant dollars if ever).

So while reopening the Strait of Hormuz (the aspect of the war with the most immediate economic impact) should lower oil prices and relieve some inflationary pressures (though not as quickly as promised), the rising debt will maintain pressure on global long-term interest rates. Monthly inflation data and bond trading data are merely data points along an upward arc of higher prices, higher bond yields, and higher borrowing costs demanded by investors funding endless government Ponzi schemes. This is the nature of things. And as the world continues on this course, the system keeps devaluing fiat currencies to sustain itself. This is also the nature of things (which means everyone should be buying gold as a hedge). One of the factors driving stocks higher – one rarely discussed or acknowledged – is the continuing devaluation of the fiat currencies in which these financial instruments are denominated. Over time, currency devaluation/inflation is a significant contributing factor to higher equity prices. It is also in the nature of things that the more a bubble inflates, the less fiat-currency denominated stocks will be worth because rising prices contribute to the very financial asset inflation that chews away at their value. This is not merely ironic; it is structural.

AI

AI bulls claim this time is different than the last technology bubble in 2000. The bull market in hyperbole is running wild. AI no doubt allows the solution of problems that vexed mankind for years (like curing certain diseases or solving certain mathematical and scientific problems) due to its ability to analyze mass data in incredibly short amounts of time. Even with its imperfections, which are being cured at a remarkably rapid rate, this is an achievement on par with the most important landmarks of human thought. There are certainly meaningful differences between the Internet companies that attracted high valuations and ultimately failed two decades ago and many of today’s AI-related companies. Transformative as it is, however, we still don’t know precisely how AI will affect jobs, earnings, and the like. Wall Street is clearly ahead of itself in extrapolating or even imagining the future for AI, but that is Wall Street’s job – to finance he future (especially the venture capital business). It doesn’t have the luxury of waiting to see which businesses will succeed. Its mission is to fund as many promising new businesses as possible before they prove their worth and let markets sort out winners and losers. This is the gravamen of venture capital (and where venture capital separates itself from private equity, which invests in businesses using debt only after they prove themselves).

So naturally many AI companies deemed most promising by the market aren’t remotely profitable or in the case of AI losing billions of dollars building data centers or more likely announcing plans to build them on earth and in space without actually breaking ground (or whatever you break in space). But we are still in the very early innings of this effort. The industry is far from proving that many of its announced projects won’t turn out to be vaporware. Right now, there remains a yawning gap between reality and fiction.

According to British research firm Panmure Liberum,most of the AI build-out only exists on paper. Of the 811 gigawatts in additional U.S. data center capacity planned through 2030, only 37 gigawatts (4.6%) are currently under construction with only a further 146 gigawatts (18%) backed by firm commitments. It is going to be challenging to find financial institutions to finance the balance, especially if the private credit industry doesn’t turn things around. The remaining 628 gigawatts (77.4%) exists only in forecasts and press releases and podcasts or as Panmure Liberum puts it “in Excel spreadsheets of investors and businesses.” This is reminiscent of the Internet companies who built or planned to build enough fiber to circle the universe (many of whom borrowed money to do so and ended up in bankruptcy) long before demand for fiber caught up to supply. AI hyperscalers shifted from self-funding to borrowing to finance their data center buildouts (with much of the financing hidden off-balance sheet). Most, but not all of them (CRWV, APLD), are capable of handling this debt but all are shifting their business models from asset-lite to capital-heavy and will be left with big debt burdens to service if AI doesn’t prove as profitable as they expect. If there is a difference today from the Internet Bubble, it is the circular nature of today’s financing structures which feature a different kind of fragility that that seen in the earlier period

Trillions of dollars are being spent without any assurance it is being spent wisely. Demand for chips feels infinite today, but it only feels that way. Nothing is infinite except the universe. There are real limitations to demand that will hit sooner than later (especially in space). Companies are over-ordering and advance-ordering to secure scarce supplies. Will companies need to continue buying so many chips once their models are up and running, or will their capex demands level out? xAI’s Grok quickly fell behind its competitors and ended up leasing its excess data center capacity to Anthropic on the cusp of the latter’s IPO but may have knocked it out of the LLM race. How many LLMs do we need? How many data centers do we need? Nobody really knows. Jensen Huang says chip demand will hit $4 trillion in 2030. How are companies going to pay for that? AI revenues are unlikely to generate $4 trillion in revenue by then. What about the competition from cheaper Chinese models. Is anybody seriously looking at how Chinese models like DeepSeek are undercutting Anthropic’s and OpenAI’s pricing and whether American companies’ pricing models are sustainable (which may be why politicians are considering banning Chinese open-source models?). And what about Small Vertical Small Language Models (SMLs) that some argue can be trained on specific tasks and run on a desktop computer or even a mobile device rather than needing data centers to run (unlike LLMs). Will SMLs eat away at LLMs? Companies are already waking up to the high costs of tokens (AMZN employees reportedly ran up $500 million in token costs in a month which probably upset Jeff Bezos as much as his rocket blowing up) If we are still in the early innings of AI as virtually every respected participant and observer argues, then it is far too early to know how all these things will play out. Yet the expected capex growth of hyperscalers from 2025 to 2028 is around 29% per year (off trillion-dollar bases). These are the largest and riskiest bets in history. If they go awry, or are delayed in any significant way, the ramifications for markets and the economy are profoundly negative.

Everybody is rushing to buy the same stocks (in many cases purchasing them through different ETFs that own the same underlying stocks without) asking questions. There are more ETFs than individual stocks today which if you think too hard is absurd (and you will get a migraine like me). SpaceX stock will be included in so many ETFs that that virtually every investor will own it whether they buy it or not. ETF-dominated market structures that didn’t exist during the last bubble render this bubble much more powerful than the earlier one. SpaceX’s IPO, which is partly a datacenter play after xAI leased out its unused data center capacity to Anthropic (though there is some confusion over the length of the lease) will benefit enormously from inclusion in index funds and ETFs that will boost demand for the stock without regard to the aura surrounding Elon Musk. These are all classic signs of a bubble, one severely divorced from the fundamentals that determine what stocks are worth.[6]

And it is a glaring - not a glancing - risk factor that this is happening against a fragile macroeconomic background rendered fragile by epic levels of public and private debt and geopolitical instability. Michael Burry correctly noted that “the market is capitalizing the most expensive phase of AI adoption as if it were normal and indicative of future demand.”[ The bubble relies on belief in the infinite capabilities (and profitability) of AI as well as the ability of governments to print infinite amounts of money. Neither belief is true. As I wrote earlier, nothing is infinite except the universe. But unsustainable debt and this stock market bubble are joined at the hip, and the hip is rotting and will eventually need to be replaced.

Still, if only for historical purposes, it is instructive to compare current stock market behavior to the Internet bubble in 2000. Through the third week of March 2000, the top ten performing stocks in the Nasdaq Composite were up 622%. Through May 5^th, BTIG points out that they were up 784%. And they’ve risen even higher since then. The hottest stocks in 2000 had higher P/E ratios than those of NVDA, GOOG and MSFT today. But the S&P 500’s trailing and cyclically adjusted P/E multiples are almost as high. This is partially due to market structure (ETFs) as noted above. There is also much more capital in the system today to pump up these stocks due to the trillions of dollars manufactured out of thin air by governments since the 2008 financial crisis. Today’s market bubble is much larger than the Internet Bubble. There are differences but the underlying etiology is the same. And it will run its course the same way. Those who invested early in AI will earn enormous profits and those who hold on too long will see those profits vaporize. These are the laws of investing because they are the laws of human nature. AI can’t change that. AI is, after all, based on human nature. It is just human nature on speed.

Government Fueling the Bubble

Rockets require fuel and stock market bubbles require liquidity to lift them into orbit. The government is happily providing that fuel. The great economist Lacy Hunt provided a compelling explanation for the current stock market moonshot at the Mauldin Conference last month when he pointed to the Federal Reserve reinitiation of QE-by-another name (they called it “Reserve Management Purchases or “RMPs”) with $40 billion monthly purchases of Treasury bills. While Treasury described this as a technical liquidity or “plumbing” operation, Mr. Hunt described it more accurately as QE:

“The Fed said that they were upping the bill purchases because the banks were short of liquidity, and this was a technical operation. Nothing could be further from the truth. This was not a plumbing issue. If the banks were in dire need of liquidity…the bulk of those purchases would have gone into idle balances, but they were not. They were directly used and explosively sold.”

In fact, the money flowed directly into bank balance sheets and was put to work in the economy (including the market). Loans and leases grew at nearly a 10% annualized rate since then while commercial and industrial lending grew at twice that rate. Instead of ending back on the Fed’s balance sheet, that money was multiplied throughout the system, expanding the money supply and increasing inflation (including financial asset inflation).

This monetary policy stimulus is in addition to the two trillion-dollar annual fiscal stimulus pouring into the economy, funded largely with short-term paper. This short-term funding raises another problem. Double Line (quoted by SocGen’s Albert Edwards) points out the following troubling scenario regarding U.S. finances:

“Nearly 70% of U.S. fixed-income issuance over the past year has come from one source: Treasury bills. It is a subtle recalibration of the government’s balance sheet – a shift toward shorter-term debt as deficits deepen. The world’s largest borrower is increasingly financing long-term fiscal obligations with instruments that mature in a matter of months… This trade-off reflects the short-term temptation at the heart of fiscal engineering. Issuing short debt reduces cost today but magnifies risk tomorrow.”

Treasury Secretary Bessent was rightfully critical of his predecessor for failing to take advantage of low long-term rates to manage the Treasury’s borrowing needs during the Biden years. Those low rates are no longer available (not should they be), forcing Mr. Bessent to follow the same path to keep the cost of servicing the deficit as low as possible. Unfortunately, this increases pressure on incoming Federal Reserve Chairman Kevin Warsh to lower interest rates which would be imprudent with inflation running well above target for years. Fiscal dominance will continue to influence monetary policy in the years ahead despite being a path to high inflation and unsustainably higher debts. But for now, we can expect Treasury to keep borrowing short-term to fund growing long-term obligations while praying that the short end of the curve won’t follow the long end too much higher. However one chooses to look at things, it is difficult to forecast a future of lower inflation or lower indebtedness for America unless you buy into the AI optimists’ argument that technology will produce enough economic growth and productivity to solve our self-made problems. Reading Iain Banks’ Colony series of science fiction novels may help you through. It has inspired them to believe we can achieve a future in which these earthly problems are solved.

The SpaceX IPO could mark the zenith of the bubble. The Blackstone IPO did so before the 2008 financial crisis; it took years for BX stock to perform. Then again, maybe the market will wait for Anthropic and OpenAI to IPO before the trifecta of these huge IPOs mark a stock market top. But parabolic market momentum in semiconductor stocks and sectors related to semis and AI is going to prove unsustainable before long. Not only does the math undergirding these epic stock moves not add up, but even the massive money creation supporting them is going to reach its limits (set by the bond markets). Financial institutions also need to decide how much more exposure they want to the most overvalued market sectors. While some firms like APO have reduced their exposure (though APO is out leading a $35 billion data center debt deal for META who seems to be falling behind in the LLM race for the moment), Wall Street remains “risk-on” and keeps pushing things to the edge. With the S&P 500 trading at 22x forward earnings and 41.6x cyclically-adjusted earnings, there is little room for error. Investors should not short this market unless they are professionals (and even professionals should use puts and be very judicious). The momentum train is as powerful as any we’ve ever seen and as noted above market structure is as pro-cyclical as we’ve ever seen (and much more so than during the Internet Bubble). But at some point, this market will collapse of its own weight and owning stocks trading like baseball cards is best left to people of the age who play with baseball cards.