Why This Short Seller Doesn’t Want to Short the AI Bubble | Russell Clark

Russell Clark sees no smoking gun for shorting AI, and considers the risk/reward finely balanced because short selling carries its own risks. The strongest short opportunities, in his view, often emerge when an industry overbuilds, not because its leaders are foolish or fraudulent. The bear case is that compute spending must grow exponentially to generate linear improvements in AI capability, making investment increasingly inefficient and potentially setting up a deflationary bust. But Google and Microsoft may keep spending even if returns look excessive because they view AI as an existential threat, while governments may support investment to stay ahead of China.

Whether demand can sustain the buildout remains uncertain. Hyperscalers say their data-center capacity is fully booked for this year and next, while GPU, rental, and cloud prices are rising rather than falling. Clark says demand could weaken if financing for companies such as Anthropic and OpenAI dries up: some of their apparent revenue may be money transferred from elsewhere in the AI system. Restrictions on data-center construction could also fail to cool prices if demand remains strong, particularly if hyperscalers build elsewhere or developers reduce capacity only after permitting more projects than they intend to complete.

Clark sees equity funding as a key test of whether the boom is ending. SpaceX’s roughly $100 billion equity-funded plan to build its own language model indicated investors were still willing to finance AI, even as credit markets appeared less receptive. A SpaceX collapse could therefore be a meaningful warning signal, while another $200–300 billion of IPO funding could prolong the boom. Clark also says AI-related borrowing is squeezing debt markets and reducing liquidity available to other corporate borrowers.

Oracle is not yet an obvious short, in Clark’s assessment. It has only begun reducing capex, its reported numbers remain strong, and it has not backed away from investment. The company announced plans to raise $10 billion from equity markets and $10 billion from debt markets, though the funds had not yet been fully raised. Clark attributes some share-price weakness to a shift from buybacks to share issuance, but says the stock is less compelling as a short if capex has peaked and cash becomes available for repurchases. A GMO study he cites found that each 1% of market capitalization issued was associated with about 4% underperformance versus average; SpaceX, Anthropic, and OpenAI coming to market within 12 months at their stated valuations could represent issuance equal to 4–6% of market capitalization.

Clark finds the short side more interesting than the long side, but is unsure AI is the right target. He says US share buybacks remain the trade that still holds up, although rising costs could limit companies’ ability to repurchase shares. He prefers gaining exposure to a guiding theme through positions that are exposed to it but not perfectly correlated, rather than concentrating a portfolio in one trade. Outside AI, he sees US home builders as potential shorts: higher labor, interest, and energy costs could reduce margins from about 15% to 5–7% and prompt substantial stock deratings, though he does not expect a global-financial-crisis-type problem.

Clark favors Japanese banks because of their large deposit bases, low valuations, and ability to reinvest deposits at higher Japanese government bond or Treasury yields. He expects Japanese growth to be good, with current growth and inflation supported by political choices to encourage investment as China is viewed as an adversary. New semiconductor fabs have become more acceptable, and Clark sees rising property prices as more politically acceptable in Japan than in many other markets. He says inflation and a weak currency have nevertheless become more salient to voters, including through anti-tourist rhetoric.

Clark attributes rising bond yields mainly to politics, large fiscal deficits, and wage growth, while allowing that demand for capital may contribute. He sees higher defense spending and investment making capital and deposits scarcer, and believes the US fiscal outlook implies higher interest rates than currently charged. In his view, rising wages in a tight labor market, rather than oil, are the structural driver of inflation; expectations of rate cuts can prompt long-term bonds to sell off if investors think wage-driven inflation remains unaddressed. He says oil can still rise and move markets in the short term, but sees road and driving oil consumption topping out or declining across China, Europe, and the US, making oil a potential distraction over the longer term.