Citrini Declares End of AI Infrastructure Trade, Warns of Late-Cycle Expansion
Cointime reports: On Wednesday, November 5, US stocks suffered their largest selloff of the year, with the Nasdaq falling as much as 6% intraday. AI infrastructure stocks bore the brunt, with many names in optical chips, liquid cooling, and power grids dropping over 10%. The same day, research firm Citrini—which last year triggered an AI selloff with a single report—announced that the “most optimistic expectations” for the AI infrastructure trade are “already in the past.” The firm has overhauled its model portfolio, shifting from betting on the continued rise of mega-themes like AI to reversal trades, pair trades, and entirely new approaches. Citrini believes the market has entered a “late-cycle expansion,” and the investment playbook must change.
Why the Trade Is Over
Citrini cites several reasons. First, timing: when Citrini launched in May 2023, stocks were cheap, sentiment was depressed, and the AI revolution was just beginning—a broad-brush approach worked. Its index has since returned more than 2.5 times the S&P 500. But over the past year, the narrative has shifted from “cheap, proven winners” to “speculative, loss-making companies dependent on 2030 market sizes.” Stocks are no longer cheap, and sentiment is no longer depressed.
Second, the gap between revenue and capex is too wide. AI ecosystem annualized revenue is at least $200 billion—unimaginable in 2023—but capex is approaching $1 trillion. Bridging the gap requires either order-of-magnitude revenue growth or a slowdown in capex; Citrini thinks both may happen somewhat. If capex doubles again, required revenue would start at “trillions,” which won’t happen. Thus, Citrini calls this the peak of capex growth.
Third, the giants are out of money, and capital markets are growing weary. Hyperscalers with a combined market cap over $10 trillion can no longer fund these projects from cash flow alone, forcing heavy debt and equity issuance. Each round drains liquidity and dilutes enthusiasm, and demand for these offerings is already weakening.
Fourth, bottlenecks have shifted from “trade themes” to “real risks.” Power, grid connection, and policy—once momentum-trade fodder—are now execution risks. If construction delays persist, gas turbine orders go undelivered, and costs rise, investors will start comparing postponed completion dates with debt maturities.
Finally, with a blue wave in the midterm elections, policy direction is clear. Texas, Ohio, and Georgia—home to the largest data center clusters—held gubernatorial elections that amounted to a referendum on data center construction. New governors are unlikely to halt construction outright, but costs will almost certainly rise sharply: more local concessions, stricter clean energy requirements, and protection for electricity ratepayers.
The report notes that last week, banks were stuck with a $15 billion bridge loan for an Anthropic-backed data center project that was halted by a Texas data center moratorium. Citrini calls the Texas moratorium one of the most potentially disruptive variables in the AI capex trade. It also says credit markets without hyperscaler guarantees have “completely stalled.” Emerging cloud providers, high-yield private credit, and BDCs seeking financing can only “wish themselves good luck.”
Positive but Wary on Anthropic
Despite widespread skepticism, Citrini is relatively positive on Anthropic. The firm notes that Anthropic’s prospectus is “much cleaner than some expected.” The company has effectively shifted most capital-intensive construction risk to others, sits on substantial cash, and turned operating cash flow positive far earlier than most anticipated. Its Opus 5.5 is performing well.
But Citrini remains wary of a price war. On July 30, OpenAI slashed token prices to regain share, with clear effect—a clear signal of commoditization. Citrini likens this to “the beginning of the end of the great railway cycle.” Anthropic’s explosive growth then “hit a wall,” with share immediately flowing to OpenAI, because in an industry with highly similar products and near-zero switching costs, customers are price-sensitive. Citrini believes both labs have long-term value, but the conditions needed to sustain capex growth are likely unattainable in reality.
The Coda, Not the End
Citrini emphasizes that the key lies in expectations, posing three questions: Will the labs’ annualized revenue growth next year be higher or lower than the past year? Will data center delivery progress in 2027–2028 likely accelerate or be delayed? Will financial markets be more or less willing to lend to this theme?
Citrini stresses it does not doubt that token usage, token spending, and AI adoption will continue to grow. What it doubts is whether the companies that have captured most of the value so far can continue to be good investments.
But has the music stopped? Citrini clarifies this is not “the music stopping” but rather a “coda”: the tone and rhythm have changed, as a transition.
Macro data show the job market posting broad vacancy increases for the first time in four years, with cyclical manufacturing employment rebounding. Surveys show service and manufacturing sentiment at multi-year highs, with price pressures still rising. Other growth drivers persist—tax cuts, tariff easing, and reshoring—even without AI. Labor supply is shrinking: Fed estimates show that with net immigration turning negative, the monthly job additions needed to keep employment stable by end-2026 have fallen to 5,000–10,000, while the three-month average nonfarm payroll gain is about 51,000, so even weak-looking numbers are absorbing supply. Electricians remain scarce; even if AI investment hard-lands, projects under construction will take years to complete. Storage and power bottlenecks can be eased by boosting output, but “you can’t conjure new electricians out of thin air.”
Yet asset prices already show stagflationary color—cyclicals vs. defensives are down 2.4% over one month, 2.9% over three months, and 6.9% over six months. Citrini concludes the market “smells like a late-cycle expansion.”
New Investment Approach: From High-Beta Baskets to “Small Ball”
Previously, Citrini used broad baskets around mega-themes, with high-beta stocks amplifying gains in bull markets and losses in bear markets. Now its trade-off: if you have reaped rich bull-market gains, perhaps it’s time to take profits and go home—or start playing “small ball.”
The new portfolio themes include: AI as relative value—pair trades rather than outright longs; life sciences and healthcare, fiscal dominance, robotics, and modern warfare, all with adjusted expressions; and four new themes: “Boomer Software” (SaaS survivors), alcohol and beverages, financials (“earn interest while waiting”), and entertainment (scarcity premium).
As New York’s changing autumn leaves remind us of a basic truth—there is a time for everything, especially in markets.
