Citrini Flags Downside Risk for AI Infrastructure

Wednesday's market rotation gained support from two competing research views, potentially prolonging volatility in AI infrastructure names as investors look to earnings for guidance on which outlook prevails. Stocks linked to data center construction and energy supply face the greatest exposure to the state-level permitting and cost risks flagged by both research houses. Hyperscaler debt issuance should be watched, given Citrini's caution about slackening demand for such securities. In Asia, the rift directly influences AI-intensive companies such as SoftBank, where valuations hinge on the global build-out narrative.

---

UBS and Citrini share a common view on the location of AI risk, but they are in total disagreement over its scale.

Summary:

• UBS lifted its AI capex projections to $1.0 trillion for 2026 and $1.4 trillion for 2027, citing sustained demand
• Citrini argues that AI infrastructure investment growth reached its peak, with hyperscaler spending close to $1 trillion against at least $200 billion in AI revenue
• According to Citrini, risks include hyperscaler bond issuance, declining backlog growth, and AI price reductions
• Both firms view state-level decisions, particularly in Texas, as more consequential for data centres than congressional control
• Citrini points to a potential shift toward late-cycle stagflation, while UBS maintains a constructive equity outlook but advises reducing tech concentration
• On Wednesday, AI infrastructure stocks experienced sharp declines even as the broader US market advanced

Two leading research firms drew opposing conclusions about the AI spending surge after examining the US election outcome. UBS increased its capital spending forecasts, while Citrini Research stated that growth in AI infrastructure investment has already peaked.

UBS raised its aggregate AI capex projections to $1.0 trillion (2026) and $1.4 trillion (2027), maintaining that demand for AI infrastructure is still robust. It does not foresee tighter congressional oversight after the Democratic victories disrupting the investment cycle, because any sweeping federal restrictions would need to overcome legislative and presidential obstacles. Nonetheless, UBS acknowledges that investors are focusing more on the risks of overinvestment, funding, and monetization, and it predicts that volatility will remain high.

Citrini holds a much more cautious stance. Its latest State of the Themes report, released Wednesday, contends that annualized hyperscaler capex is nearing $1 trillion, while AI ecosystem revenue stands at a minimum of $200 billion — a disparity it considers unlikely to narrow. It cites decelerating backlog expansion, substantial debt and equity offerings by major tech firms, and price reductions by AI providers as indicators of pressure, and asserts that the risk balance for AI infrastructure has moved to the downside.

The two firms concur on the location of political risk. Both regard state-level decisions as more critical for data centres than the shift in congressional control. UBS pointed to state government compositions, including Texas's outcome, as potentially more influential for specific projects. Citrini went a step further, viewing governor races in Texas, Ohio, and Georgia as a referendum on the infrastructure build-out and anticipating increased costs due to local concessions, clean energy mandates, ratepayer protections, and slower permit approvals. It mentioned a Texas moratorium that stopped a project tied to $15 billion in bridge loans.

The dispute is already reflected in pricing. On Wednesday, a broad rally in US stocks occurred alongside sharp declines in AI infrastructure equities. According to investingLive, Nvidia dropped roughly 3%, while Micron and Marvell fell about 7%.

Citrini additionally cautioned that the overall macroeconomic environment is moving toward late-cycle stagflation, observing that most of the asset pairs it monitors currently point in that direction and that cyclical shares have lagged defensive ones over the last half-year. In contrast, UBS keeps a positive stance on equities, pointing to robust earnings, but advises investors to reduce their excessive technology holdings.

For market participants, the division means the AI trade hinges on one question: can revenue expand quickly enough to validate the spending.

Citrini Research, a thematic investment research company, was established by James van Geelen and distributes its work mainly via a subscription newsletter reaching hundreds of thousands of subscribers. According to Bloomberg, it gained significant notice in February when a widely circulated scenario analysis about AI's potential economic disruption was associated with a decline in AI-linked equities. SemiAnalysis, a firm specializing in chip and data center research, acquired Citrini in September, with van Geelen continuing as CEO.