UBS Lifts AI Capex to $1.4T as Citrini Calls Peak

UBS raised its aggregate AI capital expenditure forecasts to $1.0 trillion for 2026 and $1.4 trillion for 2027, while Citrini Research said on Wednesday that AI infrastructure investment growth has already peaked. The disagreement landed as US stocks rallied broadly but AI infrastructure names fell sharply: Nvidia dropped about 3%, with Micron and Marvell each down roughly 7%, according to investingLive. Citrini put annualised hyperscaler capex near $1 trillion against at least $200 billion of revenue across the AI ecosystem, a gap it doubts can close. UBS said demand for AI infrastructure remains strong.

Context — Why the AI capex split matters now

Two research houses read the same US election results and drew opposite conclusions. UBS lifted its AI capex forecasts on the view that demand stays strong; Citrini argued in its State of the Themes report that the balance of risk for AI infrastructure has shifted to the downside. The report gives no prior capex forecast from UBS to compare against, so the revision's size relative to earlier guidance is not disclosed.

What both agree on is where the political risk sits. State-level decisions matter more for data centres than the change in control of Congress. UBS pointed to the composition of state governments, including the Texas result, as potentially more consequential for individual projects.

Citrini went further, treating governor races in Texas, Ohio and Georgia as a referendum on the buildout. It expects higher costs from local concessions, clean energy requirements, ratepayer protections and slower permitting.

The catalyst chain runs through project economics. Citrini cited a Texas moratorium that halted a project linked to $15 billion in bridge loans, a concrete example of how a single state decision can stop financing already in motion. UBS argued that broad federal restrictions would face both legislative and presidential hurdles, which is why it does not expect tighter congressional oversight to derail the cycle.

Data — What the numbers show

UBS's 2026 AI capex forecast of $1.0 trillion and 2027 forecast of $1.4 trillion imply a 40% increase between the two years. Against that, Citrini's estimate of at least $200 billion in AI ecosystem revenue leaves a wide funding gap.

Metric	UBS view	Citrini view
2026 AI capex	$1.0 trillion	Growth already peaked
2027 AI capex	$1.4 trillion	Approaching $1 trillion annualised
AI ecosystem revenue	Demand remains strong	At least $200 billion
Political risk	State-level, Texas key	Texas, Ohio, Georgia governor races

The single-session moves show the split is priced, not theoretical. Nvidia fell about 3% and Micron and Marvell each lost around 7% on Wednesday, even as the broader US market rallied. Citrini also flagged slowing backlog growth, heavy debt and equity issuance by the largest technology companies, and price cuts by AI providers as strain signals.

UBS, for its part, noted that investors are paying closer attention to overinvestment, financing and monetisation risks, and expects volatility to stay elevated. Neither house gave a specific index level, yield or moving average to anchor the trade.

Analysis — What it means for markets and sectors

If Citrini is right about the revenue gap, the pressure falls first on names tied to data centre construction and power, where permitting and local cost concessions bite hardest. A stalled Texas project tied to $15 billion in bridge loans shows how quickly state decisions can freeze financing.

Hyperscaler debt issuance is the second pressure point. Citrini's concern about softening demand for those securities matters because the capex figures UBS cites must be funded somewhere. Slower backlog growth and AI price cuts would compress the returns that justify that issuance.

In Asia, the debate feeds directly into AI-linked heavyweights such as SoftBank, where sentiment toward the global buildout drives pricing. Any shift in how investors read US hyperscaler spending transmits to those names.

The counter-argument sits with UBS. It remains constructive on equities, citing strong earnings, and does not expect tighter congressional oversight to derail investment given the legislative and presidential hurdles. Its recommendation is to trim oversized technology exposure, not exit. That is a positioning signal: stay long the theme, cut the concentration. The acknowledged limitation on Citrini's side is that its $200 billion revenue figure is described as at least, leaving the true gap uncertain.

Outlook — What to watch next

The next earnings season is the arbiter both camps now point to. UBS's demand-stays-strong case and Citrini's peaked-growth case cannot both be confirmed by the same set of hyperscaler results, so guidance on capex and backlog will be read as a verdict.

State-level decisions are the second catalyst. Citrini expects the Texas, Ohio and Georgia governor races to function as a referendum on the buildout, with higher costs from local concessions, clean energy requirements, ratepayer protections and slower permitting. The report gives no dates for those decisions.

Hyperscaler debt issuance is the third. If demand for those securities softens, the funding math behind UBS's $1.0 trillion and $1.4 trillion figures comes under scrutiny. Citrini also warned the broader macro backdrop is shifting toward late-cycle stagflation, noting most asset pairs it tracks lean that way and cyclical stocks have underperformed defensives over the past six months. No levels were given.

Frequently Asked Questions

What does the UBS vs Citrini split mean for AI infrastructure investors?

It means the same sector now carries two credible but opposite research anchors. UBS sees 2026 AI capex at $1.0 trillion and 2027 at $1.4 trillion with demand strong. Citrini sees growth already peaked, with hyperscaler capex near $1 trillion against at least $200 billion in AI revenue. Investors holding data centre, power and chip names face elevated volatility until earnings settle which view holds.

Why did Nvidia, Micron and Marvell fall while the US market rallied?

AI infrastructure names fell sharply on Wednesday even as the broader US market rose. Nvidia dropped about 3%, while Micron and Marvell each lost around 7%. The divergence points to rotation out of the AI buildout trade specifically, not a broad risk-off move. Citrini's report, published Wednesday, gave investors a research anchor for trimming AI infrastructure exposure while staying constructive elsewhere.

Who is Citrini Research and why does its view move markets?

Citrini Research is a thematic investment research firm founded by James van Geelen that publishes largely through a subscription newsletter with hundreds of thousands of subscribers. It drew wide attention in February when a viral scenario piece on AI-driven economic disruption was linked to a selloff in AI-related stocks. In September, chip and data centre research firm SemiAnalysis acquired Citrini, with van Geelen staying on as chief executive.

Bottom Line

UBS and Citrini agree AI risk sits in data centres and power, but split on whether spending growth has peaked or is still accelerating.

Disclaimer: This article is for informational purposes only and does not constitute investment advice. CFD trading carries high risk of capital loss.