# Citrini Cuts Chip Exposure as AI Funding Shifts to Equity and Debt

By Enna Lee

Canonical URL: https://www.tokenpost.com/news/investing/40774
Published: 2026-10-26T03:36:32.000Z
Updated: 2026-10-26T03:36:32.000Z
Section: Investing

> Founder James van Geelen said new AI infrastructure spending is more exposed to investor sentiment. He estimated AI labs need about $1 trillion in annual recurring revenue by the end of 2028.

Citrini Research has reduced its semiconductor exposure as founder James van Geelen warns that a shift to equity and debt funding has made AI infrastructure spending more vulnerable to investor sentiment.

In the first three to four years of the AI buildout, hyperscalers largely financed new infrastructure with operating cash flow, Van Geelen said in a conversation recorded Wednesday, Oct. 21. He said most incremental investment now comes from equity and debt markets. A sharp downturn in investor appetite could therefore affect construction budgets alongside company valuations.

Under the earlier funding model, hyperscalers’ cash flow helped insulate infrastructure spending from swings in market sentiment. Van Geelen’s concern is that relying more on external financing ties new construction more directly to investors’ willingness to provide capital.

Van Geelen put the annual recurring revenue (ARR) needed to support current AI infrastructure valuations at about $1 trillion by the end of 2028, compared with a current base of roughly $13.6 billion. He described the target as a test of whether AI businesses can generate enough customer revenue to sustain additional construction.

His demand concerns include his personal view that frontier models already meet 99% of his everyday needs. He also cited an estimate linking a 10% decline in task costs with a 3% increase in usage. Van Geelen estimated open-source models’ share of new ARR has climbed to about 18% from roughly 1% a year earlier. He argued the key question is whether paid AI demand can grow enough to support the cost of building infrastructure.

That question is about the economics of frontier AI spending, not whether models will improve, Van Geelen said. He sees white-collar work as the application that would have to account for most token demand growth in the coming years. He also said current demand exceeds supply.

The shift in portfolio positioning is selective: Van Geelen remains bullish on Nvidia and sees opportunities in industries adopting AI. He said Citrini would reconsider its semiconductor exposure if paid usage continued to grow and AI labs’ ARR growth reaccelerated enough for customer cash flow to support new construction. He added that his concern would weaken if current demand is mainly constrained by supply and new products prompt customers to pay for improved capabilities.