Gartner: $3.5T of AI infrastructure investment at risk

Gartner warned that more than $3.5 trillion of AI infrastructure investment could be exposed to impairment by 2030, a specific estimate that frames a broader market rout in AI and chip stocks.

The research firm’s projection arrived as shares of major chipmakers and AI infrastructure providers slid in recent trading, and as reports flagged rising financing costs and heavy debt loads at data center operators. The selloff has hit names including Nvidia, AMD, SK Hynix and Micron, and coincided with a pronounced single-session decline in the Nasdaq.

Debt strain at AI infrastructure providers

Market concern has focused on companies that provide GPU-based compute capacity to AI developers. Reports indicate that the shares of the company identified in Persian-language coverage as کورویو fell roughly 25 percent across five trading sessions, while the prices of its bonds came under pressure. Yields on some of its debt maturing in 2032 have approached the 15 percent range, reflecting heightened investor-perceived risk.

Analysts note that operators building AI-focused data centers face large upfront costs for graphics processors, construction, power and networking. If revenue from leasing compute capacity does not expand quickly enough, lenders and equity investors may see higher default risk or downward revisions to asset valuations.

Gartner’s forecasts and the investment gap

Gartner estimates companies will spend about $1 trillion on AI infrastructure in 2026 and an additional $6 trillion between 2027 and 2030. Across that entire period, Gartner projects that datacenter-related revenue tied to enterprise and consumer AI use cases will amount to at most $3.6 trillion.

According to Gartner’s calculations, datacenter owners would need about $8.5 trillion in cumulative revenue to cover capital expenditures, financing costs and to achieve a 12 percent annual return on invested capital. The firm warns that more than $3.5 trillion of expected infrastructure investment could therefore be at risk of impairment by 2030. Gartner also projects that over 80 percent of so-called neocloud providers—specialist cloud operators that supply training and inference capacity—may cease independent operations by 2030, whether through mergers or business-model change.

Those projections have circulated as investors weigh the sustainability of the AI-driven investment boom and the pace at which demand will materialize into profitable utilization of expensive hardware.

Inflation and higher borrowing costs have amplified market anxiety. The US Consumer Price Index rose 3.7 percent year on year in September, up from 3.4 percent in August, while core inflation, which excludes food and energy, increased to 2.6 percent. Higher inflation increases the risk that the Federal Reserve will keep interest rates elevated or raise them again, raising financing costs for companies that are heavily leveraged while building data centers.

The current episode of selling has not been confined to smaller or loss-making firms. Even companies that reported positive results and upbeat outlooks have seen share prices decline amid broad market pressure. For example, TSMC—the world’s largest contract semiconductor manufacturer—reported strong demand and a favorable revenue outlook but nevertheless saw its stock fall during the recent market downturn. ASML likewise issued positive reports yet could not escape the broader selloff.

Investors and analysts say the interplay of high capital intensity, rising financing costs, and uncertainty about near-term revenue from AI workloads is creating a reassessment of valuations across the semiconductor and AI infrastructure supply chain.