Gartner: $3.5T AI infrastructure at write-down risk by 2030

Gartner just put a number on the AI infrastructure gap: more than $3.5 trillion of spend could be exposed to impairment by the end of 2030. The October 14 Gartner release also predicts that more than 80% of neoclouds, the GPU-rental companies that sit under a lot of AI products, will stop operating independently by 2030. If a small vendor runs your models, this is your problem too.

What actually happened

Per Gartner's release and The Next Web's write-up of the research note:

• The spend. About $1 trillion of AI infrastructure spending in 2026, and about $6 trillion more forecast for 2027 through 2030.
• The gap. Owners would need about $8.5 trillion in cumulative revenue through 2030 to cover that spend, the financing, and a 12% return on invested capital. Gartner's forecast tops out near $3.6 trillion in business and consumer AI revenue tied to those data centers, and less than half of it goes to the companies that own the hardware.
• The neoclouds. More than 80% will lose their independence by 2030. Gartner points to cost pressure and no clear way to stand out. Many may not earn back their GPUs before the next chip generation makes them obsolete.
• The hyperscalers. Microsoft, Alphabet, and Amazon have enough other revenue to carry AI spend for years. Senior director analyst Will Sommer told Bloomberg: "All of them want to be the full-stack. They can't all be."

This is a forecast, not a loss on the books. Demand, pricing, and power supply can all move the number.

Why it matters for your business

You probably do not rent GPUs. But the AI tools you pay for do, and many of them sit on a neocloud or a small inference host you have never heard of. When 80% of that tier gets acquired, merged, or shut down, you get the downstream effects: price changes, new terms, a model that goes away, or an endpoint that moves with 30 days' notice.

The fix is not to panic. It is to know your supply chain and keep a way out:

• Map who hosts what. For each AI tool or API you run in production, find out which company actually serves the model. Ask the vendor. It is a fair question.
• Keep a second route. For any workflow that makes money, test it on a second provider or model now, while nothing is on fire.
• Read the change clauses. Check how much notice your contracts give for price changes, model retirement, and assignment after an acquisition.
• Own the glue. Prompts, evals, and integration code should live in your repo, not inside one vendor's dashboard.

Sources: Gartner press release, The Next Web.