Gartner estimates that more than $3.5 trillion of AI infrastructure spending could be exposed to impairment by the end of 2030 if investment keeps its current pace. Its analysis says data-centre owners would need about $8.5 trillion in cumulative revenue to cover their costs and a reasonable return. Revenue attributable to AI over the same period is expected to reach no more than $3.6 trillion. Gartner also predicts that over 80% of specialised AI cloud providers will no longer operate independently by 2030.

What it means: The forecast does not mean a crash is coming, but it puts numbers on the question markets keep asking: how many of the data centres being built today will earn back their cost. If the gap between spending and revenue holds, expect acquisitions, mergers and smaller providers leaving the market. For Greek businesses buying AI compute, that has two practical consequences. First, choosing a provider should account for its financial staying power, not just the hourly GPU price. A cheap specialised provider that shuts down or gets acquired can leave a service without infrastructure or on new, worse terms. Second, for financial firms DORA already requires exit plans for critical ICT providers, and Gartner's analysis gives a concrete reason to write those plans properly rather than as a formality. On the other hand, a possible oversupply of infrastructure could push prices down over the next few years, which helps companies that use AI without investing in their own hardware. The safest strategy for a small or mid-sized firm is flexible contracts, applications that can move between providers, and avoiding long commitments to a single supplier. Keeping a second provider tested and ready is cheap insurance against the consolidation Gartner expects.

Read the original on Gartner

Related service: Custom software

Related stories