📅 30 October 2026 • techzine.eu. Gartner has published a projection arguing that more than 3.5 trillion dollars of investment in artificial intelligence infrastructure could be written off by 2030, on the basis that the money coming in will not repay the money going out. The arithmetic behind the warning is straightforward. AI providers are expected to spend close to 1 trillion dollars on infrastructure during 2026 alone, with roughly 6 trillion dollars more planned for the period from 2027 to 2030. Against that, data-centre owners are projected to generate about 8.5 trillion dollars in cumulative revenue through 2030.
The gap sits in the layer above the hardware. Gartner puts AI application revenue at 3.6 trillion dollars over the same period, which leaves a shortfall approaching 5 trillion dollars when the infrastructure bill is counted in full. On those figures, more than 4 trillion dollars of asset value is exposed to write-downs by 2032. A central assumption in the models is that investors will accept a 12 percent return on invested capital, and Gartner's view is that this return is not expected to materialise before 2032 at the earliest. Assets valued on that assumption, meaning data centres, accelerators and long-term capacity commitments, would then sit well above their real worth.
The report also predicts structural consolidation. Gartner expects that by 2030 more than 80 percent of neocloud providers, the specialised cloud companies that lease compute capacity, will no longer operate independently, having been absorbed through acquisitions, mergers or partnerships with larger players. The reasoning is that scale economics favour hyperscalers, and that operators who cannot secure long-term customer commitments will struggle to justify continued capital expenditure.
The scope of the claim deserves a note. Gartner is describing a projection built on assumptions about revenue growth and returns, not a measured outcome. The forecast has been covered widely and has also drawn scepticism from parts of the industry, which argue that application revenue estimates understate how quickly AI workloads are being adopted. Projections of this kind are useful for planning rather than settled fact.
IMPACT: For compliance and risk functions the practical relevance is in vendor due diligence and financial reporting. Organisations signing multi-year compute commitments, taking capacity from neocloud providers or booking AI infrastructure as an asset on the balance sheet should test those assumptions against a scenario where the projected 12 percent return does not arrive before 2032. For procurement teams the prediction that more than 80 percent of neocloud providers consolidate by 2030 argues for contractual attention to change of control, exit terms and data portability when signing with smaller providers. For finance teams the exposure named here, more than 4 trillion dollars of asset value at risk of write-down by 2032, is the kind of figure that auditors and regulators will expect to see reflected in impairment testing and in the risk factors disclosed to investors. Anyone whose strategy assumes infrastructure costs will be recovered from application revenue before the end of the decade should now be able to say which evidence supports that assumption and what the fallback is if it fails to appear on schedule.
Sources: techzine.eu — gartner.com
Continue the conversation: https://t.me/AIAutomationArchitecture · Bluesky
Top comments (0)