Two analyst reports surfaced in the last week of March pointing at the same phenomenon from different angles: AI infrastructure investment is not approaching a natural ceiling. It’s accelerating.
The first is a BloombergNEF analysis published around March 24, covered in secondary reporting. According to that analysis, as cited in secondary coverage, AI data center capital expenditure from the 14 largest operators is on a trajectory toward approximately $750 billion in 2026, up substantially from prior-year levels. A key signal from the same report: analyst estimates for 2027 data center spending were revised upward by approximately 56%, suggesting the market doesn’t view 2026 as a peak. BloombergNEF also reportedly flagged more than 23 gigawatts of data center capacity under construction at the time of publication.
An important note on those figures: the BloombergNEF report URL is not accessible for direct verification at time of publication, and no independent cross-reference confirmed the specific numbers. These figures come from secondary coverage of the report, not from the primary source. They are directionally consistent with broader market intelligence, including the Oracle restructuring covered elsewhere in this cycle, but they should be treated as reported estimates, not confirmed data points.
The second data point comes from secondary coverage of a Gartner spending forecast. Per articles summarizing the forecast, the primary Gartner report is behind a paywall and wasn’t directly accessible, Gartner reportedly projects worldwide AI spending to reach approximately $2.52 trillion in 2026, representing around 44% year-over-year growth. Infrastructure is reportedly the dominant spending category within that total, at approximately $1.37 trillion. The same secondary coverage references a Gartner “Trough of Disillusionment” characterization, though that specific claim came from a T5 source and isn’t independently confirmed here.
The same qualification applies to the Gartner figures as to BloombergNEF’s: these are secondary-sourced projections. Gartner’s research is well-regarded in enterprise technology analysis, and the figures have circulated widely in AI market coverage. But “widely reported” and “independently verified” are different things. The numbers are reported here as widely-cited projections from a reputable source, not as confirmed facts.
What the two reports taken together suggest, even with those qualifications, is that the AI infrastructure investment cycle is in an expansion phase, not a consolidation phase. That’s consistent with the other major stories in this cycle: OpenAI closing a $122 billion funding round, and Oracle undertaking a multi-billion-dollar restructuring that analysts connect to data center capex. The money is moving in one direction.
Watch for: direct access to the BloombergNEF report (a link to the public summary page would allow verification of the specific figures); any Gartner press release or public summary that independently confirms the $2.52 trillion forecast figure; and Q1 2026 earnings calls from major hyperscalers, which will provide confirmed capex figures that can benchmark these projections against actual spending commitments.