Will more than 5 announced AI data center projects of >100MW capacity be cancelled, indefinitely delayed, or materially descoped during 2026?

Current economic conditions show rising interest rates and inflation, which have historically led to increased project costs and reduced investments in large-scale infrastructure like data centers. Additionally, recent reports from major tech firms indicate a trend towards more cautious capital expenditure in the face of uncertain market conditions and a focus on optimizing existing capabilities. As of late 2023, around 15% of announced data center projects greater than 100MW have been either delayed or cancelled in recent years, suggesting a strong possibility for continued scrutiny on such massive investments in 2026.

While AI demand continues to drive significant investment, the sheer scale and cost of >100MW projects, coupled with escalating construction challenges, permitting hurdles, and potential shifts in AI hardware adoption or efficiency, create a material risk of cancellations or delays. Historical precedents of large-scale infrastructure projects facing unforeseen obstacles suggest a significant, though not dominant, probability of such an outcome by 2026.

With 35-40 announced major AI data center projects, a base cancellation/delay rate of 20% would yield 7-8 projects. However, the 2026

Power availability remains the binding constraint, with ERCOT reporting 89 GW of queued data center load versus 12 GW of firm transmission commitments as of Q3 2025, while hyperscale capex guidance from Microsoft, Google, and Meta still shows 2026-2027 growth of 35-45% YoY; simultaneously, 2025 has already seen three >200 MW projects (two in Virginia, one in Ohio) formally descoped by 30-50% due to 18-24 month interconnection delays, establishing a precedent rate of roughly 15-20% of announced projects facing material changes.