Half of 2026 US Data Centers Face Cancellation as Grid Bottlenecks Bite

Why it matters

The AI infrastructure thesis — that capital commitments guarantee capacity — is breaking down in real time. Sightline Climate's data shows 11GW of 2026 US data center capacity sitting in the "announced" stage with zero construction activity, despite build times of 12 to 18 months. That gap is not a planning problem. It is a structural failure.

The big picture

Alphabet, Amazon, Meta, and Microsoft have collectively committed more than $650 billion in 2026 capex. JPMorgan estimates the full AI infrastructure cycle requires $5 trillion, with the US government on the hook for over $1 trillion of the gap. The money exists. The transformers, the grid interconnections, the permits, and the domestic manufacturing base do not — and no amount of capex can conjure them on the timelines hyperscalers are demanding.

Key details

• Sightline Climate's 2026 Data Center Outlook covers 140 projects totaling ~16GW of planned US capacity; only 5GW is currently under construction
• 30-50% of 2026 projects face delay or cancellation; 11GW remains in the announced stage with no construction signs
• 2027 is worse: 6.3GW under construction against 21.5GW announced — a 70% execution gap
• Beyond 2028: 37GW of planned infrastructure has no firm completion date; only 4.5GW of that has broken ground
• Transformer delivery times have stretched from 24-30 months pre-2020 to as much as five years today

What they said

"We expect 30-50% of 2026 projects to be delayed, driven by power constraints, increasingly effective community opposition, and potential grid equipment shortages. 11GW of 2026 capacity remains in the announced stage with no signs of construction, despite typical build times of 12 to 18 months." — Sightline Climate, 2026 Data Center Outlook

"Not only are the energy constraints mounting, but so are the sociopolitical ones. Something's got to give." — George Gianarikas, Canaccord Genuity

---

Thesis: The Execution Gap Is Not Priced

The market has spent two years pricing the AI infrastructure buildout as a demand story. Hyperscaler capex commitments, GPU order books, and colocation lease signings drove the narrative. What the market has not priced is the supply-side constraint — specifically, that the physical infrastructure required to convert announced gigawatts into operational compute is bottlenecked at every layer simultaneously.

This is not a single-point failure. It is a systemic one. Grid interconnection queues run years long. High-voltage transformer lead times have extended to five years in some cases. Permitting processes face legislative interference — Maine's House just passed an 82-62 vote imposing a data center moratorium through 2027. Community opposition is increasingly organized and effective. And the domestic manufacturing base for critical electrical equipment — transformers, switchgear, batteries — cannot absorb the demand spike regardless of what tariff policy does to import costs.

The ZeroHedge synthesis of the Sightline data is directionally correct. But it undersells the compounding nature of the problem.

Evidence: Every Layer of the Stack Is Constrained

Start with the transformer bottleneck, because it is the most concrete and the most underappreciated.

Philippe Piron, CEO of GE Vernova's electrification division, states the problem plainly: before 2020, high-power transformers arrived 24 to 30 months after order. That was "totally manageable in the old world." AI companies now demand delivery in under 18 months. The market has responded by extending lead times to five years — the exact inverse of what the industry needs.

::chart[GE]

GE Vernova sits at $306.25, down 0.68% on the session. The stock has been cited in recent AcidCapitalist coverage across multiple macro themes — Hormuz, private credit, oil — reflecting its position as a cross-sector infrastructure bellwether. But the transformer constraint story is the one that matters most for the AI trade. GE Vernova is one of the few domestic manufacturers with the capability to produce the large power transformers data centers require. The problem is capacity: US manufacturing cannot scale fast enough, and the tariff environment makes Chinese imports politically toxic even as they remain economically necessary. That is the Catch-22 ZeroHedge identifies correctly — the US needs components from the country it is racing against.

The grid interconnection problem compounds the transformer shortage. Grid-connected projects represent 40% of 2026 capacity by count but only a portion of the gigawatt total. The gigascale projects — New Era Energy & Digital's 7GW campus in Lea County, Homer City's 4.5GW coal-to-gas redevelopment in Pennsylvania, Crusoe's 1.8GW natural gas and renewables project in Cheyenne — are pursuing off-grid or hybrid strategies precisely because waiting for grid interconnection approval could take a decade. These projects have the capital to fund their own generation. Most announced projects do not.

The 25% of 2026 projects that have not disclosed their power strategy is the

---

This article was inspired by Half Of US Data Centers Scheduled To Start In 2026, Will Be Canceled Or Delayed from Zerohedge. AC's analysis adds original research, data context, and editorial perspective.