AI Isn’t Running Out of Capital — It’s Running Out of Power

Apr 17, 2026 - 16:00

Brief Analysis

As of April 2026, approximately half of all planned US data centre builds this year are projected to be delayed or cancelled outright — not because of a shortage of capital or demand, but because the electrical grid cannot support them. Alphabet, Amazon, Meta and Microsoft are expected to spend more than $650 billion in 2026 to expand AI capacity Tom’s Hardware, yet the infrastructure required to power that ambition simply does not exist at the pace the industry requires. The bottleneck is transformers, switchgear and grid connections — components with lead times stretching to five years in a sector where deployment cycles run under 18 months. The AI race has collided with a power grid built for a different era, and the consequences are now structural rather than temporary.

EBM Exclusive Take

The data centre power crisis reframes the entire AI investment narrative in ways that European policymakers and institutional investors have not yet fully absorbed. The United States — with the world’s most advanced AI ecosystem and the deepest capital markets — cannot build fast enough to keep pace with its own ambitions. Europe, which operates under stricter energy efficiency regulations and land-use restrictions, faces an even more acute version of the same constraint. The UK’s £500 million Sovereign AI fund and OpenAI’s withdrawal from Stargate UK — citing energy costs — are symptoms of the same underlying reality: securing power infrastructure is now a greater competitive moat than algorithmic innovation. The companies and nations that solve the energy problem first will define the AI era. Those that treat it as a secondary concern are already falling behind.

The Scale of the Blockage

The numbers are stark. Of around 140 large-scale data centre projects representing approximately 12 gigawatts of power planned to go live in the US in 2026, only a third are under construction. The remaining two-thirds are stuck in a pre-production phase and are considered highly unlikely to open on schedule, if at all. UNILAD Tech

The problem compounds further into 2027. For data centres planned to open that year, only 6.3 gigawatts of computing infrastructure are currently under construction against 21.5 gigawatts announced — a gap that represents hundreds of billions of dollars of AI capacity that exists only on paper.

The Transformer Problem

The specific chokepoint is electrical components — transformers, switchgear and batteries that are essential to every data centre build. Delivery of high-power transformers took 24 to 30 months before 2020. Waiting periods can now stretch to as long as five years. Tom’s Hardware For data centres with deployment cycles under 18 months, this is a structural impossibility. A single delayed component stops an entire project.

To address shortages, US operators have turned to suppliers in Canada, Mexico, South Korea and China — adding months to already stretched timelines. Imports of high-power transformers from China surged from fewer than 1,500 units in 2022 to more than 8,000 in 2025, creating a new dependency that sits uncomfortably alongside broader US-China technology decoupling.

The Grid Cannot Keep Up

The deeper problem is generational. The US power grid was not designed for the load that AI requires. Data centres can be built in less than three years, but power generation takes far longer — three to six years for solar or wind, around six years for a gas turbine, and more than ten years for nuclear. Network World That mismatch was manageable when data centres were smaller. At the scale AI now demands — individual facilities measured in hundreds of megawatts — it has become a hard constraint.

The IEA projects global electricity consumption for data centres will double to around 945 TWh by 2030, growing at approximately 15% per year — more than four times faster than electricity consumption growth across all other sectors combined. IEA

The European Dimension

Europe is not insulated. In West London, the allocation of power to a new cluster of data centres has created an electricity shortage that has delayed housing and commercial projects by as much as ten years. Corporate Knights Stricter EU energy efficiency regulations and land-use restrictions add further barriers to data centre development that US operators do not face.

The paradox for European AI strategy is acute — the continent is committed to AI sovereignty and competitiveness, yet the regulatory and infrastructure environment makes building the physical foundation of that ambition significantly harder than it is in the US, and considerably harder than in Southeast Asia, where data centre capacity is expanding rapidly without equivalent constraints.

What Comes Next

Operators are responding by moving to second-tier locations with available power, pursuing on-site gas generation to bypass grid connections, and exploring nuclear — both small modular reactors and existing plant restarts. Microsoft has already announced plans to reopen Three Mile Island. The trade-offs between energy security, carbon commitments and AI ambition are becoming impossible to defer.

Related Analysis

The post AI Isn’t Running Out of Capital — It’s Running Out of Power appeared first on European Business Magazine.