Why American Data Centres Are Unplugging From the Grid
Data Infrastructure / Infrastructure Management | 6 min read
The AI infrastructure boom is colliding head-on with the limits of the US power grid — and the collision is reshaping how America's most powerful technology companies source and manage electricity. As of March 2026, US data centres consume approximately 176 TWh of electricity annually — around 4.4% of the nation's total power — and the trajectory is steeply upward. With 550 planned data centre projects totalling 125 GW of capacity in the global pipeline, power availability — not chips, not land — has become the primary scaling constraint of the AI era.
The Grid Cannot Keep Up
AI workloads are redefining how data centres consume energy at a fundamental level. Traditional server racks operate at between 5kW and 10kW — yet AI-optimised racks now reach between 50kW and 100kW. This tenfold increase in density places acute pressure on power delivery, cooling systems, and the grid connections that serve them. Interconnection queues have become a defining bottleneck: construction can face multi-year delays when data centre operators attempt to connect large campuses to the grid. Analysis presented to PJM Interconnection governors warns of a 49 GW US generation shortfall by 2028 — equivalent to 49 large natural gas power plants.
"What we've seen over the last couple of years is that movement from chips being the primary scaling constraint to power generation being the primary scaling constraint."
— Peter Perri III, Managing Partner, Jupiter Island Capital
Going Off-Grid: The Island-Mode Alternative
Faced with congested interconnection queues and multi-year wait times, data centre operators are increasingly turning to off-grid, or "island-mode" systems — generating and managing their own power independently of the utility network. The appeal is straightforward: operators can bypass the queue entirely, securing power on their own timeline without waiting for grid capacity to materialise.
"We're seeing this increased demand for it because the grid-tied interconnect queue is congested and this alternative may provide a way to bypass it. We're going to see this demand for off-grid projects because the juice is now worth the squeeze."
— Maura Yates, CEO, Mothership Energy
In the near term, natural gas plays a central role in supporting off-grid data centres. Operators require dispatchable power to meet the "five nines" reliability standard — 99.999% uptime — that intermittent renewable sources cannot consistently deliver at scale without extensive storage. A 1GW gas plant may occupy around 40 acres, while equivalent renewable capacity requires far more land, often distant from the fibre networks data centres depend on. Oracle has become a prominent example of this approach, powering its Stargate data centre campuses with behind-the-meter natural gas, enabling new facilities to come online without grid connection.
The "Bring Your Own Generation" Ultimatum
For operators who do want grid access, conditions are increasingly onerous. Grid operators are offering hyperscalers priority connection — but on the condition that they agree to be disconnected or powered down when demand spikes threaten reliability. ERCOT has warned that losing roughly 2,600 megawatts of demand at once could threaten system stability — and the North American Electric Reliability Corporation (NERC) has identified sudden data centre disconnections as one of its most important emerging risks. The Data Center Coalition pushed back in a public statement, noting that a reliable, uninterrupted power supply is essential for critical data centre operations. Hyperscalers also argue that reliance on diesel generators to cover disconnection periods could violate air-quality restrictions in key markets.
The Cost Impact and the Regulatory Response
The consequences of the power crunch are already reaching household electricity bills. Retail electricity prices have risen 42% since 2019, outpacing the 29% increase in the Consumer Price Index. Goldman Sachs projects that data centre power consumption will boost core inflation by 0.1% in both 2026 and 2027. Capacity market prices in PJM Interconnection — serving 65 million people — have spiked nearly tenfold. By early 2025, around US$64 billion of US data centre projects faced delays or opposition linked to energy use, water consumption, and local community impact.
Policy responses are multiplying. In Texas, Senate Bill 6 introduces stricter requirements for large energy users above 75MW — including disclosure of on-site generation and participation in grid support during emergencies. The proposed federal Decentralized Access to Technology Alternatives Act of 2026 aims to exempt fully off-grid systems from certain regulatory oversight, provided they remain separate from the main grid. Virginia, Georgia, Indiana, and Washington have enacted or proposed legislation requiring data centre operators to fund infrastructure improvements proportional to their electricity consumption.
Key Takeaways
- • US data centres consume 176 TWh annually (4.4% of national power) and rising fast — AI racks now draw 50–100kW vs the 5–10kW of traditional server racks, creating a 10x energy density surge.
- • Multi-year interconnection queue delays are pushing operators toward off-grid "island-mode" systems powered by behind-the-meter natural gas — bypassing the grid entirely to secure faster deployment.
- • Grid operators are offering conditional access — "bring your own generation" or agree to disconnection during peak demand — while NERC identifies sudden data centre disconnections as an emerging reliability crisis.
- • Retail electricity prices have risen 42% since 2019; Goldman Sachs projects data centre demand will add 0.1% to core inflation annually; PJM capacity prices have spiked nearly tenfold.
- • Texas SB6, a proposed federal off-grid exemption act, and state-level infrastructure investment mandates across Virginia, Georgia, Indiana, and Washington are reshaping the regulatory landscape for data centre energy.