Data Center Power Demand Drives US Electricity Prices Up 76%

Electricity prices on the largest US power grid rose 76% in the first quarter of 2026. The cause is explosive demand from data centers that power AI services. The spike has prompted calls for grid operators to find a way to ease the pressure on consumers.

Pressure on the PJM grid has hit its limit

On the PJM Interconnection grid, which serves 65 million people in 13 states, demand from server facilities has spun out of control. A 76% rise in a single quarter is not market volatility but systematic pressure from new consumers that draw megawatts at a time. The grid operator has to balance states' demands to protect consumers against a different reality: data centers are built fast, pay for their electricity and are needed to keep the digital economy running.

AI data centers are eating megawatts like never before

Every service, from ChatGPT to cloud applications, requires enormous computing power. Transformer models train for months on specialized GPU clusters, consuming gigawatt-hours of electricity. And that is only training. Inference adds a constant background load.

Data centers are being deployed across the country at record speed, and their power consumption doubles every few years. The key requirement is one: be located close to internet hubs and cities.

• New data centers require 200–500 megawatts (about as much as a small city)
• Cooling servers in hot months adds 30–40% to consumption
• AI models need a constant power supply: GPUs cannot be switched off at peak hours
• Cloud services are displacing local computing, shifting load into data centers
• Every new AI startup needs its own server cluster

States demand a fix, companies look for a way out

Grid operators face a contradiction. States demand that they "ease the burden on consumers" and lower prices. But by law operators must serve anyone willing to pay. Residential and small commercial customers are already seeing the hit on their bills. Large companies and data centers pay the market price and remain attractive customers.

The power sector is not ready for the AI boom

This spike is a warning that America's energy infrastructure is not ready for the AI revolution. Data centers are multiplying faster than new power plants are being built. Solutions exist, but all are difficult: build new capacity (including nuclear), demand better efficiency from data centers, introduce consumption quotas, or move server facilities to regions with cheap energy.

What this means

The PJM story is a preview of what will start happening everywhere. Demand for electricity to power AI will only keep growing. Electricity will become one more bottleneck for the AI industry. Countries and companies that supply data centers with cheap, abundant energy will gain a competitive advantage.