Emergency Demand Cuts Followed a 160-Gigawatt Peak. The AI Buildout Is Testing the Grid.

The PJM Interconnection, which operates the largest U.S. power grid across all or parts of 13 states and the District of Columbia, crossed 160 gigawatts of electricity demand on July 1. Extreme heat, air conditioning, industrial activity, and data centers pulled on the system at the same time. Demand rose again on July 2, pushing the grid operator into emergency procedures. The tension is no longer only how much companies plan to spend on AI. It is whether the power system can support those plans when demand is already near its limits.

The key observation is that electricity is becoming a real constraint on the pace, location, and cost of AI growth.

PJM Interconnection reported a preliminary hourly peak of 161,910 megawatts between 5 p.m. and 6 p.m. on July 1, 2026.

On July 2, instantaneous demand reached about 162,700 megawatts. PJM also activated roughly 6,000 megawatts of demand response, reducing consumption during the peak.

Reuters reported that wholesale electricity prices exceeded $2,500 per megawatt-hour in parts of the Mid-Atlantic and Dominion zones, compared with roughly $40 under more normal conditions.

PJM’s 2026 load forecast projects summer peak demand growth of 3.6% per year over the next decade. The forecast puts peak demand at 222,106 megawatts by 2036.

This is a capacity test.

The issue is not simply whether enough power plants exist across the system. The test is whether dependable electricity can reach the right place at the right hour while generators, transmission lines, and substations are under pressure.

Extreme heat creates the immediate strain. Data centers and industrial facilities raise the system’s starting load, leaving less room when residential and commercial demand jumps. That combination can turn a familiar summer heat wave into a broader infrastructure test.

One key signal is the use of demand response. When a grid operator pays large customers to reduce consumption, the system has moved beyond ordinary price adjustment and into active load management.

Another signal is local congestion. A regional grid may have enough total generation while still struggling to move electricity into fast-growing areas. In data-center-heavy markets such as Northern Virginia, connection delays, transmission limits, and local prices may provide a cleaner read than the systemwide capacity figure.

A common misread is treating this as only a heat-wave story. Heat produced the immediate peak, but data-center and industrial growth increased the amount of demand already sitting on the system before temperatures climbed.

Another mistake is assuming that higher electricity demand is a clean benefit across the power sector. New demand can require costly generation, transmission upgrades, substations, and reserve capacity. Those costs may arrive years before new facilities produce steady revenue.

This won’t wait.

BloombergNEF just projected $185 trillion will move out of oil and into a single fuel source -

A fuel the International Energy Agency says holds 140 times the electricity the entire planet uses in a year.

Google has signed a 15-year deal for it.

Berkshire Hathaway wrote a check.

The Pentagon is backing it.

And on August 18th, the government is about to hand this energy source the biggest competitive advantage in modern history.

President Trump already cleared the path.

He killed subsidies for every other renewable.

Then reclassified this one alongside oil and nuclear.

Now an August 18th deadline is set to trigger a wave of capital I believe will send one small company soaring 1,000%… even 3,000%.

This is the same pattern I’ve been profiting from for 30 years.

A crisis creates a bottleneck...

The bottleneck creates a chokepoint...

The chokepoint creates a fortune.

That pattern led my readers to Palantir before it climbed up to 2,700%…

Allowed them to close 502% on Rocket Lab…

And see gains upwards of 686% on Meta.

Now I see the same setup - only bigger.

This could be the last great energy play of our lifetimes.

Focus on usable capacity, not announced investment.

AI spending plans may be measured in hundreds of billions of dollars, but servers cannot run on capital spending alone. They need dependable generation, transmission access, local substations, cooling systems, and firm connections to the grid.

When those pieces do not expand together, the bottleneck can move from chips and construction crews to electricity. AI investment may remain large while its timing, location, and cost become less certain.

Power availability is becoming part of the AI growth equation. One period of extreme heat does not settle the issue, but it shows where ambitious investment plans can meet physical limits.

Talk soon,
The Playbook Daily