OpenAI Just Told the White House That Electricity Is the New Oil

OpenAI didn't send the White House a polite memo. They sent a formal pitch with a thesis so blunt it belongs on a protest sign: electricity is the new oil. Not metaphorically. Not eventually. Right now. And if America wants to stay competitive in AI, it needs to solve for power the way it once solved for petroleum—aggressively, strategically, and at national scale.

The stakes? OpenAI's Stargate data centers alone will demand 10 gigawatts of power—enough to run eight million homes. And that's just one project. The company is pushing for 100 gigawatts per year nationally to support next-generation AI workloads. For context, the entire U.S. currently generates about 1,200 GW of total power capacity. OpenAI just asked for nearly 10 percent of that. Annually. Just for AI.

If you thought the AI race was about algorithms and talent, surprise—it's about electrical grids and generator contracts.

When Data Centers Start Competing With Cities

Let's put 10 gigawatts in perspective. That's roughly the output of ten nuclear power plants. It's more electricity than Ireland uses in a year. And OpenAI needs it for one infrastructure project. Not their entire operation—one. The Stargate project is just the beginning. To train and run the next generation of models—the ones that might actually deliver on AGI promises—the U.S. needs to add 100 GW of new power capacity every year.

That's not an infrastructure project. That's a Marshall Plan for electricity.

The pitch to the White House isn't subtle: if America doesn't solve for power, China will. And whoever controls the energy supply controls the AI future. It's the same logic that drove 20th-century geopolitics around oil, except this time the resource isn't buried in the desert—it's generated by turbines, solar farms, and nuclear reactors. And right now, the U.S. grid isn't built for this.

The AI Arms Race Is an Energy Arms Race

Here's the uncomfortable truth: AI progress is now bottlenecked by watts, not weights. You can have the smartest researchers, the best algorithms, and the cleanest code, but if you can't power the data centers to train and run those models, you lose. OpenAI knows this. Microsoft knows this. Google, Meta, Amazon—they all know this. That's why they're all scrambling to lock down power contracts, build private energy infrastructure, and lobby for regulatory fast-tracking on everything from natural gas plants to small modular reactors.

The companies winning the AI race aren't just the ones with the best models. They're the ones with guaranteed access to gigawatts. And right now, that's a zero-sum game. Every watt OpenAI secures is a watt someone else doesn't get. Every data center location is chosen not just for connectivity or tax incentives, but for proximity to power plants.

This is why OpenAI's pitch matters. They're not asking the government for funding or subsidies. They're asking for energy policy to be restructured around AI as a strategic priority. Build more plants. Speed up permitting. Treat electricity generation the way we once treated oil refining—as critical infrastructure for national competitiveness.

The Climate Contradiction No One Wants to Address

Let's talk about the elephant in the coal-fired room: this is an environmental disaster in the making. The U.S. was supposedly transitioning to renewable energy, decarbonizing the grid, hitting net-zero targets. And now OpenAI is asking for 100 GW per year—much of which will, realistically, come from natural gas and nuclear because renewables can't scale fast enough to meet AI's voracious, always-on power demands.

AI companies love to talk about using AI to solve climate change—optimizing supply chains, predicting weather patterns, designing better batteries. But they conveniently skip over the part where training and running AI models produces carbon emissions equivalent to small countries. GPT-4's training run reportedly used as much electricity as 1,000 U.S. homes consume in a year. GPT-5 will be worse. The "legitimate AI researcher" OpenAI wants to build by 2028? It'll need its own power plant.

This isn't to say AI isn't worth it. But let's stop pretending there's no trade-off. Every gigawatt powering Stargate is a gigawatt not powering homes, hospitals, or electric vehicle charging stations. Every dollar spent fast-tracking nuclear permits for AI is a dollar not spent on residential solar or grid modernization. We're making choices here. Let's at least be honest about what we're choosing.

What This Means for Everyone Not Building AGI

If you're a marketer, a business leader, or just someone trying to understand where this is all headed, here's your signal: AI isn't just expensive in money—it's expensive in energy. And energy costs don't scale down. They scale up. Which means the AI tools you're using will get more powerful, but also more expensive to run. The free tiers will shrink. The API costs will climb. And the companies that can't secure cheap, reliable power will lose.

For enterprises, this is your reminder that cloud costs are about to become energy costs. If your AI strategy depends on unlimited compute at predictable prices, rethink it. Power scarcity is coming. And when it does, the companies with long-term energy contracts will have a structural advantage over everyone else.

For everyone else? Watch the energy policy fights. Because whoever controls the grid controls the future of AI. And right now, OpenAI just told the White House that future depends on treating electricity like we once treated oil—as the resource wars are fought over.

Need to build an AI strategy that accounts for the real costs—not just the hype? Let's talk. Because waiting for someone else to solve the power problem isn't a strategy.