The AI Bubble May Pop, But Its Power Plants Will Fuel Our Future

If the AI boom proves to be a bubble, it may turn out to be one of the most productive bubbles in history. The reason is simple: power. Unlike past technology waves that required only modest infrastructure, AI’s demands are immense. Training and operating modern AI models require electricity on a scale previously reserved for entire cities or nations. This appetite has forced the world’s most powerful corporations to become energy developers, reviving nuclear plants, erecting massive solar fields, and planning gigawatts of new capacity. If demand for AI falters, the power infrastructure will remain, just as the fiber optic cables of the dot-com era still carry the internet today.

The parallel to the telecom and dot-com boom of the 1990s is instructive. Companies laid tens of millions of miles of fiber, so much that by 2002 only 2.7% was in use. Prices for bandwidth collapsed by 90%, and many firms went bankrupt. Yet the infrastructure became the backbone of modern broadband, enabling everything from Netflix streaming to remote work. What was once derided as waste proved to be an extraordinary surplus that society later tapped for growth. Overbuilding can be destructive in the short term but transformative in the long term.

So it may be with AI. Already there are nearly 500 AI unicorns, worth a combined $2.7 trillion. Major players such as Microsoft, Google, Amazon, Apple, and 𝕏-backed ventures are racing to secure power for their data centers. These centers are no longer measured in megawatts but gigawatts. A single AI campus can demand as much power as New York City. In Chicago alone, utilities have received requests for 40 gigawatts of data center power. Even if most projects never materialize, the scale of planning shows how AI has forced a new industrial revolution in electricity.

This shift has already created remarkable energy projects. Microsoft struck a deal to restart a dormant reactor at Three Mile Island, securing 837 megawatts of nuclear power for two decades. Amazon has purchased land adjacent to nuclear plants and invested in 20 gigawatts of renewable energy capacity. Google is building data centers co-located with solar and wind farms, ensuring that each new load is matched with new supply. Apple, too, has built its own solar fields and even bought a hydro plant to keep its data centers running on renewables. Oracle plans to power an AI campus with small modular nuclear reactors. These are not marketing gimmicks. They are billion-dollar investments in hard infrastructure that will endure whether AI’s valuations prove inflated or not.

Critics will object that if the bubble bursts, we will be left with stranded assets. That risk is real, but overstated. A nuclear reactor or solar farm does not vanish when a data center goes dark. Electricity can be rerouted to homes, factories, and electric vehicles. Indeed, if history is a guide, the greater risk is not stranded assets but underestimating how quickly demand will rise to meet them. Just as internet traffic eventually caught up with fiber capacity, electrification of transport, heating, and industry could easily absorb today’s AI-driven overbuild. Even a surplus of capacity could lower prices, making energy cheaper and more reliable for ordinary Americans.

There is another reason to welcome this surge: pollution. Much of the power now being built for AI is pollution-free, because corporations face investor pressure to meet climate pledges. Microsoft, Google, and Apple are leading a nuclear and renewable renaissance, financing infrastructure the government has failed to build. Even if AI fizzles, this legacy will remain. An overbuilt clean energy system could accelerate the electrification of cars, buses, and industry, cutting emissions in ways that would have seemed impossible a decade ago.

The grid itself will also be modernized. AI data centers require new substations, high-voltage transmission lines, and cutting-edge cooling technologies. Utilities are being pushed to upgrade faster than regulators ever could have mandated. These upgrades improve reliability for everyone, ensuring that blackouts become less frequent and that the system can handle new types of load, from electric trucks to advanced manufacturing. Once again, the analogy to the dot-com bubble is clear: speculative investment forces technology forward, and the wreckage becomes the scaffolding of future progress.

Of course, risks remain. But with most hyperscalers focused on clean energy, the likely outcome is an oversupply of renewables and nuclear. That is a problem worth having.

Bubbles are painful when they burst, but they often leave behind gifts. The railroad mania of the 19th century built the steel arteries of the nation. The fiber glut of the 1990s gave us cheap bandwidth for the digital economy. If the AI frenzy ends in disappointment, its monuments will be power plants and substations. That is an outcome conservatives and progressives alike should welcome, for it ensures that America will have the energy foundation needed for decades of growth.

If you enjoy my work, please consider subscribing: https://x.com/amuse.

Sponsored by the John Milton Freedom Foundation, a nonprofit dedicated to helping independent journalists overcome formidable challenges in today’s media landscape and bring crucial stories to you.

READ NEXT: Mom Uses Sacred Right To Save Her Baby, Stop The Threat