Amazon has put $500 million into the nuclear power startup X-energy and secured 5 gigawatts of nuclear power through 2039. The company is now going public with an IPO target of up to $814 million. Anyone who thinks this is a climate project doesn't understand the business model behind it.
This is infrastructure securing. For AI, specifically.
#Why AI Needs Nuclear Power
AI data centers have a problem that solar and wind can't solve: they need power around the clock, without interruption, in constant amounts. That's called baseload power. The sun sleeps at night. Wind is fickle. Nuclear power just runs.
This isn't opinion, this is physics.
A large AI data center consumes between 100 and 500 megawatts depending on the expansion stage. Continuously. According to the International Energy Agency (IEA), global electricity consumption from data centers will rise to over 1,000 terawatt-hours per year by 2026. That's roughly equivalent to Japan's entire electricity consumption.
Amazon knows this. That's why Amazon is buying capacity before the market does.
#This Isn't an Isolated Case
Amazon isn't alone. Microsoft has struck a deal with Constellation Energy to bring the Three Mile Island nuclear power plant back online. Google has signed contracts with Kairos Power for small modular reactors. Meta is exploring similar options.
The pattern is the same everywhere: tech giants are securing energy capacity for the next 10 to 20 years. Not because they suddenly want to become energy providers, but because energy is the limiting factor for AI growth.
Whoever has the energy can scale. Whoever has to rent it pays the price others set.
#X-energy and What's Behind It
X-energy builds so-called high-temperature gas reactors. The reactor uses helium as coolant and a fuel called TRISO, where uranium is encased in ceramic and carbon spheres. This is considered safer than older reactor designs, but isn't yet widely deployed.
The company has raised around $1.8 billion in investments so far. The IPO attempt via a SPAC deal failed in 2023. Now the second attempt, this time a regular public offering.
Whether X-energy can deliver what it promises is an open question. Small modular reactors are technologically promising, but none of them are operating at commercial scale yet. 2039 as a target date is far away. A lot can happen.
But that's not Amazon's primary concern. It's about owning an option. A capacity commitment that others don't have.
#The Gap That's Emerging Right Now
Here's the part that concerns me.
The big tech giants are currently securing infrastructure for the next 15 years. Power, data centers, network capacity. Whoever owns this infrastructure controls the price long-term for everyone who uses it.
For SMBs, little changes in the short term. AWS, Azure, and Google Cloud keep running. AI tools remain accessible. Prices are still affordable today because the giants want to build market share.
But the dependence that's emerging right now is real. Anyone who bases their entire AI infrastructure on one provider gives up pricing control. Not immediately. But eventually.
This isn't alarmism. This is simply how infrastructure markets work. Railroads, telecommunications, cloud storage. Always the same pattern: cheap at first to create dependence. Then more expensive once the dependence is there.
#What You Can Take Away From This Concretely
You're not an energy provider and you won't be building a nuclear reactor. But there are a few things you can do right now:
Understand where your AI dependencies lie. Which tools do you use daily? Which provider is behind them? If a provider doubles prices, what happens to your workflow?
Diversify where it makes sense. Not everything on one cloud provider, not everything on one AI tool. That costs a bit more effort today, but gives you negotiating power tomorrow.
Watch pricing developments. AI APIs are cheap today. OpenAI, Anthropic, Google. That will change when infrastructure costs rise and investors want to see returns. Those who factor this in plan more realistically.
And above all: understand the dependencies before you enter them. Not after.
The decisions Amazon, Microsoft, and Google are making right now will affect the prices you pay for AI services in 5 to 10 years. This isn't speculation. This is the logic behind every infrastructure investment.
Those who understand this now can prepare. Those who wait until the bills come pay double.
