The tech industry’s narrative was straightforward for roughly ten years: chips, chips, and more chips. The winner was the person who purchased the most GPUs. Quietly, that story has reached its end. The new bottleneck is electricity, which is much more difficult to produce on demand and far less glamorous than silicon. dependable, unfailing, and unaffected by clouds passing electricity. Baseload, as the industry refers to it, is currently insufficient to supply what is being constructed in Texas, Virginia, and the American West’s high deserts.
You can see the scope of the issue before anyone has to explain it if you stroll past one of the new hyperscale campuses rising outside of Phoenix. The buildings, which are quiet, fenced, and humming in a low register that you can feel more than hear, stretch across what was once farmland. A small city’s worth of power is consumed by each one. Additionally, the companies that are responsible for them—Meta, Microsoft, and Amazon—have made public commitments to clean energy for the better part of 20 years. Wind contracts, solar farms, the entire green portfolio. It was largely successful until the AI race began consuming power at a rate that no one had anticipated.
| Category | Details |
|---|---|
| Sector Focus | Nuclear Energy & Artificial Intelligence Infrastructure |
| Primary Players | Meta Platforms, Microsoft, Amazon, Oklo Inc. |
| Key Catalyst | AI data center baseload power demand |
| Notable Stock | Uranium Energy Corp. (NYSEAMERICAN: UEC) |
| Options Activity Date | January 9, 2026 |
| Call Volume Spike | 35,884 contracts (35% above daily average) |
| Reactor Type Trending | Small Modular Reactors (SMRs) |
| Recent U.S. Plant | Vogtle Unit 4, Georgia (14+ year build) |
| Last New Reactor Before Vogtle | Watts Bar Unit 2, Tennessee (2016) |
| Policy Voice | White House Office of Science and Technology Policy |
| Market Signal | Widening uranium supply deficit |
| Investment Lanes | Fuel supply + new infrastructure construction |
This load cannot be met solely by renewable energy. The wind goes out. The sun goes down. Batteries are helpful, but they continue to be stubbornly expensive at gigawatt scale, and a data center training a frontier model cannot pause in any meaningful way. Therefore, the topic of discussion in Silicon Valley has changed to nuclear power, which would have seemed a little ridiculous in those rooms only a few years ago. Not as a mental exercise. as a purchase order.
If you know where to look, you can find the signals everywhere. With about 35,884 call options in a single session on January 9, traders moved on Uranium Energy Corp. with a level of urgency that doesn’t typically accompany drowsy resource stocks. This is significantly more than the daily average. That’s not noise from a store. That is a thesis supported by institutional funding. The general argument is that the spot market is where the leverage resides and that uranium is about to become important once more.
Those who keep a close eye on this believe that the partnerships that have been announced in recent months are merely the outward manifestation of something bigger. In order to secure future capacity from unbuilt plants, tech giants are entering into agreements with advanced reactor developers. The market has interpreted Oklo’s agreement with Meta as a sort of de-risking event, demonstrating that the small modular reactor pitch is no longer merely a slide deck. It’s still genuinely unclear if that confidence is justified.
The irony in all of this is difficult to ignore. The same industry that used wind farms and solar arrays to establish its environmental credibility is now aiming for technology that, in the eyes of the general public, has never fully shed its reputation from 1979. “This is stuff we figured out in the 1940s and 1950s, before the personal computer was invented,” Last Energy’s Michael Crabb recently told a reporter. “We’ve gotten a lot better at it.” Perhaps. Watts Bar 2 was the only new reactor built in the United States since 1996, and Vogtle 3 and 4 took more than fourteen years each. To put it mildly, the track record of actually building reactors in America is not encouraging.
Therefore, the wager on small modular reactors is based in part on speed, in part on loosening regulations, and in part on the idea that the demand curve from AI is just too steep to ignore. It appears that investors think the math is sound. The question that no one yet knows the answer to is whether engineering, politics, and public opinion all work together on the same timeline. As you watch this develop, it seems unlikely that the next stage of the AI boom will be decided in a server room. A substation will make the decision.