Chinese AI startup DeepSeek has stunned the tech world with the release of its R1 model, which reportedly performs nearly as well as leading models from Google and OpenAI, despite being trained on a relatively modest number of GPUs. This breakthrough has significant implications for the energy demands of data centers, which have been predicted to consume as much as 12% of all electricity in the U.S. by 2027.

DeepSeek claims to have trained its model using 2,048 Nvidia H800 GPUs for two months, a fraction of the compute power rumored to be used by OpenAI. This raises questions about whether AI really needs the massive hardware outlays that have been predicted, and whether the surge in power demand from AI has been overblown.

The news has sent shockwaves through the tech industry, with Nvidia's share price plummeting 16% at the time of publishing. Startups and power producers that have bet big on new nuclear and natural gas capacity are also likely to be affected, as the demand for their services may be lower than anticipated.

Nuclear power, in particular, has been on the cusp of a renaissance, driven by advances in fuel and reactor designs that promise to make new power plants safer and cheaper to build and operate. However, nuclear power is still relatively expensive compared to wind, solar, and natural gas, and next-generation nuclear has yet to be tested at commercial scale.

Despite this, tech companies have been racing to secure new supplies of power, with Google pledging to buy 500 megawatts of capacity from nuclear startup Kairos, Amazon leading a $500 million investment in another nuclear startup, X-Energy, and Microsoft working with Constellation Energy on a $1.6 billion renovation of a reactor at Three Mile Island.

However, if DeepSeek's breakthrough is genuine, it may render these investments unnecessary. As one analyst noted, "While DeepSeek's achievement could be groundbreaking, we question the notion that its feats were done without the use of advanced GPUs." Still, history suggests that even if DeepSeek is hiding something, someone else will probably find a way to make AI cheaper and more efficient.

The current wave of new reactors aren't scheduled to come online until 2030, and new natural gas power plants won't be available until the end of the decade at the soonest. In that context, tech companies' power investments appear to be hedges in case their software bets don't pan out.

If AI's power needs do indeed ebb, expect tech companies to scale back their power ambitions. When given the choice between spending billions on physical assets or software, tech companies almost always choose the latter. This could leave nuclear startups and energy companies struggling to adapt to a changing market.

However, it's worth noting that the world is electrifying, and demand for electricity was expected to grow even before the AI boom. Some nuclear startups and energy companies may be able to produce power at a low enough cost that it won't matter if AI's power needs ebb. Wind, solar, and batteries are cheap and getting cheaper, and they're inherently modular and mass-produced, making them more adaptable to changing demand.

In the end, the safer bets in energy will probably flow to proven technologies that can be rapidly deployed and scaled according to a rapidly evolving market. Today, renewables fit that bill. As the AI landscape continues to shift, one thing is clear: the future of energy production and consumption is likely to be shaped by the intersection of technology and innovation.