Chinese AI startup DeepSeek surprised the world with the release of its R1 model, which looks very similar to the main models from Google and OpenAI, although the company says it uses a small amount of GPUs for training. DeepSeek’s relative efficiency has experts and investors questioning whether AI really needs the huge hardware costs that have been predicted. And it can change the demands of data centers – and the energy needed to power them. The company claims to have run 2,048 Nvidia H800 GPUs for two months to train the slightly older model, a fraction of OpenAI’s rumored computation. Few companies look like Nvidia, whose stock price fell 16% when it was published. Perhaps more vulnerable are startups and power producers that are betting big on new nuclear and natural gas capacity. Nuclear power, in particular, has been on the cusp of a renaissance for years, fueled by advances in fuel and reactor design that promise to make a new generation of power plants safer and cheaper to build and operate. So far, there is no reason to move forward. Nuclear is still expensive relative to wind, solar, and natural gas. In addition, next-generation nukes have yet to be tested on a commercial scale. The surge in power demand from AI is changing that equation. With data centers predicted to use up to 12% of all electricity in the US – more than three-quarters by 2023 – and AI data centers forecast to run low on power by 2027, tech companies are racing to secure new supplies, and spending billions. dollar in problem. Google has pledged to buy 500 megawatts of capacity from nuclear startup Kairos, Amazon is leading a $500 million investment in another nuclear startup, X-Energy, and Microsoft is working with Constellation Energy on a $1.6 billion renovation of its reactor at Three Mile Island. But what if the problem is overblown? There are no hard and fast rules suggesting that the only way to improve AI performance is to use more computing. For a while, the tactic worked well, but more recently, more computations have not yielded the same results. AI researchers have come up with a solution, and maybe DeepSeek found the R1 model. Not everyone is convinced, of course. “While DeepSeek’s achievements may be groundbreaking, we question whether these achievements were achieved without the use of advanced GPUs,” Citigroup analyst Atif Malik wrote. However, history shows that even if DeepSeek hides it, others will find ways to make AI cheaper and more efficient. After all, it is easier and potentially faster for the task of some PhD to develop a better model than to build a new power plant. The current wave of new reactors isn’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 earliest. In that context, the tech company’s investment of energy is seen as a hedge if the software bet doesn’t pan out. If that’s the case, expect tech companies to tone down their power ambitions. When given the choice between spending billions on physical assets or software, tech companies almost always choose the latter. Where will that leave nuclear startups and energy companies? It depends. Some may be able to generate power at a fairly low cost, so it doesn’t matter if the AI power needs to ebb. The world is electrifying, and even before the AI bubble begins to inflate, the demand for electricity is expected to grow. But absent demand from AI, these cost pressures are likely to increase. Wind, solar, and batteries are cheaper and cheaper, and they are modular and mass-produced. Developers can roll out new renewable plants in stages, delivering electricity (and revenue) before the entire project is completed while offering some control over the future in the face of uncertain demand. The same cannot be said about nuclear reactors or gas turbines. Tech companies know this, so they are quietly investing in renewable energy to power their data centers. Few people are predicting an AI boom today, and no one knows what it will be like five years from now. As a result, safer bets in energy are likely to flow to proven technologies that can be deployed quickly and scaled according to rapidly evolving markets. Today, renewable energy fits the bill.
How DeepSeek’s efficient AI can prevent nuclear resurgence
