OpenAI’s CEO, Sam Altman, has praised Chinese startup DeepSeek for its groundbreaking artificial intelligence model, R1, calling it an “impressive model.”
In a social media post on Monday, Altman said, “It’s invigorating to have a new competitor,” adding that the model stands out for its performance and affordability.
DeepSeek’s R1 model has caused a stir, not just in the tech industry but also in the energy sector.
The company claims it developed the open-source model using only 2,000 Nvidia chips, much fewer than previously thought necessary to train advanced AI systems. This innovation could significantly reduce both the cost and energy demands of AI development.
The AI industry has long been associated with high energy consumption, particularly by data centers that power AI applications. According to the International Energy Agency (IEA), data centers accounted for 1% of global electricity use in 2023, with projections showing that consumption could double by next year.
DeepSeek’s model, however, offers a potential solution by improving computing efficiency. “R1 illustrates the threat that computing efficiency gains pose to power generators,” wrote Travis Miller, an energy strategist at Morningstar.
The energy sector felt the impact immediately, with investors dumping energy stocks on Monday. Constellation Energy, a major player planning to expand energy capacity for AI, saw its shares plummet by over 20%.
Andrew Lensen, a senior lecturer in artificial intelligence, noted that DeepSeek’s approach could lead to a decrease in overall energy requirements. “If DeepSeek was to replace models like OpenAI’s, there would be a net decrease in energy use,” he explained.
Despite the promise of energy savings, some experts warn of a potential rise in demand. This is tied to the Jevons paradox, where improved efficiency leads to increased usage. Microsoft CEO Satya Nadella acknowledged this concern, stating, “As AI gets more efficient and accessible, we will see its use skyrocket, turning it into a commodity we just can’t get enough of.”
Additionally, DeepSeek’s “chain-of-thought” model, while efficient, is more energy-intensive because it processes queries through multiple steps. Lensen believes US tech companies might use DeepSeek’s methods to build even larger, more powerful AI models instead of focusing solely on efficiency.
In recent years, major tech companies like Google, Microsoft, and Amazon have invested heavily in data centers, spending the equivalent of 0.5% of the US GDP in 2023 alone, according to the IEA. Many of these companies are also exploring renewable and nuclear energy to meet growing demands.
While DeepSeek’s innovation could reduce the need for massive energy investments, its long-term impact on both the tech and energy industries remains to be seen.
For now, the startup has sparked discussions about the future of AI efficiency and energy use, with Altman’s endorsement further solidifying its position as a game-changer in the AI race.