DeepSeek's Budget Chatbot: Implications for AI Energy Use and Climate Impact

Chinese startup DeepSeek has taken the technology world by surprise with claims that it has developed a highly popular chatbot at a significantly lower cost than those produced by major American tech companies.

This assertion raises questions about the extensive financial resources U.S. firms are investing in expansive data centers, which they argue are essential for powering the next generation of artificial intelligence.

If DeepSeek's chatbot operates effectively on a fraction of the resources, could it suggest that the AI sector requires less electricity than previously believed? This potential reality carries important consequences for global efforts to combat climate change, as the data centers that support AI systems consume vast amounts of energy, much of which is sourced from fossil fuels, exacerbating environmental issues.

“There has been a very gung ho, go ahead at all costs mentality in this space, pushing toward investment in fossil fuels,” said Eric Gimon, senior fellow at Energy Innovation. “This is an opportunity to tap the brakes.”

Experts argue that improving the efficiency of AI technologies could lessen their environmental impact, even if their demand for electricity remains high.

DeepSeek’s chatbot has attracted attention and is now the top downloaded free app on Apple’s iPhone, surpassing established competitors like ChatGPT and Google’s Gemini. Jay Woods, chief global strategist at Freedom Capital Markets, commented, “It caused a bit of a panic. These were the hottest stocks in the world.”

DeepSeek’s AI assistant exhibits strong performance across multiple tasks, such as coding, solving complex math problems, and providing multi-step reasoning in its answers. Analysts have been examining DeepSeek’s research papers closely, noting that the training cost for its main AI model, R1, was only $5.6 million. This is remarkably low when compared to the billions spent developing products like ChatGPT.

However, it is important to mention that this figure reflects only the training phase of the chatbot and does not include expenses incurred during earlier research and development stages. DeepSeek operated under constraints posed by U.S. export controls on high-performance AI chips and relied on a less powerful chip from Nvidia, which remains available for sale in China.

Data centers in the U.S. currently account for approximately 4.4% of national electricity consumption, and this is expected to rise to between 6.7% to 12% by 2028, as reported by the Lawrence Berkeley National Laboratory.

It has been a common assumption that U.S. tech giants must invest significant resources into data center infrastructure to support AI systems. For instance, Meta Platforms, the parent of Facebook and Instagram, plans to allocate up to $65 billion this year for various projects, including a major data center complex in Louisiana. Similarly, Microsoft has earmarked $80 billion for its initiatives.

Experts believe that making AI more efficient may lead to higher overall usage, thereby increasing energy demand even with lower operational costs. Vic Shao, founder of DC Grid, stated, “When there’s an innovative technology that’s useful and affordable, people will use it.” This means that while data centers may become more efficient, they will still be constructed to meet growing demands.

If DeepSeek’s assertions prove accurate, certain AI tasks could potentially be performed without the need for a data center, shifting some operations to mobile devices. Rahul Sandil, vice president at MediaTek, noted that this shift could reduce the need for data center resources and allow more time to develop renewable energy options.

Some companies in the AI sector, like Bloom Energy, experienced stock price declines following the revelations about DeepSeek's chatbot. Founder and CEO KR Sridhar emphasized the importance of U.S. leadership in AI, particularly in incorporating clean energy solutions for data centers to mitigate reliance on fossil fuels.

Rick Villars of IDC highlighted that while DeepSeek's news may inspire a new wave of AI research and development, the need for electricity and data centers will persist. He stated, “We still think the capacity is required.”

AI, Climate, Energy