Google DeepMind CEO: AI Will Save More Energy Than It Uses

Demis Hassabis, CEO of Google DeepMind, told ABC News that artificial intelligence will save more energy over the next decade than it consumes. The remark comes as the AI industry faces growing criticism over data centers' appetite for electricity, which has been driving up energy bills in many countries in recent months.

Hassabis's forecast comes as critics of artificial intelligence point to data centers' rising electricity consumption as one of the industry's biggest problems. The Google DeepMind chief acknowledged that tech companies must take responsibility for this cost themselves, rather than passing it on to ordinary energy consumers.

Return on investment argument

In the interview with ABC News, Hassabis said outright that "AI, over the next 10 years, will save significantly more energy than it uses." The argument rests on the premise that AI models will speed up scientific discoveries in areas such as new materials, batteries and power grid management systems, which over the long run should more than offset the electricity data centers currently use to train and run these models.

AI, over the next 10 years, will save significantly more energy than it uses. - Demis Hassabis, CEO of Google DeepMind

Hassabis also noted that large tech companies, including Meta, Microsoft and Google, are building their own power plants and energy infrastructure to secure electricity supply for AI data centers without passing the cost on to ordinary energy consumers. This is a nod to criticism that the buildout of AI computing capacity in parts of the US and Europe is driving up electricity prices for households.

The scale of change

Beyond energy, Hassabis spoke about AI's broader impact on the economy and science. He said the transformation triggered by artificial intelligence would have ten times the impact of the industrial revolution and could unfold up to ten times faster. He stressed that for years he has viewed AI as the "ultimate tool for accelerating scientific discovery," and that this vision is what drives his work at DeepMind.

The DeepMind chief did not shy away from tougher questions either. He acknowledged that the development of artificial intelligence "will disrupt the labor market" while also creating new career opportunities. He called for civil society, academia and the social sciences to work together to chart the direction humanity should take as the technology advances.

Who decides AI's direction

Asked about concerns that the future of artificial intelligence is being decided by a narrow group of companies and individuals, Hassabis said he favored "an approach based on international standards" rather than concentrating power over the technology in the hands of a few players. His position fits into a months-long debate about global oversight of the most powerful AI models, playing out at forums including the G20 and the European Commission.

The interview also touched on figures illustrating the scale at which Google is growing its AI ecosystem. Gemini, the company's flagship model, reached 900 million monthly active users, more than double the 400 million users it had a year earlier. That growth in the user base translates directly into demand for the computing power and electricity needed to handle queries.

Why it matters for Poland

Hassabis's statements carry significance beyond the tech industry itself, since energy consumption by AI data centers is a topic increasingly raised in Poland as well, where grid operators have warned about data centers' growing share of national electricity demand. If Hassabis's promise of energy savings from AI holds true, it could partly ease the pressure on energy prices felt by both businesses and households.

For now, though, the claim remains a forecast without a specific timeline or calculation methodology, and critics of the AI industry point out that current electricity consumption by data centers is a measurable fact today, while future savings from AI-assisted scientific discoveries remain, for now, a promise. Whether artificial intelligence's energy balance will actually turn positive within a decade will only become clear over time, as companies publish data on their data centers' energy consumption.