KUALA LUMPUR (Oct 15): Big Tech’s hunger for electricity to run artificial intelligence (AI) could help pay for the clean energy the world needs, said Professor William Chueh, director of Stanford University’s Precourt Institute for Energy.

Hyperscalers are willing to pay a premium for power and have long-term climate commitments and that spending can be channelled into cleaner and more reliable sources of supply, he argued.

Chueh was speaking at AVPN Academy’s Ask the Expert: AI, Climate and Energy Connection online session, where he set out how AI is both straining power systems and offering ways to run them more efficiently.

Although AI contributes to rising energy demands, it also creates new opportunities to improve how energy systems operate. Chueh pointed out AI's ability to optimise electricity grids and improve weather forecasting.

“AI has now exceeded the prediction power of most weather models, making it possible to predict weather more accurately using less computational power,” said Chueh.

“AI is also now being used to more efficiently route electricity across a high-dimensional system like the electricity grid,” he added.

However, the rapid development of AI presents a challenge for the infrastructure needed to support it.

Currently, there is mismatch between the development timelines of software, computer hardware and energy infrastructure, said Chueh. While software can evolve within a month, hardware development may take around a year, and major infrastructure projects can take a decade.

“Yes, we can innovate AI very quickly, but it takes us a decade to build up the infrastructure needed to power AI, at the same time, it’s also an opportunity, can we use the power of AI to accelerate, for example, how we deploy infrastructure,” said Chueh.

This creates uncertainty for capital allocators and venture investors making long-term bets while AI continues to evolve, he said.

Investments in energy infrastructure require careful planning because projects take years to develop, while the technologies they are designed to support may change much faster.

“It is absolutely likely that in the next five years, AI then will look very different from today, and I think that poses a challenge for investors who are investing five to ten years out again. I don't think there's any ambiguity on the growth of the AI and the demand of AI, but how it is architected could be changing,” said Chueh.

Despite these uncertainties, Chueh stressed on the importance of keeping pace with innovation rather than focusing solely on energy demands of current AI systems.

“So, because of this rise in demand and the growing of the pie, I think innovation is going to be critical, and I think it's extremely important if you're a capital allocator not to only think about the AI and the tech of today, but also the AI and the tech of tomorrow,” he said.

The impact of this expansion extends to local communities, particularly across Asean and developing nations.

Chueh noted three main dimensions of concern: ensuring that sustainable energy remains affordable addressing the limited job creation associated with data centre developments and finding ways to mitigate the impacts on the local community.

However, he recognises a major opportunity for technological advancements to transform how data centres interact with their host environments.

“AI algorithms have increased together with GPU. The efficiency has gone up by a factor of 1,000 in the past 10 years, and there is very little constraint in how far the efficiency increase can go, so that will lead to lower energy use, lower waste heat and therefore lower water usage. So, I think it's important to think about what innovation can do in terms of the community impact,” said Chueh.

Despite concerns over hyperscalers building massive compute capacity at any cost, Chueh believes this presents an economic opportunity. Because major tech firms are willing to pay a premium for power and hold long-term climate commitments, their financial resources can be channelled into clean firm power and carbon capture.

Adopting these green technologies adds only a modest 2% to 3% premium for end users, he argued, adding that aligning business models would ensure tech giants' spending directly accelerates sustainable deployment and drives down future clean energy costs.

“So, I'm quite excited that with the growth in demand for energy for AI, we can use that willingness to pay to advance the sustainability goals,” said Chueh.

He also stressed the importance of mapping solutions across realistic timelines rather than treating sustainability as a single immediate goal.

“Laying things out in terms of this timeline will give us a better chance of maintaining that reliability, that security and affordability to make sure that there is a licence to continue to advance toward our sustainability goals,” said Chueh.