Investment implications of AI’s huge appetite for energy
In response to the anticipated rise in demand for electricity, Big Tech companies have been investing in nuclear, solar and wind power
I ASKED Meta’s artificial intelligence (AI) tool to suggest a headline for this article. By doing so, I may have used 10 times the electricity required to answer a traditional Google query.
As the use of AI becomes more widespread, demand for electricity is expected to rise. Currently, data centres account for about 2 per cent of electricity consumption in America, said the US Department of Energy. However, the Electric Power Research Institute (EPRI) estimated that data centres could grow to consume 4.6 to 9.1 per cent of US electricity generation annually by 2030.
Before ChatGPT arrived in November 2022, the growth in power demand from computing and data storage had been largely offset by data centre efficiency gains, as small, relatively inefficient corporate data centres were replaced by large cloud computing facilities.
Furthermore, Koomey’s Law also suggests that as computing devices advance, the amount of energy necessary to perform a set amount of computing falls by half every two-and-a-half years. The arrival of AI computations, however, represents a steep increase in power demand that could well exceed efficiency gains. Part of the uncertainty in EPRI’s estimated range is due to varying assumptions on the extent of efficiency gains in the future.
Big Tech is getting ready
The Big Tech companies, which are investing heavily into AI, are already responding to the anticipated rise in power demand. With aggressive clean energy targets, many of the technology companies are likely to seek low or zero-carbon solutions.
Several nuclear power deals have been making headlines. Nuclear plants do not emit carbon. They also have the advantage of operating continuously, unlike renewable power sources – such as solar and wind – that depend on the weather. However, nuclear power plants can take a long time to build, are very expensive, and come with safety concerns.
BT in your inbox
Start and end each day with the latest news stories and analyses delivered straight to your inbox.
In March 2024, Amazon Web Services (the largest cloud service provider) bought for US$650 million a data centre campus in Pennsylvania which is directly connected to a nuclear power facility. In October 2024, Amazon said it would finance the construction of several small modular nuclear reactors (SMRs) in Washington state and invest in a startup which will build the reactors. SMRs do not currently operate in the US but are faster and less costly to construct compared to large nuclear plants.
Earlier in October, Alphabet’s Google also announced it was backing the construction of seven SMRs, to add 500 megawatts of nuclear power in total. The first reactor is expected to be online by 2030, followed by others up till 2035, to power Google’s global data centres and offices.
Meanwhile, Microsoft announced in September 2024 that it was backing the restart of a nuclear plant in Pennsylvania to provide electricity for its data centres in a 20-year supply deal.
Besides nuclear power, the Big Tech companies have also been investing in solar and wind power – which are more readily available today compared to nuclear power – as well as geothermal power, another form of renewable energy. But given the significant increase in power demand driven by AI, new nuclear capacity is seen as essential by the Big Tech companies.
Investment implications
The growth in US electricity consumption this year and the anticipated future increase have been major drivers of the strong performance of the utilities sector, one of the top performing sectors in the US this year. We believe demand for clean power, especially from Big Tech companies, favours independent power producers that have the ability to craft customised power contracts including long-term power purchase agreements.
Nuclear power operators and the associated supply chain are also well positioned in this environment. However, in the near term, the US utilities sector appears overbought – which is likely to result in some consolidation. We would use any pullback in US utilities to build a core allocation within a portfolio.
In addition, there are companies in the industrial sector that can benefit from growing electrification and the expansion of utilities’ power capacity. There are also attractive companies that benefit from the buildout of data centres, such as companies that provide cooling systems that are critical for the efficient operation of the data centres, as well as power and monitoring solutions. An uninterruptible power supply system is seen as essential for data centres, given the need to provide backup power and prevent costly downtimes.
AI development remains in its early days, in our view, with further software and hardware applications to come. This should be supportive of the demand for advanced semiconductors used in AI applications, as well as software companies that can adopt AI to improve efficiency and enhance value for their users.
We have a favourable view of the US technology sector, driven by ongoing investments in AI. Although valuations are elevated, we expect interest rate cuts to mitigate such concerns and earnings growth to drive the sector higher.
Risks to our view would include a significant slowdown in AI investments; weaker-than-expected economic growth that curtails investments into technology; and escalation in geopolitical tensions.
The writer is senior investment strategist at Standard Chartered Bank’s Wealth Solutions Chief Investment Office
Copyright SPH Media. All rights reserved.
