Deirdre Cooper

The amount of electricity consumed by AI is forecast to exceed 85 terawatt hours annually within a couple of years, more than many small countries^[1]. The latest whitepaper by Deirdre Cooper, Head of Sustainable Equity, and Graeme Baker, Co-Portfolio Manager, Global Environment, ‘Powering the AI revolution: a superscale cleantech opportunity?’ ^[2] examines the resultant investment opportunities across renewable energy and energy-efficiency solutions.

AI has the potential to significantly change trends in electricity demand in developed countries. McKinsey and the US Energy Information Administration (EIA) expect a 38% increase in US power demand over the next two decades, a startling increase from the 9% increase seen over the previous 20 years.

Increase in 5-year energy-demand growth forecasts by US utilities

Source: NextEra Energy, 2024

Deirdre Cooper, Head of Sustainable Equity: “While AI could be powered by dirty energy in the short term, over the medium term AI power demand is likely to be met by renewables. This is partly because renewables continue to be the most economic option for new power generation, and most of the assets in the transmission queue are renewables, which is the critical bottleneck. In addition, the biggest investors in AI have been companies with strong commitments to net zero.

As a result, we expect stronger tailwinds for companies involved in delivering clean energy. There are also significant growth opportunities for providers of solutions that improve datacentre efficiency, specifically server and infrastructure efficiency.

Cleantech pricing headroom is growing

As demand for renewable energy increases, power purchasers are already becoming less price sensitive when negotiating offtake contracts. This gives pricing headroom to companies across the cleantech value chain, not just developers like NextEra Energy, Iberdrola and Orsted, but also renewable and battery equipment manufacturers like Vestas Wind Systems, Sungrow, and CATL.

Power delivery and datacentre efficiency are crucial

The biggest challenge for AI datacentres is electrical power, the maximum capacity in megawatts that a datacentre needs available to be able to operate. A typical co-location datacentre^[3[] needs 20-30MW, but a dedicated AI campus might need 100- 500MW. Graeme Baker, Co-Portfolio Manager, Global Environment: “We see an opportunity for companies that provide technologies and services that are crucial to solving this power challenge. For example, locating datacentres close to high-voltage transmission infrastructure can help. Consequently, developers with better sites – such as NextEra Energy, which has significant undeveloped land rights in good locations – are likely to see their competitive advantages strengthen.”

Demand for power has also resulted in an increased need for certain electrical equipment, including high-voltage transformers and switchgear. This has led to shortages, with lead times for some technologies now running to years. Cooper adds: “Given the power-supply challenges facing datacentres in an AI world, the demand for efficiency gains will drive structural growth for companies that provide relevant solutions, such as Schneider Electric and Delta Electronics,.”

Cooper concludes: “It is important to note that most of the forecasted new datacentres are yet to be built, so investors need to treat power-demand forecasts with caution. But if they are even half right, the growth of the market for decarbonisation solutions will accelerate rapidly, expanding the opportunity set for investors.”

Read the report.

——–

Notes:
[1] Scientific American, October 2023.
[2] Powering the AI revolution: a superscale cleantech opportunity?
[3] In a co-location datacentre, multiple companies/third parties rent space. An enterprise datacentre is owned/used by one company

The post Powering the AI revolution: a superscale cleantech opportunity? appeared first on AdviserVoice.