Alan Greenshields, Director of Europe at ESS Inc.
The Artificial Intelligence (AI) surge is driving demand for electrical power, with datacentres predicted to consume over 20% of global energy by 2025.
At a time when the climate crisis is intensifying, there is a critical need to power these datacentres sustainably and meet their electricity requirements with renewable energy. This pressure is exacerbated by the need to ensure energy security and limit the energy cost volatility that has challenged businesses and consumers in recent years.
At the core of the problem is the sheer power required by AI chips. The most popular AI chip, Nvidia’s A100, requires ~400W of power for each instance. When this is replicated tens of thousands of times over in the datacentres of Google, Amazon, Microsoft and others, it adds up to AI becoming a formidable consumer of electrical power.
Addressing energy needs
While AI power consumption might be seen as a challenge for a world trying to reduce carbon emissions, the reality is that it is a necessity for any developed economy to ensure global competitiveness.
However, as technology advances, so do power requirements. The top-performing Nvidia H100 AI chip consumes 700W of power which is equivalent to a standard microwave oven and thousands of these will be required in an AI datacentre. UK-based AI Supercloud company NexGen Cloud, for example, is planning to deploy 20,000 Nvidia H100s by summer of 2024, while Microsoft has committed to a £2.5 billion investment in the UK, with the ambition of doubling its AI datacentre capacity in the UK by 2026.
Impact of increased AI datacentre deployment
Increasing deployment of datacentres to meet rising demand has major implications for the environment, grid stability and energy sources.
The power used for the day-to-day running of data centres has meant they now contribute more to global carbon dioxide levels than the aviation industry underscoring the need to meet this demand with renewable energy.
Thankfully the adoption of renewable energy for AI is already taking place with major datacentre providers like Google Cloud committed to a carbon-free operations by 2030, while Digital Reality has a global renewable coverage of 62% across its data centres.
Others need to follow suit if an AI-driven climate crisis is to be avoided. It is crucial new datacentre power demands are met through the mass deployment of wind and solar energy.
Energy storage provides clean baseload energy
Solar and wind are low-cost and low-carbon, but their intermittency poses a challenge to AI datacentres that operate 24/7. A potential solution is energy storage, in particular long-duration energy storage (LDES).
New LDES technologies, like ESS’ iron-flow batteries, become crucial in addressing the intermittency issue. These solutions can store up to 12 hours of renewable energy, meeting datacentre demands even when the sun is not shining and wind is not blowing.
However, adding renewables and energy storage to the grid at large is not straightforward. Despite national climate targets, the UK grid is facing delays of up to a 15-year wait to connect new clean energy projects to the grid.
Developing modern grid infrastructure
When powered by the grid, datacentre demand can affect local grid reliability in addition to broader climate impacts. Last year, the head of a Norwegian arms company blamed cat videos for their adverse effect on his organisation’s production of munitions for Ukraine; a datacentre located near the munitions factory was consuming vast quantities of electricity from the grid, affecting their production.
New energy storage technologies afford an opportunity to drive the deployment of new architectures, such as microgrids, that can balance the unpredictability of renewables with the resilience and reliability needs of AI.
This is not theoretical. Today, a microgrid installed at a technology recycling company in the USA using an ESS Energy Warehouse has been integrated with a solar array. This system is helping not only reduce peak demand and provide 24/7 clean energy but is also mitigating the impacts of grid outages to ensure reliable, resilient operations.
New AI technologies, powered by new clean energy technologies, have the potential to deliver a climate-positive feedback loop as further datacentre demand is met by new clean technologies deployed at scale. This portends a future where datacentres’ potential climate liability has become an asset delivering global benefits.
This article appeared in the May 2024 issue of Energy Manager magazine. Subscribe here.