AI Data Centres Drive Battery Demand as Power Surges Become a Grid Risk

Battery start-ups are moving deeper into the data centre market as artificial intelligence workloads create new challenges for power stability, grid connection and infrastructure planning.

According to the Financial Times, companies developing advanced battery technologies are seeing strong demand from AI data centre operators that need systems capable of smoothing sharp power surges. The issue is not only the total amount of electricity consumed by AI facilities, but also the speed and volatility of demand when thousands of processors operate in parallel. Some AI workloads can create energy spikes of tens of megawatts within milliseconds, placing pressure on conventional backup and power management systems.

AI Is Changing the Power Profile of Data Centres

The shift is opening a new market for battery suppliers that were originally focused on electric vehicles, grid storage or industrial applications. The Financial Times cited companies including Alsym Energy, Nyobolt and QuantumScape as examples of battery developers exploring the opportunity. Their technologies, including sodium-ion, high-power lithium-ion and solid-state systems, are being positioned as tools for fast response, high reliability and improved power quality in facilities where downtime can be extremely costly.

The development comes as AI infrastructure becomes one of the fastest-growing sources of electricity demand. The International Energy Agency expects global electricity consumption from data centres to more than double by 2030 to around 945 terawatt-hours, with AI as the main driver. In its base case, the IEA estimates that data centres will account for just under 3% of global electricity consumption by 2030, but growth will be highly concentrated in specific regions and grid connection points.

This matters because AI data centres are increasingly dense. Modern GPU clusters can draw far more electricity per rack than conventional server halls, while liquid cooling, power conversion equipment and grid interconnection systems take up a larger share of facility design. Racks running high-performance AI chips can require many times more electricity than regular web servers, changing the layout and engineering requirements of large-scale computing sites.

Batteries Move Beyond Emergency Backup

For data centre operators, batteries are becoming more than emergency backup assets. Traditional uninterruptible power supply systems and diesel generators were designed mainly to prevent outages and bridge short disruptions. AI facilities increasingly need equipment that can respond almost instantly to fluctuations in load, helping to protect processors, manage voltage and frequency stability, and reduce stress on the local grid.

This creates an opportunity for battery technologies that can deliver rapid discharge, frequent cycling, and stable performance under demanding conditions. Some systems are designed to provide high power over short periods, while others are being developed for longer-duration support or safer chemistry profiles. The common requirement is responsiveness. AI workloads can ramp up and down quickly, and power systems must keep pace without compromising reliability.

In practice, batteries can sit between the data centre and the grid, absorbing sudden changes in demand and providing power during brief disturbances. They can also help operators manage peak demand charges, improve power quality and reduce dependence on diesel backup generation. In some markets, they may allow data centres to provide flexibility services to electricity networks, although this depends on regulation, market design and operational constraints.

Grid Reliability Becomes a Strategic Concern

The rapid expansion of AI data centres is drawing attention from utilities, regulators and grid operators. Large computing facilities can require hundreds of megawatts of capacity, sometimes in areas where grid infrastructure is already constrained. Connection queues, transformer shortages and transmission bottlenecks can delay projects or increase costs.

The reliability challenge is different from simply adding more demand. AI workloads may be concentrated, fast-changing and difficult to predict. If several large facilities are connected to the same regional grid, their combined behaviour can affect local voltage, frequency stability and peak-load planning. This is why some grid operators are paying closer attention to emerging large loads, including data centres, hydrogen production and industrial electrification.

Battery systems could help reduce some of these risks by acting as a buffer between highly variable AI workloads and the electricity network. They can smooth short-duration demand spikes, support critical systems during grid disturbances and help facilities coordinate more effectively with utilities. However, they do not remove the need for new generation, substations, transmission capacity and better long-term planning.

A New Market for Battery Developers

For battery companies, AI data centres may become an attractive customer segment. Electric vehicle markets have become more competitive, particularly because of manufacturing scale, falling cell prices and pressure from Asian suppliers. By contrast, data centres are high-value customers where reliability, safety and fast response can justify premium technologies.

This could benefit start-ups working on alternatives to conventional lithium-ion batteries. Sodium-ion systems may offer advantages in material availability and safety. Solid-state batteries may eventually provide higher energy density and improved performance, although commercial scale-up remains challenging. High-power lithium-ion systems, including those optimized for rapid charging and discharging, may also find a role in short-duration data centre applications.

The opportunity is not only about selling battery cells. Data centre operators need integrated systems that include battery management software, power electronics, thermal controls, fire safety, grid interface technology and maintenance services. Suppliers that can deliver complete, reliable and certified solutions may be better placed than companies offering cells alone.

Sustainability Benefits Depend on System Design

From a net-zero perspective, batteries can support cleaner and more resilient data centre operations, but their climate impact depends on how they are used. If battery systems reduce diesel generator runtime, support renewable energy use or help avoid grid instability, they can contribute to lower emissions. If they are used mainly to enable rapid growth in fossil-powered electricity demand, the benefits are more limited.

Battery manufacturing also carries environmental impacts. Material extraction, refining, cell production and end-of-life management all need scrutiny. Data centre developers that adopt large-scale battery systems will need to consider lifecycle emissions, responsible sourcing, recycling pathways and fire-safety standards. These factors are particularly important as companies make claims about low-carbon digital infrastructure.

Clean electricity procurement remains central. Batteries can shift and stabilize demand, but they do not generate power. AI data centre operators will still need credible renewable power purchase agreements, direct investment in clean generation, demand flexibility, energy efficiency improvements and transparent reporting of electricity use. Hourly matching of clean power may become more relevant as electricity demand from AI grows.

Implications for Net-Zero Stakeholders

The rise in battery demand from AI data centres highlights a broader shift in the energy transition. Digital infrastructure is becoming a major power-sector actor. Its sustainability will depend not only on renewable energy contracts, but also on where facilities are built, how they connect to the grid, how flexibly they operate and whether they invest in technologies that reduce stress on electricity systems.

For utilities, the priority is better coordination with data centre developers. Grid studies, connection agreements and demand-response arrangements will need to reflect the distinctive power profile of AI workloads. For regulators, the challenge is to ensure that new computing loads do not raise costs or reliability risks for other customers without appropriate planning and investment.

For corporate buyers of AI services, energy transparency may become a more important part of procurement. Companies using AI at scale will increasingly need to understand where computing is hosted, what electricity mix powers it and whether infrastructure providers are managing grid impacts responsibly.

For battery suppliers, AI data centres could become a valuable new growth market. For the energy sector, they are another reminder that electrification, digitalisation and decarbonization are now tightly connected. The next stage of AI growth will be shaped not only by chips, models and software, but by the ability to deliver stable, low-carbon power at speed.

Source: www.ft.com