As AI becomes increasingly integrated into everyday life, data centres powering the technology are expanding, consuming unprecedented amounts of electricity and pushing power grids to their limits. Amid concerns, energy storage is stepping in to keep AI operations reliable, resilient and sustainable.
According to the International Energy Agency, data centre electricity consumption is set to more than double to around 945 terawatt-hours by 2030. In the US, which accounts for the largest share of this projected increase, data centres are expected to consume more power than all other energy-intensive industries – including aluminium, steel, cement and chemicals – combined.
Go deeper with GlobalData
Discover B2B Marketing That Performs
Combine business intelligence and editorial excellence to reach engaged professionals across 36 leading media platforms.
To support this transformation, energy storage is emerging as the answer. Power Technology explores its role in managing AI’s power demand and how hyperscaler-storage partnerships and various deal structures are shaping the growth of the storage market.
Why AI needs energy storage
Data centres – especially those operated by hyperscalers like Google, Amazon and Microsoft – are designed for non-stop performance. They are energy-intensive, high-stakes operations that cannot afford downtime.
However, as their power demand rises, so too does the strain on the grid, increasing the risk of disruptions.
Adding to concerns is hyperscalers’ new-found mission to decarbonise. “Not only do they need access to huge amounts of power, but they are focused on getting renewable power,” says Michael Hunter, head of commercial and the UK data centre portfolio at Apatura.
US Tariffs are shifting - will you react or anticipate?
Don’t let policy changes catch you off guard. Stay proactive with real-time data and expert analysis.
By GlobalDataWhile renewables are expanding, current grids have struggled to manage the rapidly increasing load and intermittency of renewables. In the UK, grid constraints are projected to cost bill payers around £3bn ($4bn) by 2030.
Enter energy storage – offering a “strategic buffer to counterbalance fluctuating energy requirements [of data centres] and facilitating the incorporation of renewable energy sources”, says Rehaan Aleem Shiledar, senior energy analyst at Power Technology’s parent company GlobalData. “It diminishes dependence on the grid during periods of peak demand and supports uninterrupted operations in the event of power disruptions.”
Allowing renewable energy to be fully utilised “offers savings from an operating expense (opex) perspective through energy arbitrage – charging batteries when costs are low and discharging when energy is more expensive”, Hunter adds.
Beyond optimising energy consumption, storage also strengthens cybersecurity. “Modern grids exhibit vulnerabilities to cyberattacks, such as inadequate authentication measures in legacy systems, dependence on outdated protocols, hastily implemented digital tools and an absence of strong encryption,” Shiledar confirms.
“Energy storage can serve as an essential safeguard by supplying localised backup power and bolstering grid resilience by facilitating a more decentralised grid infrastructure, mitigating the risk of disruptions resulting from cyberattacks on the grid.”
For facilities that can’t afford even seconds of downtime, such resilience isn’t a luxury – it’s a requirement.
Batteries take the lead
When it comes to choosing the right type of storage for data centres, experts agree on one technology: batteries.
“Battery energy storage systems (BESS), as the technology stands today, offer the most flexibility and provide opex benefits from energy arbitrage and easy access to renewable energy,” says Hunter. “Another emerging benefit is using batteries as part of the uninterruptible power system that data centres need, thus also saving on capital expenditure (capex).”
Battery tech has improved dramatically in recent years, in addition to “shattering the cost curve downwards”, says Andrés Acosta, director of innovation at LevelTen Energy, “displacing other types of storage that were available and enabling things that weren’t thinkable a few years ago”.
While solutions like hydrogen and pumped hydro storage have found their niches, “the market for wholesale hydrogen is still in nascent stages”, Shiledar notes, while hydro applications remain restricted by geographical constraints.
BESS therefore offers the technological maturity and flexibility that hyperscalers need today, “making it the safest investment”, Hunter affirms. “As long-duration battery systems continue to mature, that is where the [data centre] market will shift towards, rather than entirely new technologies.”
However, for an AI-powered future, energy storage – including mature BESS – must scale further.
Win-win: building up the storage market for an AI future
Shiledar stresses that scaling will require persistent innovation to enhance the energy density and longevity of storage solutions to satisfy the requirements of AI data centres’ load profiles, “rendering energy storage more attainable and cost-effective”. This starts with policy support of research and development in the sector.
He also notes governments’ roles in stimulating the market through “financial incentives, comprehensive regulatory frameworks, clearly defined mandates and targets, and market reforms”. Hunter exemplifies the UK’s AI Opportunities Action Plan announced in January 2025, which established AI Growth Zones (AIGZs) that will see accelerated development of data centres, with AIGZs’ key criteria including land suitability for battery storage.
As such, “policymakers are best placed to bring together grid operators, storage developers and tech companies to align their problems and solutions”, says Hunter.
But even without the policy push, hyperscalers can fuel the storage market’s growth through strategic partnerships with storage developers.
Hyperscaler-storage collaboration is mutually beneficial. Large tech companies have the financial resources to back storage development – solving the cost issue for scaling – while storage helps hyperscalers “enhance both the efficiency and sustainability of their operations”, says Shiledar, not only allowing them to “lead the way in reducing the carbon footprint of the tech industry” but also providing operational and financial value.
Acosta notes: “The systems, technology and willingness to develop projects are there – all we need now are contracting structures that are simple and viable for hyperscalers while allowing developers to properly finance projects.”
Making the business case work
Structuring these hyperscale-storage deals isn’t always straightforward.
Speaking from his experience at Google, Hunter says that a major hurdle to fostering hyperscaler-storage partnerships has been a lack of transparency. “Hyperscalers have traditionally been secretive about future plans, not wanting to reveal locations to avoid land speculation or giving competitors a heads-up,” he says.
These attitudes have evolved as hyperscalers have begun to recognise that “the commercial value of strategic partnerships – opex and capex savings at that scale – make openness more worthwhile”.
On LevelTen’s platform – an online marketplace facilitating energy transactions – Acosta has observed that while renewables power purchase agreements (PPAs) had been popular between hyperscalers and energy providers in recent years, buyers (hyperscalers) are now finding them “less attractive and effective” due to “the cannibalisation of prices and surge of negative hours in many markets”.
“That is where batteries come in, allowing buyers to capture higher prices,” the LevelTen official says. “Sellers (energy providers) need project finance and stable revenue since they are in the business of making money through energy. Buyers are not. They just want 24/7 matching and delivery profiles that can adjust to their consumption profile.
“Adding batteries into the equation helps bridge this gap.”
This is easier said than done, however, as Acosta explains: “With traditional renewables PPAs, you value a PPA by comparing the capture price of a certain delivery profile against the PPA price – it is relatively easy… But when you add batteries, you are not just evaluating a long-term price forecast, but also how you are going to operate the battery, like how well the battery asset can capture those low-price hours and sell during high-price hours, among other factors.”
Deal structures for storage must accommodate for the complexities of battery operations, long-term energy pricing and risk exposure to represent a true win-win for hyperscalers and storage developers.
Acosta has seen various transactions emerge, which depend on the type of battery asset as well as the market.
Stand-alone batteries are simpler and often rely on tolling agreements (where the asset owner is paid a fee for use of their generation or storage asset) with “fixed revenue, no market exposure”, he says, ideal for developers seeking stable financing. Others prefer partial tolling – where part of the battery capacity is tolled while the developer keeps a portion – to “gain exposure through multi-market optimisation strategies”.
For co-located batteries – like solar-plus-storage – deal-making gets more creative. “You not only have to manage exposure to risk but also the relationship between the generation and storage asset,” Acosta explains.
Fixed-shape PPAs have become more common as they are easily understood by offtakers – hence, “easier to sell” – but “leave potential upside of not extracting the best value in different markets”. On the other, more complex extreme in which offtakers take energy only from the solar asset and share the battery revenue, providers gain full market exposure but struggle to “negotiate this with offtakers because offtakers aren’t in the business of making money with batteries”.
“Lastly, there is everything in between… this is a more balanced approach in which you have covered some risk from having a fixed price for energy while still having some exposure to various markets,” he summarises.
Acosta adds that in markets such as Italy with government tools like the MACSE (Mechanism for the Procurement of Electric Storage Capacity), regulation defines how battery capacity is allocated across state mechanisms versus in a merchant strategy.
Newer deal models like storage-as-a-service (STaaS) are also emerging. “By diminishing the need for substantial initial capex, STaaS provides flexible and scalable storage solutions that enhance operational cost efficiency,” says Shiledar. “This model enables businesses to pay exclusively for the storage capacity they utilise, obviating the necessity for significant investments in hardware and infrastructure.” Companies such as ABB and GridBeyond have entered battery STaaS partnerships to strengthen renewables integration and cost efficiency.
Meanwhile, “energy-integrated service level agreements (SLAs) can support the environmental sustainability objectives”, Shiledar adds, as SLAs guarantee that a portion of energy usage is derived from clean sources or compensated for through energy storage.
In the end, however, tech companies must now acknowledge that energy storage is a strategic asset, not just an ESG (environmental, social and governance) box-ticker.
The future is AI, undoubtedly so, as the technology is poised to play a critical role across most – if not all – industries, including energy. However, its benefits come with risks that must be managed now to assure that it becomes a core enabler of grid reform and the energy transition, not the grid’s greatest source of strain.
Energy storage systems will be key to managing these risks, and their expansion starts with hyperscalers recognising their value. “We can’t wait for policy alone. Private sector innovation and collaboration is needed for scalability,” says Hunter.
He claims that the tech industry is already taking the first steps to embedding storage within the AI supply chain: “Embedding BESS within the data centre footprint – using it not just for green power, but also uninterruptible power – makes sense. Industry is realising the opex and capex benefits [of storage] while increasing resilience.
“We will need innovation in generation – like small modular nuclear reactors – to manage AI data centres’ power load, but that is more medium-term,” he notes. “For now, energy storage is the most logical way to maintain continuity between generation and demand.”
Whether it is smoothing grid pressures, enabling decarbonisation or supporting 24/7 data processing, energy storage is well-positioned to become the heart of resilience in the digital age.
