A new category of buyer is entering the storage market—and they're not building wind farms
In Northern Virginia, the queue for new grid connections has grown long enough that some data center developers have stopped treating power access as a procurement problem. They have started treating it as a site selection variable.
More than a quarter of the world's hyperscale capacity is concentrated in the region, yet utilities are struggling to absorb a wave of load requests that has little precedent in commercial electricity planning. The constraint is not generation. It is the physical infrastructure required to move power from where it exists to where it is needed—transmission lines, substations, interconnection approvals—and that infrastructure moves on a timeline that has no relationship to the competitive pressures driving AI deployment. Similar pressures are visible in Texas, Ireland, and parts of continental Europe, where regulators and network operators are working through a version of the same problem with different grid architectures and different levels of urgency.
The effect is beginning to show up in the storage market. Several developers active in North American projects say conversations with data center operators have become more frequent over the past year or two, and the nature of those conversations has shifted. What typically started as a discussion about backup power requirements has expanded into questions about phased energization strategies, demand charge exposure, and how batteries might reduce dependence on grid capacity during the years it takes for transmission upgrades to clear regulatory and construction hurdles. That is a different commercial conversation from the one that has historically driven utility-scale storage procurement.
The demand numbers help explain why. The IEA estimates global data center electricity consumption at roughly 415 TWh in 2024, with a base case projection of around 945 TWh by 2030, most of it driven by AI workloads operating at computing densities that conventional cloud infrastructure rarely approached. Lawrence Berkeley National Laboratory data puts the median wait time for grid interconnection in the US at around five years, up from under two in the early 2000s. Some operators are completing construction on facilities they cannot yet energize. The capital sitting idle in those buildings has a way of concentrating minds on alternatives.
Not every data center operator is a natural storage customer. In markets where grid connections are available and electricity costs are low, diesel generation remains competitive on a pure backup basis, and transmission investment is still the only credible long-term answer to capacity constraints. What has changed is the threshold at which batteries enter the evaluation. Developers facing multi-year interconnection delays are running numbers they would have dismissed a few years ago. The option has not become cheap. It has become worth pricing.
The demand charge angle has drawn particular attention from operators running large AI training workloads. Unlike inference, which tends to draw power at relatively stable rates, training clusters can create sharp fluctuations in consumption when large numbers of processors run simultaneously. In markets where demand charges represent a meaningful share of total electricity costs, those fluctuations translate directly into operating expenditure. Storage developers working with data center customers describe an evaluation process that looks quite different from a utility procurement—less about energy arbitrage or grid services revenue, more about reducing peak exposure and managing tariff risk. It changes the return model, and it removes the requirement for the customer to have a sophisticated view of wholesale electricity markets.
Corporate energy procurement adds another dimension. Microsoft, Google, and Amazon have each expanded commitments to match electricity consumption with carbon-free generation on an hourly rather than annual basis. That target becomes progressively harder to meet as total demand rises and AI infrastructure becomes more geographically concentrated. Storage appears in those conversations with increasing frequency, though developers are careful about overstating the current commercial traction. The economics remain sensitive to local renewable supply, grid tariff structures, and the specific terms of any associated power purchase agreement.
Utilities still account for the overwhelming majority of large-scale battery procurement globally, and nothing in the current pipeline suggests that is about to change. What has shifted is the geographic distribution of where serious commercial inquiries are coming from. In regions where power availability has become genuinely uncertain, storage companies report that technology firms and data center operators now represent a growing share of the conversations they are having. Whether that inquiry volume converts into a procurement wave, or stays at the level of evaluation and option-pricing, is the question the market is still working through.
