Welcome back to this week’s Battery Business Insights article, where we’re looking at one of the most consequential demand stories shaping the battery industry right now. Most of the public conversation around AI data centers focuses on memory chip shortages and soaring electricity bills — and for good reason. But there is a second, less-discussed chapter to this story that matters directly to everyone in the battery sector. Artificial intelligence hyperscalers — the same companies pre-buying DRAM wafers by the hundreds of thousands — are fast becoming anchor customers for large-scale battery energy storage systems (BESS), both on-site and off. Understanding why, and how fast, is essential for anyone investing in, building, or supplying battery storage today.
Two Storage Markets, One Demand Driver
- Hard-drive lead times have stretched from weeks to over one year due to AI data center demand; some backlogs extend to two years.
- OpenAI’s “Stargate” project reportedly reserved up to 900,000 DRAM wafers per month from Samsung and SK Hynix — close to 40% of global DRAM output.
- A single 30 TB enterprise SSD jumped 257% in price in under a year (from ~$3,000 to ~$11,000); SSDs now cost up to 16× more per terabyte than comparable HDDs.
- Goldman Sachs projects global data center power demand will grow 50% by 2027 and 165% by 2030, vs. 2023 levels — reaching ~92 GW by 2027.
- Deloitte projects US AI data center power demand surging from 4 GW in 2025 to 123 GW by 2035 — a more than 30× increase.
- The global data center sector is projected to add 97 GW between 2025 and 2030, effectively doubling in five years.
- The average wait time for a grid connection in primary data center markets now exceeds four years
- US Q1 2026 battery storage installations hit 9.7 GWh — up 32% year-over-year, the highest-ever first-quarter figure.
- The data center battery market is forecast to grow from $4.82 billion in 2026 to $10.23 billion by 2032 (13.4% CAGR).
- Benchmark Mineral Intelligence: energy storage will account for 41% of total US battery demand in 2026, up from 26% just two years earlier.
How AI Turned Data Centers Into Insatiable Storage Consumers
For most of the past decade, data center storage demand grew steadily but predictably. Cloud storage expanded, enterprise IT budgets rose, and HDD and SSD manufacturers managed supply cycles in the familiar boom-and-bust rhythm of the semiconductor industry.
Generative artificial intelligence broke that rhythm. Large language models are trained on petabyte-scale datasets, requiring high-throughput storage systems capable of feeding GPU clusters continuously. Each node in a large training cluster can need hundreds of gigabytes of DRAM and multiple terabytes of flash storage. As training clusters scaled from roughly 40 MW systems in 2020 toward projected 1–5 GW clusters by 2030, the associated storage requirements scaled in proportion. Dell’Oro Group now projects the combined HDD and SSD drive market will grow at a CAGR of more than 20% over the next five years — driven almost entirely by AI data center demand.
The power dimension followed the same trajectory, only faster. Traditional data center racks consumed 5–15 kW each. AI-optimized racks now require 30 kW to over 100 kW — a tenfold increase at the top end. The result is that a campus requiring 500 MW or more of grid power is no longer unusual; it is the new normal for hyperscaler builds. Goldman Sachs and McKinsey both project that roughly 70% of all new data center capacity built through 2030 will be AI-ready, with McKinsey forecasting 33% annual growth in AI-capable capacity from 2023 to 2030. The IEA puts global data center electricity consumption at 945 TWh by 2030, rising to 1,200 TWh by 2035.
From Grid Bottleneck to Battery Opportunity
The shift that matters most for the battery industry is not the electricity demand itself — it is what AI data center operators are doing about it. And what they are doing is buying battery storage.
In primary US data center markets, the average wait for a grid connection now exceeds four years. In many ISO and RTO regions, interconnection queues already stretch 36–48 months. For hyperscalers that have committed to delivering AI capacity in months, that timeline is simply not acceptable. The response has been to go around the queue. According to ING, more than 55 GW of behind-the-meter power capacity is now planned for US data centers — a figure that exceeds New York State’s entire installed generation capacity. Battery storage is a central component of that strategy: it allows operators to right-size their initial grid connection, buffer renewable generation, and manage the MW-scale load swings that AI workloads produce within a single facility.
The project activity has shifted from pilot-scale to commercial-scale in a short time. Meta and Enbridge announced a partnership pairing 365 MW of solar with 200 MW / 1.6 GWh of Tesla Megapack batteries to power a Meta data center under construction near Cheyenne, Wyoming. Tesla and NatPower signed a multiyear agreement covering more than 25 GWh of BESS across Italy and the UK, specifically designed to deliver dispatchable power to large consumers including data centers. In parallel, Sunrun, Renew Home, and Tesla announced a framework capable of delivering more than 16 GW of fast, flexible power for data centers and large loads using distributed home battery resources and managed demand. The idea that battery storage is becoming a frontline infrastructure asset for the AI build-out, rather than a backup or ancillary service, is now showing up clearly in both project pipelines and corporate announcements.
What the Hyperscaler Playbook Means for Battery Manufacturers
The clearest signal that this is a structural shift — not a temporary spike — is the supply-chain behavior of the hyperscalers themselves. When these companies decide something is a strategic input, they pre-commit at a scale that restructures entire markets. The DRAM example is instructive: OpenAI’s “”Stargate”” deal for up to 900,000 wafers per month reportedly covers close to 40% of global DRAM output. The same logic is now being applied to battery storage.
Battery manufacturers are reading this correctly and repositioning. Panasonic Energy announced in June 2026 that it will convert its EV battery plant in Kansas to produce batteries for data center applications, starting in Q3 2029, with parallel retooling of production lines in Japan and Mexico. Samsung SDI unveiled AI data center battery solutions in January 2026, including UPS systems with up to 15-year lifespans and a next-generation Samsung Battery Box (SBB) 2.0 using LFP chemistry for large-scale applications. CATL presented dedicated AI data center energy storage solutions at Intersolar Europe 2026 in June, alongside the debut of its TENER Sodium system — and in April 2026 had already signed a three-year, 60 GWh sodium-ion supply contract with HyperStrong, the world’s largest sodium-ion commercial order at the time. Meanwhile, EnergyStrat describes AI data centers as “”the industry’s most lucrative anchor tenants”” for new BESS projects, with battery pack costs now at or below $80/kWh making large-scale procurement commercially viable.
One caution worth noting: the transition is real but uneven. Linda Li, CFO of Re-Teck, told Forbes in June 2026 that “”nobody has a battery farm to supplement the energy demand on their AI data centers yet, because it is too expensive.”” The practical reality today is that most data centers still rely on battery storage primarily for UPS and short-duration backup — though ZincFive’s 2026 industry survey shows that 52% of data center respondents now cite managing AI workload power as a core storage priority, up sharply from 37% in 2025. The larger grid-side and co-located BESS projects represent the next phase, and the deals are starting to flow.
The Storage Crunch Parallel: A Useful Warning for Battery Supply Chains
It is worth pausing on the memory and storage shortage story, because it offers a direct warning for battery supply chain planners. The hard-drive and NAND flash markets did not see this crunch coming — or rather, the speed and concentration of demand caught suppliers flat-footed. HDD and NAND flash are now simultaneously constrained for the first time in years. Up to 70% of all memory production in 2026 is estimated to be going to AI data centers. A 30 TB enterprise SSD that cost $3,000 in early 2025 now costs close to $11,000. Consumer electronics, PCs, automotive, and telecom — every segment that shares components with the AI data center supply chain — is absorbing the price impact.
Battery storage is not DRAM. The manufacturing base is broader, the technology mix is more varied (LFP, NMC, sodium-ion, nickel-zinc), and pack costs have already fallen more than 90% since 2010. But the structural dynamic is comparable: a small number of extremely well-capitalized buyers are starting to commit to very large volumes, and they have both the incentive and the financial capacity to pre-commit supply years in advance. BloombergNEF confirmed in 2025 that the levelized cost of battery storage has fallen below that of gas turbines for the first time — making large-scale BESS a cost-competitive option for data center power management at the precise moment demand is accelerating. In 2022, just one GWh-scale battery facility existed anywhere in the world. By late 2025, 42 were operational, with more than 250 additional projects expected by 2027. That trajectory is being shaped, in increasing measure, by AI infrastructure demand.
China is already ahead on volume. China Mobile Energy Technology reported 24.5 GWh of Chinese telecom sector storage installs in 2024, and executives speaking at the Energy Storage International Conference in April 2026 projected the AI data center storage market reaching tens of GWh by 2028, with a CAGR above 50% over traditional data center deployments. CATL’s own energy storage system sales reached 69 GWh in 2025 — up 46.81% year-over-year.
(Sources: Tom’s Hardware October 2025; Research data; Prosperous America May 2026; BloombergNEF via Prosperous America May 2026; Caixin Global April 2026; CATL/WEF January 2026)
The Structural Case for Long-Term Battery Demand
The demand signal from AI data centers is not a one-cycle event. Iron Mountain projects AI inference workloads alone will grow at a 79% CAGR through 2030, accounting for roughly 80% of total artificial intelligence IT load capacity by the end of the decade. Every inference interaction generates outputs, logs, and metadata that require persistent storage — and every facility serving those workloads requires power that the grid, in most cases, cannot yet reliably provide at the speeds hyperscalers require.
This creates a structural, multi-year case for battery storage demand of a type the industry has not seen before: buyers with trillion-dollar balance sheets, procurement horizons measured in decades, tolerance for large upfront capital commitments, and an operational need for power reliability that makes battery storage non-negotiable. The question for the battery industry is not whether this demand will materialize — it already is. The question is whether the supply chain, manufacturing capacity, and project development pipelines can scale fast enough to meet it.
FutureMarkets projects data centers will be the fastest-growing commercial and industrial BESS application through the late 2020s. IDTechEx forecasts the global C&I BESS market will grow at a 17.4% CAGR from 2026 to 2036, with data centers as a primary driver. Bloom Energy’s 2026 Power Report found that expectations among hyperscalers and colocation providers for fully onsite-powered data centers increased 22% in just the past six months — and more than one-third of data centers are expected to use 100% onsite power by 2030.
Battery storage manufacturers, project developers, and investors who treated data centers as a niche application until recently may want to revisit that assumption.
Bottom Line
AI data centers started by squeezing the world’s DRAM and hard-drive supply. They are now doing the same to grid capacity — and the response to that grid constraint is redefining who the battery industry’s biggest customers are. Grid interconnection queues of four or more years, rack power densities of up to 100 kW, and hyperscaler capex commitments in the trillions are turning battery storage from a grid ancillary service into a core infrastructure requirement for the artificial intelligence economy. The deals are being signed, the manufacturing lines are being retooled, and the demand projections are pointing in one direction. For the battery industry, AI data centers are no longer a side story — they are one of the most significant new demand drivers on the horizon.
