SoftBank plans large-scale battery cell manufacturing for AI services

SoftBank is getting into the battery business. The Japanese tech and investment giant is converting a shuttered Sharp LCD factory in Sakai, Osaka, into a large-scale battery manufacturing facility, with production expected to begin as early as fiscal 2027.

The goal isn’t to compete with CATL or Panasonic in the consumer battery market. These batteries are being built to power SoftBank’s own AI data centers, which are running headlong into one of the industry’s most stubborn problems: there simply isn’t enough electricity to go around.

Why a telecom giant is building batteries

Data center power demand is projected to triple by 2030, driven almost entirely by the proliferation of AI technologies. That’s a staggering ramp, and it’s forcing companies that depend on AI infrastructure to think creatively about where their watts will come from.

SoftBank, which owns chip designer Arm and has poured billions into AI ventures across its Vision Fund portfolio, apparently decided that waiting for the grid to catch up wasn’t an option. Instead, it’s building its own power supply chain from scratch.

Converting existing industrial space is significantly faster and cheaper than building from the ground up, which helps explain why production could begin within roughly two years.

No lithium, no cobalt, fewer problems

Perhaps the most interesting detail is what won’t be inside these batteries. SoftBank is developing cells that use neither lithium nor cobalt, the two materials that have turned battery supply chains into geopolitical chess matches.

Cobalt carries well-documented ethical sourcing concerns, with a significant portion of global supply coming from mines in the Democratic Republic of Congo.

Stationary storage, unlike EVs, doesn’t need the highest possible energy density. It needs to be cheap, durable, and scalable. Removing lithium and cobalt from the equation could meaningfully lower per-kilowatt-hour costs.

What this means for the broader AI energy landscape

Microsoft has signed nuclear power deals. Amazon has invested in small modular reactors. Google is exploring geothermal. Each of these hyperscalers is essentially saying the same thing: the current electricity grid cannot support our AI ambitions, so we need to build our own energy infrastructure.

SoftBank is taking a slightly different approach by focusing on storage rather than generation. Batteries don’t create new electricity, but they do allow companies to stockpile power during off-peak hours and deploy it when AI workloads spike. That flexibility is valuable in Japan, where electricity prices are high and grid capacity is constrained.

The Osaka facility is described as potentially one of Japan’s largest battery production lines.

The risk, of course, is execution. Battery manufacturing is notoriously difficult to scale. Yield rates for new chemistries can be unpredictable, and moving from pilot production to factory-scale output has tripped up far more experienced manufacturers than SoftBank. Fiscal 2027 is an ambitious timeline for a company that has never mass-produced battery cells before.

SoftBank owns key pieces of the AI stack, from Arm’s chip designs to various AI application companies in its portfolio. Securing a dedicated, cost-effective power source for that stack removes one of the biggest constraints on scaling AI operations.

Disclosure: This article was edited by Editorial Team. For more information on how we create and review content, see our Editorial Policy.