Target output is one gigawatt-hour a year. Partners are South Korea's Cosmos Lab and DeltaX. Production opens for the fiscal year starting next April, with zinc-halide chemistry following in 2027.

SoftBank Group's mobile-services subsidiary will begin manufacturing large-scale battery cells at the Sakai, Osaka site that once belonged to Sharp, the company said on Sunday, targeting roughly one gigawatt-hour per year of output once the line is at scale.

The plant supplies storage for the AI data centres SoftBank is already building, and for grid, industrial, and residential customers beyond that.

The conversion has been signalled since April, when SoftBank confirmed that the 440,000-square-metre former LCD factory it bought for about ¥100 billion would house Japan's largest battery production line.

The Sunday disclosure sharpens the timeline. Mass production begins in the fiscal year starting April 2026, in partnership with two South Korean firms: Cosmos Lab, which contributes zinc-halide cell chemistry, and DeltaX, which contributes systems integration.

Zinc-halide manufacturing is targeted for 2027, with lithium-iron-phosphate and other chemistries supplying earlier volumes.

The economic logic is the same one that drove the Sharp acquisition. AI data centres need large, fast-discharge storage to smooth power demand and survive grid events; the lithium-iron-phosphate and zinc-halide combinations SoftBank is building are well suited to that profile and avoid the rare-earth and cobalt supply chains that have come under Chinese export-control pressure.

Zinc-halide in particular is non-flammable and uses widely available materials, which matters for data-centre fire-code approval and for siting batteries inside or near server halls.

The Sakai conversion is one of the more concrete examples of Japanese industrial reuse: a high-volume display factory built for a category that lost out to OLED, now repurposed for the storage category AI has made urgent.

The economics of that reuse only work if utilisation runs high, which is why SoftBank has chosen to vertically integrate. The company is both the data-centre operator and the battery customer.

SoftBank's earlier conversion of the Sharp LCD factory described that integration as a hedge against the multi-year lead times facing equivalent storage projects elsewhere.

The wider context is a global storage-capacity shortage. The $1.4 trillion of utility capex US grids are spending to keep pace with AI demand is testimony to that pressure, and the wave of energy-storage startups already chasing that demand has so far produced more demos than deployable gigawatt-hours.

SoftBank is betting that owning manufacturing, rather than buying cells, gives it both control over delivery schedules and a margin on units it does not consume internally.

The partner choice tells its own story. Both Cosmos Lab and DeltaX are South Korean, which positions the joint venture inside the South Korean battery-tech ecosystem rather than the Chinese one. That has commercial and policy advantages.

South Korean battery companies have continued to invest in research and pilot lines through the AI build-out; Chinese-sourced chemistry would carry both Japan's tariff-and-friction risk and the prevailing investor preference for non-China supply chains.

SoftBank invested separately in Cosmos Lab earlier this month, a deal that reads as preparation for this announcement.

One gigawatt-hour a year is meaningful but not dominant. The number puts Sakai on the list of Japan's largest battery facilities; it does not get it into the global top tier, where Chinese CATL plants run at 100GWh-plus and the largest US sites are heading toward 35GWh.

The strategic value is less about volume than about onshore supply, control over chemistry, and the option to extend the line if demand justifies.

Production schedules in the energy-storage sector have a track record of slipping. SoftBank's own framing has been transparent about that risk; the company has previously described the Sakai project as a multi-year build whose timing matters less than the existence of the optionality. The Sunday update tightens that timing without claiming it is firm.

Investors read the disclosure as long-term positive but an immediate non-event. SoftBank shares closed roughly flat in Tokyo on Monday, with the battery story doing less to move the stock than the ongoing OpenAI-margin-loan negotiations have.

Group debt remains the bigger conversation. Sakai is, for now, a hedge that buys SoftBank a manufactured commodity it would otherwise be paying spot prices for.