SoftBank picks water over fire risk: Why the AI data center build is rewriting battery chemistry

SoftBank Corp. (TSE: 9434) has launched a Japan-based battery manufacturing business aimed at supplying the next-generation power infrastructure required to support escalating electricity demand from artificial intelligence workloads. The end-to-end operation will span cell development through energy storage system manufacturing, anchored at the former Sharp Corporation factory site in Sakai City, Osaka Prefecture, which SoftBank Corp. is converting into an AI data center complex. SoftBank Corp. plans to begin manufacturing battery cells and energy storage systems from the fiscal year ending March 31, 2028, with mass production targeted at gigawatt-hour scale by around FY2028 and annual domestic battery revenue of more than 100 billion yen by FY2030. SoftBank Corp. shares last traded around 220 yen on the Tokyo Stock Exchange, sitting in the lower half of a 52-week range of 206.5 to 247.9 yen, even as the parent SoftBank Group Corp. (TSE: 9984) has captured most of the recent AI-driven re-rating in Japanese technology equities.

Why is SoftBank Corp. entering gigawatt-hour battery manufacturing now and what does it signal about Japan’s industrial strategy?

The timing of SoftBank Corp.’s battery launch is not incidental. Japan was, two decades ago, the centre of gravity for lithium-ion innovation and commercialisation, with Sony Corporation pioneering commercial cells in 1991 and Panasonic Holdings Corporation later anchoring the global supply chain through its Tesla, Inc. partnership. That position has been comprehensively eroded. Chinese manufacturers led by Contemporary Amperex Technology Co., Ltd. and BYD Company Limited now control close to 70 percent of global electric vehicle battery installations, with Panasonic the only Japanese name remaining in the global top four. South Korean producers including LG Energy Solution, Samsung SDI and SK On are also losing share, running domestic plants at roughly half capacity as Chinese rivals operate near full utilisation.

SoftBank Corp.’s decision to re-enter battery manufacturing is therefore not a play to recapture the electric vehicle cell market against entrenched gigafactory incumbents. It is a deliberate sidestep into stationary energy storage tied to AI infrastructure, where the demand profile, safety requirements, and cost structure differ materially from automotive applications. The strategic logic is to build domestic capacity in a battery chemistry that does not require competing on lithium carbonate procurement, nickel supply chains, or the manufacturing scale curves that Contemporary Amperex Technology Co., Ltd. has spent a decade flattening.

The use of the former Sharp Corporation Sakai plant is itself a strategic signal. Sharp Corporation sold the loss-making liquid crystal display facility to SoftBank Corp. in 2024 after years of restructuring, and the site is being repurposed into what SoftBank Corp. is calling the Osaka Sakai AI Data Center. Attaching a battery manufacturing hub to that complex bundles the demand source and the supply source on the same industrial campus, a configuration that has obvious operational benefits but also reduces SoftBank Corp.’s exposure to the merchant grid storage market it has not yet proven it can serve at scale.

How does the zinc-halogen battery chemistry change the safety and supply chain calculus for AI data centers?

The cell technology at the centre of SoftBank Corp.’s announcement comes from its collaboration with COSMOS LAB, led by Chief Executive Officer Ju-Hyuk Lee. The chemistry pairs a halogen-based cathode with a zinc anode and uses pure water as the electrolyte rather than the flammable organic solvents that dominate lithium-ion cells. SoftBank Corp. is claiming this configuration as a world first among commercially deployed aqueous zinc-halogen cells to achieve energy density comparable to lithium-ion, as of May 11, 2026, based on its own research.

The safety argument is the more strategically important of the two claims. Lithium-ion thermal runaway has become a non-trivial concern for large-format stationary storage, and several high-profile battery fires at energy storage sites in the United States, Australia and South Korea over the past three years have hardened insurance underwriting and slowed regulatory approvals for new sites. For an operator planning to colocate hundreds of megawatt-hours of storage with high-value computing infrastructure, the fire risk profile of the storage chemistry directly affects siting flexibility, insurance cost, and grid interconnection timelines. An aqueous chemistry that cannot ignite by structural design is a meaningful differentiation for data center applications even if its raw cost per kilowatt-hour eventually settles above the cheapest Chinese lithium iron phosphate cells.

The supply chain argument is more defensive in nature. Halogen feedstocks and zinc are both materials that can be sourced domestically in Japan, removing exposure to the lithium, cobalt and graphite chokepoints that currently constrain non-Chinese cell makers. This matters less as a cost story than as a procurement resilience story, particularly given the deteriorating United States and China trade environment and Japan’s recent willingness to treat economic security as a battery industry policy variable. Whether the cost of mining and refining zinc and halogen at industrial scale in Japan can be made competitive remains an open execution question that the press release does not address.

What does the DeltaX partnership tell us about the energy storage system architecture and the energy density target SoftBank Corp. is chasing?

The second collaboration partner is DeltaX Co., Ltd., led by Chief Executive Officer Stephen Kim, which is supplying the energy storage system architecture. DeltaX brings two proprietary technologies: a Cell Connecting System design that interconnects multiple cells with reduced loss, and a Cell to Pack approach that eliminates intermediate modularisation and packages cells directly into the system enclosure. The combination has allowed DeltaX to reach 5.37 megawatt-hours of storage in a standard 20-foot containerised lithium-ion BESS unit, which SoftBank Corp. is using as the engineering benchmark for its zinc-halogen system.

Reaching comparable or higher density with a different cell chemistry is the central technical risk in the roadmap SoftBank Corp. has just laid out. Aqueous chemistries have historically traded energy density for safety, and pushing zinc-halogen cells to lithium-ion equivalent volumetric performance is a non-trivial materials science problem. SoftBank Corp. is essentially asking the market to believe two things at once: that COSMOS LAB can scale a chemistry that has not previously seen gigawatt-hour deployment, and that DeltaX’s packaging gains will transfer cleanly across cell formats. Neither claim is impossible, but both will need to be demonstrated before utility offtakers, industrial customers and grid operators commit to procurement contracts.

The integration with an energy management system equipped with an artificial intelligence-based power demand forecasting layer is the third element of the architecture. SoftBank Corp. has been operating an electricity business for several years and has accumulated forecasting capability that it can now bundle with the hardware. For data centre buyers in particular, an EMS that anticipates AI workload power draw and stages charge-discharge cycles accordingly offers value beyond the raw kilowatt-hour, and helps SoftBank Corp. defend pricing against pure-play cell competitors.

What are the financial and competitive implications for SoftBank Corp. as it commits capital to gigawatt-hour-scale manufacturing?

The 100 billion yen revenue target for FY2030 is modest in the context of SoftBank Corp.’s overall financial profile. The company recorded record revenue of 5,195 billion yen for the third quarter of fiscal year 2025, on a market capitalisation of approximately 10.55 trillion yen. A 100 billion yen battery business contributes a low single-digit percentage to total revenue at the target year, suggesting that the strategic value SoftBank Corp. sees in this initiative is less about direct revenue contribution and more about three adjacent benefits.

First, it secures internal supply for the AI data centre buildout, where SoftBank Corp. is competing with NTT, Inc., KDDI Corporation and a series of hyperscaler-backed projects for limited grid capacity in the Kanto and Kansai regions. Owning the storage stack reduces both procurement risk and the marginal cost of dispatching power into computing loads.

Second, it positions SoftBank Corp. as a participant in Japan’s evolving industrial policy on critical technologies. Tokyo has indicated through the Ministry of Economy, Trade and Industry that domestic battery manufacturing is a strategic priority, and government subsidy programmes for storage and supply chain resilience are likely to expand. A domestic gigawatt-hour-scale producer with a non-lithium chemistry fits the political brief better than a pure-import strategy.

Third, the medium-term option value of selling cells and systems into global grid, factory and residential markets is real but unproven. The global stationary storage market is projected to grow rapidly through 2030, and a differentiated safety-first chemistry could find buyers in jurisdictions that have tightened fire codes for lithium-ion installations. SoftBank Corp. has not, however, demonstrated a track record of exporting industrial hardware, and the gap between announcing a gigawatt-hour production target and converting it into deliverable international revenue is wide.

The competitive response is the more interesting question. Panasonic Holdings Corporation, Toyota Industries Corporation and GS Yuasa Corporation are the natural Japanese incumbents in stationary storage. None has so far made a comparably aggressive commitment to a non-lithium aqueous chemistry at scale, and SoftBank Corp.’s announcement effectively dares them to either accelerate their own next-generation chemistry roadmaps or cede the safety-differentiated segment to a telecommunications operator with no battery manufacturing heritage.

For the equity, the market reaction to date has been muted. SoftBank Corp. has been treated by investors primarily as a high-yield Japanese telecommunications name with optionality on artificial intelligence infrastructure, and the battery announcement does not immediately change that thesis. The stock’s 3.92 percent dividend yield and 18.8 price-to-earnings multiple remain the dominant valuation anchors. If SoftBank Corp. can demonstrate execution on the FY2027 mass production milestone with COSMOS LAB cells, the optionality argument becomes more defensible. If the timeline slips or the chemistry fails to scale at the claimed energy density, the initiative will be quietly folded into the AI data centre capex line and the storyline will revert to the telecommunications fundamentals.

What are the key takeaways from SoftBank Corp.’s gigawatt-hour battery business launch for investors, competitors and Japan’s industrial base?

• SoftBank Corp. (TSE: 9434) is committing to gigawatt-hour-scale battery manufacturing in Japan from FY2028, targeting more than 100 billion yen in domestic battery revenue by FY2030, anchored at the Osaka Sakai AI Data Center site.
• The strategic logic is vertical integration with SoftBank Corp.’s AI data centre buildout, not direct competition with Contemporary Amperex Technology Co., Ltd., BYD Company Limited or LG Energy Solution in the electric vehicle cell market.
• The cell chemistry is aqueous zinc-halogen, developed with COSMOS LAB, using pure water as the electrolyte and eliminating the thermal runaway risk that dominates lithium-ion stationary storage underwriting.
• Energy storage system architecture is supplied by DeltaX Co., Ltd., which has already reached 5.37 megawatt-hours per containerised unit using lithium-ion and is targeting comparable density with the new chemistry.
• An artificial intelligence-driven energy management system developed inside SoftBank Corp.’s electricity business will bundle with the hardware, offering differentiated value to data centre buyers facing volatile AI workload power demand.
• Halogen and zinc feedstocks can be sourced domestically in Japan, reducing exposure to lithium, cobalt and graphite supply chains that currently constrain non-Chinese cell makers.
• The 100 billion yen FY2030 revenue target is small relative to SoftBank Corp.’s 5,195 billion yen quarterly revenue base, indicating that the project is valued primarily as strategic optionality rather than as a near-term earnings driver.
• Execution risk centres on whether COSMOS LAB can scale aqueous zinc-halogen cells to gigawatt-hour production while maintaining lithium-ion equivalent energy density, a transition that has no precedent at this scale.
• The announcement implicitly challenges Panasonic Holdings Corporation, GS Yuasa Corporation and Toyota Industries Corporation to accelerate their own next-generation chemistry roadmaps or cede the safety-differentiated stationary storage segment.
• For Japan, the initiative represents a calculated bet that the country can re-enter battery manufacturing by reframing the contest around AI infrastructure storage and non-lithium chemistry rather than attempting to claw back lost ground in lithium-ion electric vehicle cells.