SoftBank is moving into battery manufacturing because AI data centers are becoming power projects as much as computing projects. The bet is simple: whoever controls energy storage may control more of the AI buildout.
SoftBank is no longer treating AI infrastructure as a matter of chips, servers and software alone. The Japanese group has now moved from planning into execution, launching a battery business aimed at supporting the data centers behind Masayoshi Son's wider AI ambitions.
As The Wall Street Journal reported on Monday, SoftBank Corp., the group's mobile unit, will work with South Korea's Cosmos Lab and AI firm DeltaX Co. to produce large-scale battery cells and storage systems at its Sakai facility in Osaka. Battery production is expected to begin in the fiscal year ending March 2028, with mass production targeted for the following year. That is a sharper timeline than earlier reports that described production within the next five years.
That matters because the next AI bottleneck is not only Nvidia GPUs or advanced packaging capacity. It is the ability to secure reliable electricity, smooth grid demand, provide backup power and keep enormous computing sites running without interruption. Batteries sit directly in that pressure point. They can help data centers manage peaks, support on-site generation and reduce dependence on strained utility grids.
For years, the AI investment story was dominated by model labs, cloud providers and chipmakers. That was understandable. OpenAI, Anthropic, Nvidia and Arm gave investors obvious ways to think about where value might accrue. But the physical side of AI is getting harder to ignore. A powerful model is useless if the data center cannot get enough electricity to train or serve it at scale.
This is why SoftBank's battery move should be read as part of a broader infrastructure strategy, not as a side project. Son has already tied SoftBank's future to AI through Arm, OpenAI investments, data-center ambitions and large industrial proposals in the United States. If battery manufacturing becomes part of that same operating system, SoftBank is trying to own more of the chain that turns capital into compute.
The specific details now look more concrete. SoftBank is working with Cosmos Lab on next-generation zinc-halogen battery cells and with DeltaX on high-performance, high-energy-density battery designs. The company is targeting more than 100 billion yen, or about $638 million, in annual revenue from the battery business by fiscal 2030. The batteries are expected to support AI data centers first, with possible uses in grid and industrial systems in Japan and overseas later.
Still, execution remains the hard part. Building batteries at data-center scale is not the same as announcing interest in the category. It requires supply contracts, chemistry decisions, safety systems, grid integration and enough demand visibility to justify the capital spend. SoftBank has more clarity than it had a few weeks ago, but the commercial proof will come when named customers, capacity targets and deployments start to appear.
Still, the direction is clear. AI data centers increasingly need energy storage for both resilience and speed. Waiting years for grid upgrades can slow a project before the first server rack is installed. Pairing batteries with on-site generation, renewables or utility connections gives operators more flexibility. It also changes which suppliers become strategically important. Battery integrators, power electronics firms, grid software providers and industrial real estate owners all move closer to the center of the AI economy.
Vertical Integration Is Becoming A Test
SoftBank's approach also raises a sharper question: is vertical integration becoming the line between serious AI infrastructure players and hype-driven capital spenders? It is one thing to announce a giant AI campus. It is another to secure chips, land, fiber, cooling, power generation and energy storage in a coherent plan. The more constrained the system becomes, the more valuable coordination becomes.
That is where Son's style cuts both ways. SoftBank has a history of making enormous directional bets before the market has fully formed. Sometimes that has produced spectacular gains, as with Alibaba and Arm. Sometimes it has exposed the company to painful misjudgments, as WeWork showed. Batteries for AI data centers sit somewhere between those two histories. The demand case is logical, but execution will decide whether this becomes infrastructure leverage or another expensive expression of conviction.
For founders, this is the useful lesson. AI infrastructure is no longer just a cloud resale opportunity or a model wrapper opportunity. The value chain is widening. Companies that solve boring physical problems, from battery management to thermal design to substation interconnection, may find themselves serving some of the highest-growth customers in technology.
For investors, the signal is similar. Capital is moving into the layers beneath AI applications because those layers determine how fast the entire market can grow. If power availability becomes the binding constraint, energy storage will not be a peripheral clean-tech theme. It will become a core component of AI capacity planning.
The next thing to watch is whether SoftBank turns this plan into a real supplier network with measurable production capacity and live data-center deployments. A factory conversion in Sakai is only the beginning. If Son can connect batteries to his wider AI buildout, SoftBank will be making a larger argument: the winners in AI will not only own intelligence, they will own the infrastructure that keeps it running.
Also read: ChatGPT is becoming the life advisor young users already expect • Alphabet turns to yen bonds as AI becomes a balance sheet race • JPMorgan says Korea's AI memory rally can push the Kospi to 10,000
