The latest move landed Thursday. SoftBank Group's mobile unit plans to transform part of its factory in Sakai City, Osaka into one of Japan's biggest production lines for large-scale batteries, in an ambitious attempt to power its own AI data centers. SoftBank Corp. aims to bring that production online within the next five years. After SoftBank executives considered different purposes for the plant — including robotics manufacturing — they decided to pursue energy. Arcads The decision to manufacture batteries for self-supply rather than buying from external vendors is a significant escalation. It transforms SoftBank from a data center customer into a vertically integrated energy producer — controlling, in theory, the full stack from stored electrons to AI inference output.
The timing makes sense when you understand the scale of what SoftBank is building on the other side of the Pacific. In March, SoftBank and American Electric Power announced a $33.3 billion project anchored at the former Portsmouth Gaseous Diffusion Plant in Pike County, Ohio — once a Cold War-era hub for uranium enrichment, now being repurposed as the "PORTS Technology Campus." The centerpiece is a 10-gigawatt data center campus, with SoftBank's energy arm SB Energy planning to build 10 gigawatts of new power generation — including 9.2 gigawatts of natural gas — to feed both the local grid and the facility. AdStellar To put 10 gigawatts in context: that is enough electricity to power 7.5 million American homes. All of it dedicated to artificial intelligence.
SB Energy and AEP Ohio have struck a separate $4.2 billion deal to upgrade and expand electrical transmission infrastructure across Southern Ohio, with both parties claiming it will lower utility bills for local homes and businesses. AEP expects power to reach the site by 2029. Evergreen The project is framed as consistent with President Trump's Ratepayer Protection Pledge, which aims to shield consumers from energy price increases driven by surging AI infrastructure demand — a political cover that has helped SoftBank move quickly through federal approvals.
SoftBank has announced a "Portsmouth Consortium" of companies and financial institutions interested in participating, including Japanese firms Hitachi, Mitsubishi Electric, Toshiba, TDK, and Sumitomo Mitsui Banking, alongside US names including Bechtel, Morgan Stanley, Goldman Sachs, and JPMorgan. Evergreen When Goldman, Morgan Stanley, JPMorgan, and the Japanese industrial establishment are all in the same room for a single project, the signal is clear: this is not a vision document. This is a build.
The Ohio play connects directly to SoftBank's relationship with OpenAI through the Stargate program. SB Energy received a $1 billion equity investment from OpenAI and SoftBank to fund the deployment of renewable energy assets, and was selected by OpenAI to build and operate a 1.2-gigawatt data center site in Milam County, Texas, powered by new solar generation and battery storage. Search Engine Land Masayoshi Son, for his part, has described the broader infrastructure vision as potentially channeling up to $500 billion into such campuses — a number that requires a certain suspension of disbelief until you start totaling up the projects actually under construction.
The through-line connecting all of it is the power constraint. The bottleneck in AI buildout is no longer chips. It is power. eMarketer Every major hyperscaler is running into the same wall: you can order Nvidia GPUs, wait in the queue, take delivery — and then discover that your data center site doesn't have the grid capacity to run them. Permitting new power connections in the US can take years. Building new generation takes longer. SoftBank's answer to this problem — build the batteries, build the generation, lease the federal land, bring the consortium — is the most aggressive attempt yet to solve the power problem from the supply side rather than waiting for grid operators to catch up.
The Japan battery plant announcement adds a dimension the Ohio story alone doesn't capture: SoftBank is now building energy infrastructure for its own data centers in its home market simultaneously. The Sakai City factory pivot from potential robotics manufacturing to energy production reflects a deliberate choice by SoftBank's leadership about where the most critical constraint in its AI infrastructure plans actually sits. Arcads Batteries at this scale serve a dual function in data center operations — they provide backup power during grid interruptions and, increasingly, act as grid-scale storage that allows operators to charge during off-peak hours and discharge during demand spikes, lowering overall energy costs significantly.
For the startup and venture world watching all of this, the SoftBank energy infrastructure push carries a lesson that is worth absorbing. The AI investment thesis has moved decisively downstream from software and models into physical infrastructure — and the companies that control the energy feeding the compute are positioning themselves to extract value from every layer of the AI stack, regardless of which model wins or which cloud platform dominates. Masayoshi Son put it simply: "AI will transform every industry, and the PORTS Technology Campus will help deliver the next-generation infrastructure needed to unlock those breakthroughs." Incognito Worldwide
Son has been wrong before — spectacularly, publicly, expensively. WeWork. Katerra. The first Vision Fund. But on AI infrastructure, the direction of his conviction has been consistent and, so far, directionally correct. Battery factories in Osaka. A 10-gigawatt campus on a nuclear site in Ohio. A 1.2-gigawatt solar-powered data center in Texas. It is the kind of portfolio of bets that only makes sense if you believe AI compute demand will grow at a rate that makes today's energy constraints look trivially small in hindsight. Son clearly believes that. He's betting his balance sheet on it.
