Japan's biggest power company is building its own gas plant just to feed one AI data center

JERA, Dell Technologies, and UK-based RHAELM signed a memorandum of understanding to build a $15 billion AI data center at JERA's Chiba Thermal Power Station near Tokyo. The project will draw up to 400 megawatts of dedicated power capacity, with operations targeted to begin in phases around 2028 and reach full capacity by 2029.

Why it matters: at $15 billion and 400 megawatts, this would be Japan's largest single-site AI infrastructure deployment, and among the largest anywhere in Asia. A deal this size signals serious confidence that AI computing demand in the region will keep growing for years.

Background: the project is structured as "behind-the-meter," meaning the data center draws power directly from JERA's own generation capacity rather than pulling from Japan's shared electrical grid.

In real life it's like building your own well instead of waiting in line at the municipal water department, you get water (or in this case, electricity) exactly when you need it, without depending on anyone else's queue.

Why it matters: getting a new facility connected to a national power grid can take years of approvals and infrastructure upgrades. RHAELM CEO Bradd Lewis said the behind-the-meter approach can "deliver a 400MW facility years ahead of a conventional grid-connected timeline," which is likely the entire point of designing it this way.

JERA supplies large-scale, round-the-clock power generation along with its existing LNG supply chain, covering procurement, shipping, and regasification. Dell Technologies provides standardized, rack-scale AI computing hardware through its Dell AI Factory platform. RHAELM leads overall project execution, tying the power and computing pieces together into a single deployment.

Why it matters: none of these three companies could easily pull this off alone. JERA knows power, Dell knows AI hardware, and RHAELM specializes in stitching large infrastructure projects together, a division of labor that mirrors how complicated building serious AI infrastructure has become.

Apollo Global Management is serving as the strategic investment and financing partner for the project.

Why it matters: a private equity giant backing the financing signals this isn't being treated as a speculative bet so much as a structured, long-term infrastructure investment, the kind of financing usually reserved for projects investors expect to generate predictable returns for decades.

JERA CEO Yukio Kani framed the move directly: "We can help AI infrastructure come online faster with access to the large-scale, reliable energy it needs."

Why it matters: AI data centers have become some of the largest new sources of electricity demand anywhere in the world. Rather than simply selling power to whoever builds a data center nearby, a utility like JERA capturing that demand by building the facility itself is a meaningful shift in how the AI boom's infrastructure is getting financed and built.

Background: this deal follows a wave of massive AI infrastructure investments worldwide, where energy companies and tech firms are increasingly pairing up to solve computing capacity and power supply together, rather than treating them as two separate problems handled by two separate industries.

Why it matters: that pairing reflects a hard lesson the AI industry has learned over the past couple of years: having enough chips means little if there isn't enough electricity to run them. Deals like this one are as much about solving an energy problem as a computing one.

Background: JERA was formed in 2015 as a 50-50 joint venture between Tokyo Electric Power Company (TEPCO) and Chubu Electric Power, two of Japan's biggest utilities. Over the following years, it gradually absorbed its parent companies' thermal power stations, fuel trading, and overseas power businesses, eventually taking over roughly 67 gigawatts of generation capacity.

Why it matters: that history matters because JERA isn't a startup experimenting with data centers as a side project. It's the product of a deliberate consolidation of Japan's thermal power industry, built specifically to operate at national scale, which is exactly the kind of infrastructure backbone a $15 billion, multi-decade AI project needs behind it.

By the numbers: data center investment tied to AI has been accelerating sharply worldwide, with hyperscalers and energy companies alike committing tens of billions of dollars to new computing capacity over the past two years. Power availability, not chip supply, has increasingly become the limiting factor on how fast that capacity can come online.

Why it matters: that shift in bottleneck explains why a utility is now a lead partner on a flagship AI infrastructure project rather than simply a vendor supplying electricity after the fact. When power becomes the scarce resource, the companies that control power generation gain real leverage in deciding what gets built and how fast.

Background: Dell AI Factory bundles standardized computing hardware, which is PowerEdge servers equipped with AI accelerator chips from Nvidia and AMD, along with high-performance storage and networking, into a single, repeatable package. Dell says more than 5,000 customers are already deploying it.

Why it matters: "standardized" is the key word. Rather than custom-engineering computing infrastructure for each new site, Dell is betting that a repeatable, pre-packaged setup lets a project like this one get built faster, since the computing side doesn't need to be reinvented for every new facility JERA and RHAELM develop together.

• If you use AI tools, infrastructure deals like this are part of what keeps up with growing demand, even though you'll never interact with the Chiba facility directly.
• Watch for more "behind-the-meter" data center deals globally, as developers increasingly bypass slow grid connection processes by building dedicated power supply alongside the computing facility itself.
• Power companies entering the data center business directly is a trend worth tracking, since it changes who controls AI infrastructure, not just who sells electricity to it.
• This signals continued, serious long-term investment in AI computing capacity in Asia, not a short-term bet, given the scale of financing and the multi-decade power agreement involved.