Japan has a problem familiar to every nation chasing the AI boom. Power. Land. Time. The usual bottlenecks that slow data center construction to a crawl now threaten to leave the country behind the United States and China in the race for compute capacity.

Enter JERA, the world’s largest liquefied natural gas trader and Japan’s biggest power generator. On October 1 the company signed a memorandum of understanding with Dell Technologies and UK-based AI infrastructure developer RHAELM. Their immediate target: a single site in Chiba, east of Tokyo. The price tag exceeds $15 billion. The capacity reaches 400 megawatts. Operations could begin in phases around 2028.

That project stands as only the opening move. According to reporting in the Financial Times, the broader ambition could command as much as $140 billion across multiple sites. JERA global chief executive Yukio Kani told the paper the partners aim to develop 3 to 4 gigawatts of AI data centers and associated gas power infrastructure within five years. One gigawatt, he estimated, costs between $35 billion and $45 billion. The arithmetic lands near that eye-popping total.

The approach sidesteps the grid entirely for the first facility. No waiting years for new transmission lines or regulatory approvals. The data center sits directly beside JERA’s existing Chiba Thermal Power Station. Power flows behind the meter under a 15- to 25-year agreement. Speed to power, as Kani put it. Years shaved off typical timelines.

JERA supplies the land and the electricity. RHAELM handles development, construction, operation and financing, with private equity giant Apollo Global Management acting as strategic investment and financing partner. Dell delivers standardized rack-scale AI infrastructure, its AI Factory servers forming the compute core. The trio wants to create a repeatable template. One that can roll out at other JERA power stations across Japan and, eventually, abroad.

This matters. Japan currently holds just 1.7 gigawatts of data center power capacity, according to figures cited by the Financial Times. The United States commands 29.2 gigawatts, or 43 percent of the global total. China sits at 8.5 gigawatts. Tokyo has set its own targets, including roughly $206 billion in public-private investment in cloud and data centers by 2035 under Prime Minister Sanae Takaichi’s strategic sector plan. The new partnership aligns with that goal while addressing the practical barriers.

Critics might question the carbon footprint. JERA’s thermal plants burn gas. The country’s nuclear restarts move slowly. Yet the immediate demand for AI training and inference leaves little room for perfect solutions. JERA generates about one-third of Japan’s electricity. Its integrated LNG supply chain, from procurement to regasification, offers reliability that renewable-only builds cannot yet match at this scale.

The model carries risks. Construction at $15 billion for 400 megawatts implies enormous capital intensity. Land costs near Tokyo run high. Equipment prices for the latest GPUs and associated cooling fluctuate. And the $140 billion headline remains an extrapolation, not a firm commitment. JERA’s official release speaks of multi-gigawatt ambitions in the 2030s without attaching that precise figure.

Still, the structure shows sophistication. By co-locating with existing generation assets, the partners avoid one of the longest lead times in the industry. They standardize the integration of power infrastructure, cooling, and compute. They bring together expertise from energy trading, server manufacturing, and specialized data center development. The result could serve as blueprint beyond Japan’s borders.

Dell stands to gain significantly. The company has seen explosive demand for its AI servers. Placing its rack-scale solutions as the standard compute layer in what could become Asia’s largest single-site AI project outside China gives it a strong foothold in a market hungry for Western technology partnerships.

RHAELM, formerly known as Sovereign AI in some references, brings operational know-how in building and running these facilities at hyperscale. Apollo’s involvement signals confidence from sophisticated capital allocators who understand both the returns and the execution challenges of such projects.

Recent coverage underscores the momentum. A Reuters dispatch detailed the memorandum and the Chiba project’s specifics, noting how the facility would rank as Japan’s largest single-site AI infrastructure deployment. Channel Insider highlighted the potential $140 billion total while stressing that JERA’s own statements focus on the standardized model rather than a fixed investment sum.

Japan’s existing players have not stood still. NTT Data and SoftBank operate significant facilities. Blackstone has signaled a $30 billion push into the country. Yet the gap with global leaders remains wide. This initiative, built on behind-the-meter power and repeatable design, attempts to close that gap faster than conventional approaches allow.

Executives emphasize the integration. One described the collaboration as stretching from the molecule, in the form of LNG, all the way to the AI token output. Ambitious language. But it captures the holistic thinking required when power constraints dominate every conversation about AI expansion.

Challenges persist. Environmental opposition to new gas capacity could surface. Supply chain issues for specialized cooling and power equipment might delay phases. Global competition for chips and talent will not ease. Yet the partners have moved with unusual speed. The memorandum was signed and announced in coordinated fashion across companies and time zones.

What comes next will test the model. Detailed engineering for Chiba. Securing the full financing stack. Delivering the first 400 megawatts on something close to the 2028 schedule. Success there would validate the template and open the door to replication at JERA’s other sites. Failure, or even significant delays, would cast doubt on the $140 billion vision.

Either way, the announcement marks a serious attempt by Japan to treat AI infrastructure as a matter of national industrial policy. Not just another data center build. A coordinated effort that pairs the country’s energy assets with global technology leaders to secure a larger slice of the AI compute pie. The coming years will reveal whether that bet pays off.