You can't build modern artificial intelligence infrastructure without running into the world's most stubborn bottleneck: electricity. While tech companies scramble for compute power, energy grids are buckling under the strain. Japan is taking a radically different approach to solve this crisis. Instead of waiting for public utilities to upgrade outdated electrical grids, the country is teaming up with Dell and energy giant Jera to build massive AI data centres directly next to power plants.
This joint push targets a staggering $140 billion in infrastructure investments across the next few years. It marks a dramatic shift in how industrial economies plan to handle the energy demands of machine learning models. If you want to understand where enterprise computing is heading, you have to look closely at this Japanese model because conventional grid planning is officially dead.
The Chiba Blueprint and Direct Power Supply
The partnership kicks off with a massive 400-megawatt project situated right next to Jera's thermal power station in Chiba. Estimated at roughly $15 billion for its initial phases, this campus is designed to bypass standard transmission headaches entirely.
Most data centres fail or face years of bureaucratic delays because they try to plug into overloaded local power grids. The Chiba campus takes a smarter route. It hooks up straight to the power generation source. Rhaelm is developing the site, Dell is providing rack-scale computing architecture, and Apollo Global Management is backing the financing. Operations are slated to begin around 2028.
By keeping the generator and the compute cluster side-by-side, the partners eliminate the friction that usually bogs down large tech developments. You don't have to guess when the local utility will approve new transmission lines if the electricity is already right outside your server room wall.
Scaling Up to Multigigawatt Ambitions
Japan isn't stopping at a single site. The long-term vision encompasses rolling out this integrated framework across multiple Jera locations, scaling toward several gigawatts of total capacity through the 2030s. When you tally up the cost of building out several gigawatts of data centers along with dedicated gas and electrical infrastructure, projections point toward a massive $140 billion capital injection.
This is a stark contrast to Japan's historical posture in the data infrastructure race. For years, the country lagged far behind the United States in raw data center capacity. While US clusters devoured tens of gigawatts, Japanese infrastructure remained modest. That gap is closing fast. Prime Minister-backed initiatives have set aggressive targets for public and private cloud investments, pushing toward trillions of yen by 2035.
Why Co-Location Changes Everything for Investors
If you're watching enterprise technology or energy markets, this blueprint solves the exact problem that plagues Western projects. In many regions, developers buy land first, sign massive hyperscale tenants, and only later discover that local substations or gas pipelines are tapped out. That mismatch creates massive financial exposure.
By locking down the power supplier, the land, the hardware provider, and the financial sponsor before laying down heavy capital, the Jera-Dell-Rhaelm alliance flips the script. You reduce permitting chaos and regulatory friction. You also secure predictable long-term energy pricing, which matters when your server racks are drawing megawatts around the clock.
Other nations will likely copy this playbook. When power generation and computing hardware merge into a single localized campus, everyone wins except the traditional grid bottleneck. Keep an eye on how these 2028 timelines hold up. If Chiba succeeds, expect direct-to-source data centres to become the gold standard for global AI expansion.
