JERA, Dell and RHAELM Target National AI Infrastructure

Japan’s biggest power generator, JERA, is joining Dell Technologies and RHAELM to develop large-scale artificial intelligence infrastructure. The companies have signed a memorandum of understanding focused on creating a standardized development model in Japan. Their first project centers on Chiba and could involve more than $15 billion in total capital deployment. JERA plans to support the development with up to 400 megawatts of power capacity. The partnership combines power generation, data center development and AI computing within one infrastructure strategy. JERA contributes its energy capabilities, while Dell brings rack-scale AI infrastructure to the project. RHAELM will lead development and project delivery. Moreover, the partners aim to create a model that they can apply to additional locations across Japan.

Chiba Anchors the $15 Billion Development

The first development will sit at JERA’s Chiba Thermal Power Station. JERA and RHAELM are working together on the AI infrastructure project at the site. The companies expect total capital deployment across all phases to exceed $15 billion, or about 2.3 trillion yen. That amount covers land, power infrastructure, facility construction and AI compute. The project could access up to 400 MW of power capacity at the Chiba site. The partners expect operations to begin in phases around 2028. RHAELM says the project could become Japan’s largest single-site AI infrastructure deployment at the planned scale. It would also give the partners a major test site for their standardized infrastructure model.

The distinction between total capital deployment and construction spending matters. The $15 billion-plus estimate does not represent the cost of the data center buildings alone. It includes the broader infrastructure required to turn electricity and land into usable computing capacity. That wider scope shows how capital-intensive large AI deployments can become when power and compute develop together.

Behind-the-Meter Design Connects Power and Compute

The partners plan to develop the Chiba project using a behind-the-meter configuration. The approach will connect AI infrastructure more directly with JERA’s generation assets at the site. It could reduce dependence on the conventional process of securing an entirely new grid connection. The development therefore places power availability near the center of the project design. That approach changes how the partners can plan the physical infrastructure. Large AI facilities require electrical equipment, cooling systems, computing hardware and supporting infrastructure to work together. Each layer must reach operational readiness before customers can use the resulting compute capacity. The Chiba project aims to coordinate those layers from an earlier stage of development.

Power availability alone does not create usable AI capacity. Developers still need facilities, electrical distribution, cooling equipment and computing systems before workloads can run. The partners want to standardize how these components connect across future developments. That could reduce some of the site-specific engineering required for subsequent projects.

Dell Adds Rack-Scale AI Infrastructure

Dell’s role focuses on the computing layer of the development. Its rack-scale infrastructure will form part of the standardized architecture planned for the project. The partnership will connect that technology with JERA’s power capabilities and RHAELM’s development expertise. Together, the companies aim to create an infrastructure model that they can repeat at other sites. This approach differs from building every major infrastructure layer as a largely separate project. Instead, the partners want to coordinate power, facilities and compute within one development strategy. The Chiba project will provide the first large-scale opportunity to test that concept. Its execution could influence how the partners structure later developments in Japan.

RHAELM will lead project delivery and coordinate the infrastructure development. Its role includes connecting site and power resources with the computing platform required for AI workloads. The project therefore brings together an energy producer, a technology supplier and an infrastructure developer. Each partner controls a different part of the physical path from electricity to compute.

Apollo Plans Strategic Financing Role

Apollo Global Management intends to participate as a strategic investment and financing partner for RHAELM’s deployment. Its planned involvement adds a major capital provider to the development structure. The project requires funding across land, energy infrastructure, construction and computing equipment. Those components help explain why the overall capital requirement extends well beyond the facility itself. The partners have not described the $15 billion-plus figure as a single Apollo financing commitment. Instead, the estimate covers total capital deployment across several project components and phases. That distinction provides important context for the scale of Apollo’s planned role. It also prevents the overall development value from being confused with one financing package.

Meanwhile, the project connects technology procurement with energy infrastructure and long-term capital planning. JERA provides the energy platform, while Dell contributes the computing architecture. RHAELM brings development execution, with Apollo expected to support investment and financing. The structure links several infrastructure layers that AI developers often need to coordinate simultaneously.

JERA Sites Could Support Further Expansion

Chiba represents the first project under a potentially much larger development strategy. JERA and RHAELM plan to explore similar AI infrastructure deployments at other JERA sites. Their longer-term plans contemplate multi-gigawatt-scale capacity across Japan during the 2030s. Existing power-generation locations could provide potential sites for that expansion. Such locations can offer access to energy infrastructure that already serves large power requirements. They may also allow developers to plan computing capacity closer to the underlying source of electricity. However, every future project would still require its own technical, commercial and regulatory development work. The Chiba project will provide an early test of how effectively the model works at scale.

The companies also plan to examine whether they can use the standardized approach beyond Japan. Any wider expansion would depend on local power systems, infrastructure requirements and project economics. The Chiba development therefore serves both as a major project and a reference architecture. Its progress could determine how aggressively the partners pursue the model elsewhere.

JERA Puts Energy at the Center of AI Development

JERA’s participation gives the project a direct connection to Japan’s electricity infrastructure. The company is Japan’s largest power generator and operates extensive generation assets. Its role places energy supply alongside computing and construction from the beginning of development. That arrangement makes power strategy a core part of the AI infrastructure proposition. The Chiba project also highlights a broader change in AI infrastructure planning. Securing accelerators represents only one part of delivering large-scale computing capacity. Developers must also provide enough electricity, cooling and physical infrastructure to operate those systems. Failure to align those elements can limit how quickly installed equipment becomes usable.

For AI customers, that difference can become commercially significant. A project may have hardware commitments before every supporting infrastructure layer reaches operational readiness. Connecting power planning more closely with facility and compute development could help address that mismatch. Chiba will test the model at a scale large enough to provide meaningful evidence.

Power and Compute Converge in Japan’s AI Buildout

The strategic importance of the project reaches beyond its $15 billion-plus capital estimate. JERA, Dell and RHAELM are testing a development model that starts with access to substantial power capacity. They then plan the facility and computing architecture around that foundation. The approach treats electricity as an integral part of AI capacity rather than a separate supporting service. The planned 400 MW site will bring power, facilities, financing and rack-scale computing into one development structure. Phased operations are expected to start around 2028. The longer-term test will be whether the partners can reproduce the model across additional locations. Execution at Chiba will therefore matter as much as the project’s headline scale.

For Japan, the development could add a significant block of domestic AI computing infrastructure. It also offers a practical test of closer coordination between energy and digital infrastructure. The project’s value will depend on converting planned power into reliable, usable computing capacity. That makes Chiba an important infrastructure project to watch as Japan expands its AI capacity.