# Apollo Backs Japan’s $15 Billion AI Power Bet

> A Chiba project pairing Apollo, JERA, RHAELM, and Dell makes power access the organizing asset in Japan’s AI buildout.

Apollo Global Management plans to finance a more than $15 billion AI infrastructure project in Chiba Prefecture with developer RHAELM, power generator JERA, and Dell. The planned 400-megawatt campus makes Japan’s domestic compute ambitions inseparable from power supply, construction finance, and industrial policy.

- Section: AI & Compute Infrastructure
- Published: 2026-10-02T05:30:00Z
- By: AI-Assisted Editorial Staff
- HTML edition: https://culled.org/articles/apollo-japan-ai-infrastructure-chiba/

---

![An engineer overlooks a coastal power-and-data-center construction site at dusk, with cranes, transmission lines, and wind turbines](/images/articles/apollo-japan-ai-infrastructure-chiba-hero-1790918158745.webp)

Apollo Global Management plans to become a financing and investment partner for a more than $15 billion AI infrastructure venture in Chiba Prefecture. The project brings together U.K. developer RHAELM Holdings, JERA, Japan’s largest power generator, and Dell Technologies. Its planned operations date is 2028; its proposed power draw is as much as 400 megawatts.

That combination explains why the project matters. Japan is not simply proposing another data-center shell. It is trying to bind land, generation, construction, servers, and long-duration capital into one domestic-compute strategy.

## The scarce asset is the powered site

JERA says Japan could become Asia’s largest data-center demand center outside China, with power demand potentially reaching 10 gigawatts. For a developer, that is an opportunity. For the grid, it is a claim on generation and transmission that must be financed long before a customer runs its first model.

The Chiba site gains an unusual advantage from JERA’s nearby thermal-power station. Physical proximity does not remove permitting, interconnection, or fuel questions. It does change the project’s bargaining position: a power company is present at the beginning rather than added after a developer has designed a campus.

That is the same constraint visible in [the fight over who pays for AI grid upgrades](/articles/who-pays-for-ai-data-centers-senate-permitting-deal/). Compute capacity is not just a chip order. It is a financing structure around electrons delivered on a date certain.

![A coastal power-station switchyard and data-center building connected by dense high-voltage equipment](/images/articles/apollo-japan-ai-infrastructure-chiba-mid-1790918175932.webp)

## Private credit meets industrial policy

Tokyo has committed ¥102 trillion to AI and semiconductors as part of a broader manufacturing strategy. Apollo’s participation makes that policy legible to private capital: a global asset manager can finance the physical layer while local partners supply power-market knowledge and an operating foothold.

Apollo has already financed AI-related projects involving Nvidia, Broadcom, and Anthropic, while also expanding its Japan presence. The firm’s recent increase of a loan to SoftBank illustrates the broader pattern. The AI buildout increasingly depends on institutions that can underwrite assets whose cash flows will arrive years after construction begins.

> Japan’s AI strategy is becoming an infrastructure strategy: secure the megawatts, then finance the machines.

The risk is execution. A 400-megawatt site is a large initial commitment, and JERA has discussed a possible three-to-four-gigawatt expansion that could require roughly $150 billion. Those numbers are a roadmap, not delivered capacity. The market will test them through construction milestones, power contracts, and tenant commitments.

For now, the Chiba venture offers a clearer signal. Japan’s AI ambition is moving from semiconductor subsidies toward the harder business of building powered, financeable campuses where domestic demand can actually run.

---

Sources: Bloomberg reporting published by The Japan Times on the Apollo, RHAELM, JERA, and Dell Chiba project; prior Culled reporting on AI power constraints.
