Japan’s Data Centre Market After STTGDC’s ¥93bn Loan

Japan’s next data centre investment cycle will be decided by the ability to match a specific customer with a credible power and delivery plan. Capital is still flowing into established Tokyo locations, even as developers look elsewhere for electricity. Those movements can reinforce each other: they serve different workloads and transfer different risks.

The immediate news is STTGDC’s 27 October 2026 announcement of a ¥93 billion green loan for the fit-out and subsequent operation of its TYO1 and TYO2 campus in Inzai, Greater Tokyo. The operator describes 70.8 MW of combined IT capacity across the two buildings. STTGDC’s announcement

For investors, operators and prospective customers, the implication is more useful than the headline amount. An established location can continue attracting capital when the financing supports a defined asset and a route to customer service. A cheaper or more power-rich site elsewhere must demonstrate why a customer would move there, and whether the resulting operating advantage survives the cost of connectivity and delivery.

My reading is that Japan is developing a more differentiated infrastructure market. The most attractive project will depend on which constraint a buyer is paying to remove.

Financing strengthens a location; customers determine its value

The financing announcement identifies TYO1 as operational and TYO2 as under development. It also sets environmental eligibility criteria, including a design PUE of 1.3 or below and an agreed green building certification. It does not disclose borrowing margins, utilisation or customer-level availability.

That distinction leaves room for a commercial interpretation without pretending to know the economics of the transaction. For a developer, access to funding can allow equipment procurement and fit-out to progress with greater confidence. For a tenant, however, the value still sits in the block that can be accepted under its technical requirements and contractual timetable.

An established campus may offer a useful combination of local operating experience, supplier access and proximity to customers’ existing infrastructure. Those advantages are worth paying for when relocation would introduce integration work or delay. They are less decisive for a workload whose data movement can be planned and whose economics are dominated by sustained electricity consumption.

This also explains why more financing need not immediately produce lower market rents. A funded expansion can be absorbed by existing customer growth, reserved for a particular configuration or commissioned in stages. The effect on competitive pricing depends on how much serviceable capacity is offered to new customers, on what terms and when. The next commercially meaningful evidence is the leasing and delivery progression.

Japan’s geography is changing by workload

The broader constraint is documented independently. CBRE’s March 2026 analysis identifies high-voltage grid access as a major obstacle in Greater Tokyo and, to a lesser extent, Greater Osaka. It describes increasing interest in regional locations and alternative power arrangements.

JLL’s Japanese market analysis, drawing on November 2025 research, presents Hokkaido and Kyushu as complementary regional hubs. That is useful structural context; it is not a current availability schedule for any individual site.

The commercial question is how far a workload can move before the benefits of electricity access are offset by other costs. An enterprise application tied to a dense local ecosystem has a different location tolerance from a training cluster with planned transfers of large datasets. Even within AI, interactive inference and batch processing need not share the same siting logic.

A national forecast can therefore conceal two simultaneous outcomes: persistent competition for suitable metropolitan capacity and investment in new locations with a different demand profile. Regional growth does not automatically weaken Tokyo. It can take workloads that would otherwise struggle to find a viable large-scale power path.

The developer’s task is to identify the relevant demand before committing the full campus. A site that is inexpensive on paper can become expensive if commercial demand arrives slowly or needs substantial network integration.

Power access creates a second route to scale

JERA, Dell Technologies and RHAELM offer a contrasting approach. Their 1 October 2026 memorandum of understanding describes an integrated development model, with the first application at Chiba. JERA reports up to 400 MW of power capacity, operations targeted around 2028, and more than US$15 billion of expected capital deployment across land, power infrastructure, facilities and AI compute. The project is intended to use generation behind the meter at JERA’s thermal power station. JERA’s release

This remains a development proposition. The power figure is not published IT load, and the capital scope includes compute. It cannot be compared directly with a campus fit-out loan.

The strategic opportunity is to bring the energy and compute delivery schedules under a more coordinated arrangement. If that produces a credible service date for an anchor customer, a project can compete on the economic value of earlier deployment as well as its running cost.

The corresponding exposure becomes more concentrated. A project tied to generation must resolve fuel-price allocation, maintenance coordination, backup arrangements, emissions expectations and responsibility for interrupted supply. Faster access to electricity has value only if the customer can accept the resulting service and commercial terms.

This model also shows why the distinction is not simply “Tokyo versus the regions”. Chiba sits within the wider Tokyo economic area. A project can remain close to a major demand centre while changing the way power is secured. Location and energy procurement are related choices, but they are not the same choice.

What the latest PowerlandMap review adds

PowerlandMap’s Japan records were checked against the public operator material on 2 November 2026. The comparison below brings together three relevant propositions. The Osaka OS5x record was refreshed on 1 November; this is a database verification date, not a new project announcement.

The Equinix expansion timeline keeps Osaka relevant to near-term procurement discussions. It supports a prospective opening window, not a conclusion about vacant capacity or a guaranteed handover date.

The point of this comparison is to identify competition at the level of the customer decision. A buyer needing service soon may evaluate metropolitan expansion. A customer seeking a much larger future deployment may be able to accept a longer development programme. Treating both as one pool of Japanese megawatts would hide that distinction.

An electricity advantage must cover the cost of moving

A simple sensitivity helps explain when a location change becomes attractive. Consider a hypothetical 20 MW IT installation, operating at an average 80% IT load, with an assumed annual PUE of 1.30. Its modelled annual facility electricity use is:

20 × 1,000 × 80% × 1.30 × 8,760 = 182.208 million kWh.

This is an illustrative load model, not an estimate of any project above. The price differences below are assumptions, not Japanese tariff benchmarks.

At the middle case, a ¥3/kWh advantage creates approximately ¥546.6 million of annual gross electricity savings. If the alternative location adds ¥300 million per year in network, staffing and other recurring costs, the remaining advantage is about ¥246.6 million per year, before migration costs, financing, taxes and any service-performance effect.

Those additional costs are also hypothetical. The calculation holds PUE constant across load levels for transparency; actual efficiency varies. It excludes demand charges and other tariff components unless already included in the compared effective energy prices. It does not value earlier revenue, and it does not treat gross savings as profit.

The useful conclusion is that utilisation changes the value of location. An electricity advantage matters much more when the installation carries a sustained load. During a slow customer ramp, the same infrastructure can produce a smaller saving while retaining much of its fixed cost. A developer should therefore connect its power strategy to a credible demand ramp, rather than assuming that a favourable energy position will fill the building.

The counterargument: existing hubs can remain the better investment

It is tempting to read power constraints as a reason to move every large project. That conclusion is too broad. Existing metropolitan demand may be more valuable than an unproven regional cost advantage, particularly when customers need proximity, established connectivity or an earlier deployment.

For the same reason, a generation-adjacent development should not be dismissed because its model is unfamiliar. It can be the better option when an anchor customer accepts the location and power structure, the delivery sequence is credible, and the total cost advantage survives realistic network and utilisation assumptions.

The evidence that would change my view is commercial: a binding customer ramp, a documented energisation and acceptance sequence, and operating terms that allocate the remaining risks clearly. Those milestones matter more than an impressive end-state footprint.

For Japan, the next phase should reward operators that can combine location, electricity and customer demand into a coherent service. STTGDC’s financing shows continued capital commitment to an established market. The emerging power-led alternatives put pressure on incumbents to explain the value of their location and on challengers to demonstrate demand. Both can succeed, but on different underwriting assumptions.

Sources, scope and next step

This analysis uses public announcements and operator pages read on 2 November 2026, alongside PowerlandMap records for the named projects. It is a selected case comparison, not a national census, vacancy estimate or ranking of investment returns. The STTGDC release is dated 27 October; the JERA release is dated 1 October. Earlier research is identified as context. No private customer information is used.

The numerical sensitivity is PowerlandMap’s own illustrative calculation. All input assumptions are stated, and no capacity figures from the three cases are added together. The cover is an original conceptual illustration, not a photograph of a named facility.

For teams moving from market screening to operator discussions, the Global Data Center Capacity Availability study provides a framework for separating current offers, future supply and contracted capacity across selected markets. Its public sample and evidence dates are available on the study page; it should not be read as a guarantee of today’s Japanese availability.

— Matthieu Gallego, Founder, PowerlandMap