Japan Could Be the Next Growth Engine for Singapore’s Data Centre REITs — But There Is a Catch

For years, the data centre investment thesis was relatively straightforward.

More cloud computing. More AI. More servers. More demand for data centres.

For Singapore-listed data centre REITs expanding into Japan, however, the scarce commodity is increasingly not the data centre itself. It is electricity.

That changes the investment thesis.

Keppel DC REIT’s proposed Tokyo acquisitions and Digital Core REIT’s increased exposure to Osaka are not simply bets on Japanese demand for computing capacity. They are increasingly bets on the value of existing, operational assets with secured power.

And that distinction could determine which REITs create attractive returns over the next decade and which merely buy expensive exposure to a fashionable theme.

Japan has more than 1.8 GW of operational data centre capacity, a relatively low vacancy rate and substantial additional development planned. AI, cloud adoption and enterprise workloads provide powerful structural demand drivers.

But the same AI boom that is creating demand is also making data centres dramatically more electricity-intensive.

That creates a fascinating investment paradox:

The shortage of power can increase the value of existing data centres while simultaneously making it harder for REITs to create new ones.

For investors, this means Japan may offer data centre REITs a long runway for growth—but not necessarily at any price.

The most important asset in a Japanese data centre may be invisible

Investors normally look at a data centre and ask:

• How much capacity does it have?
• Who are the tenants?
• How long are the leases?
• What is the occupancy rate?
• What is the rental growth?

Those remain important.

But AI is introducing another question:

How much reliable power can the site actually access?

A building can theoretically be expanded.

A tenant can be replaced.

Rent can be renegotiated.

But securing additional high-voltage grid capacity can take years.

That makes power access a potential barrier to entry.

It also means two physically similar data centres can have very different economic values depending on their ability to secure future electricity supply.

This is particularly significant around Greater Tokyo, where grid constraints are becoming an increasingly important consideration.

The implication for REIT investors is subtle.

The value of a data centre may increasingly depend not just on its existing rental income, but on its future optionality.

An asset with available power and expansion potential could command a premium even before that capacity generates additional revenue.

Why Japan is becoming strategically important to Singapore REITs

Japan offers something that many emerging data centre markets cannot easily provide: scale and maturity.

The market already has a substantial operational base, established hyperscale customers and a growing ecosystem of cloud and enterprise users.

Hyperscalers account for an estimated 60–70 per cent of Japanese data centre capacity.

That is important because hyperscale customers can provide significant long-term demand visibility.

Japan also offers geographic diversification.

For Singapore-listed REITs, expanding into Japan reduces dependence on the relatively constrained domestic Singapore market.

This is particularly relevant because Singapore has its own limitations around land, energy and data centre development.

Japan therefore offers an attractive combination:

large market + established demand + institutional-quality infrastructure + potential AI-driven growth.

But there is a catch.

The easiest assets to buy are increasingly the assets that everybody else wants.

The market may be shifting from a construction story to a scarcity story

This could be the most important change investors are missing.

The first phase of the data centre boom was fundamentally a development story.

Developers could acquire land, obtain power, build facilities and lease them to hyperscalers.

As AI demand accelerates, the bottleneck moves upstream.

Power becomes harder to secure.

Construction timelines lengthen.

Grid infrastructure needs upgrading.

Existing facilities become more valuable because they are already connected and operational.

That shifts the economics towards scarcity rents.

If an investor owns a fully operational data centre with long-term access to electricity while competitors struggle to obtain grid capacity, the asset could become more strategically valuable.

This helps explain why investors may accept lower initial yields for high-quality operational assets.

They are not necessarily buying today’s income alone.

They are paying for certainty in a market where future supply is becoming harder to create.

But scarcity can be both an opportunity and a warning

This is where investors need to be careful.

Scarcity is good for existing asset owners.

It is not automatically good for REIT investors buying those assets at elevated valuations.

Imagine two scenarios.

In the first, power scarcity allows landlords to raise rents and asset values continue appreciating.

In the second, power scarcity prevents new supply, but high acquisition prices and financing costs consume much of the potential return.

Both scenarios can coexist.

That is why investors should separate the question:

“Is the Japanese data centre market attractive?”

from:

“Can this particular REIT buy Japanese data centres at prices that generate attractive returns for unitholders?”

The first question increasingly looks like yes.

The second is much harder.

AI could make older data centres less valuable

There is another second-order risk.

AI does not simply require more data centre capacity.

It requires different data centre capacity.

AI workloads can involve much higher power densities than traditional workloads.

That means an older facility may have plenty of physical space but insufficient electrical infrastructure, cooling capability or rack density to support next-generation AI deployments economically.

In other words:

“Existing data centre” does not automatically mean “AI-ready data centre.”

This could create a bifurcation within the market.

High-quality, power-rich facilities may become increasingly valuable.

Older facilities requiring substantial electrical and cooling upgrades could face higher capital expenditure requirements.

For REIT investors, that means headline occupancy may become less informative.

A 100 per cent occupied facility with limited ability to support higher-density workloads may ultimately be less attractive than a slightly less occupied asset with substantial expansion and power potential.

Tokyo versus Osaka: geography is becoming an investment decision

Japan should not necessarily be treated as one homogeneous data centre market.

Greater Tokyo remains the dominant hub, but power constraints are becoming an increasingly important limitation.

Osaka offers a potentially different proposition.

Its comparatively better power availability has supported additional development, with more supply expected from 2027.

That creates a trade-off.

Tokyo

Potential advantages

• Deepest demand
• Major hyperscaler presence
• Mature ecosystem
• Scarcity of power-ready facilities

Potential disadvantages

• Higher competition
• Greater power constraints
• Potentially higher asset pricing
• More difficult expansion

Osaka

Potential advantages

• Better relative power availability
• Development runway
• Growing hyperscale demand
• Potential for future supply growth

Potential disadvantages

• More incoming supply
• Potentially less scarcity-driven pricing
• Greater competition as new capacity comes online

This suggests an important portfolio question for Singapore REIT investors:

Should a REIT maximise exposure to the most supply-constrained market, or seek markets where it can actually expand?

There is no universal answer.

But the distinction could materially affect long-term growth.

The sponsor advantage is real—but investors should not overpay for it

The growing importance of sponsor-to-REIT transactions is another feature worth watching.

Sponsors can develop or acquire assets, stabilise them and eventually sell them to their associated REITs.

This can provide a powerful pipeline.

For a specialised data centre REIT, that is valuable because finding suitable assets independently can be difficult.

But there is an obvious conflict investors should recognise.

A sponsor pipeline is an opportunity, not a guarantee of value creation.

The critical question is not:

“Does the sponsor have more assets?”

It is:

“At what price can the REIT acquire them, and what return will those assets generate?”

If scarce data centre assets become increasingly expensive, a strong sponsor pipeline could actually create pressure to deploy capital rather than preserve capital discipline.

That distinction is crucial.

Growth is not automatically accretive.

The CapitaLand Ascendas REIT example matters

This is particularly relevant to investors in diversified Singapore REITs.

CapitaLand Ascendas REIT and a Mitsui-managed fund’s acquisition of a 45.9 MW Greater Osaka data centre illustrates that data centre exposure is no longer restricted to specialist data centre REITs.

That changes the competitive landscape.

Specialist REITs such as Keppel DC REIT and Digital Core REIT have obvious sector expertise.

But diversified industrial REITs can also compete for the same assets.

That matters because competition for high-quality facilities can compress acquisition yields.

In effect, data centres are becoming an institutional asset class rather than simply a niche REIT category.

More capital chasing the same limited pool of power-ready assets is positive for existing owners—but potentially negative for future acquisition returns.

The bull case: Japan could provide a long runway for DPU growth

The bullish thesis rests on several structural factors.

AI demand is still in its early innings

The global expansion of AI infrastructure requires enormous computing capacity.

Japan is strategically positioned to serve domestic enterprises, cloud providers and hyperscale demand.

Power scarcity can strengthen incumbent assets

Owners of operational facilities with secured power may enjoy stronger pricing and lower competitive pressure.

Supply takes time

Japan has an estimated 3.5 GW of additional data centre pipeline through 2030, but a meaningful portion of capacity is already pre-leased.

That suggests demand is arriving before all future supply is available.

REITs can recycle capital

Specialist REITs can potentially acquire, stabilise and eventually recycle mature assets while using sponsor pipelines or third-party acquisitions to replenish growth.

If executed well, that can support recurring distribution growth.

The bear case: investors may be paying tomorrow’s growth today

The biggest risk is valuation.

If everyone recognises that power-ready data centres are scarce, investors will compete aggressively for them.

The result can be capitalisation-rate compression.

That sounds positive because it raises asset values.

But it creates a problem for future buyers.

Suppose a REIT owns an asset purchased years ago at an attractive yield. Rising scarcity pushes its valuation higher.

Excellent.

But when the REIT buys another asset at today’s lower yield, the new acquisition may contribute much less incremental income relative to the capital deployed.

Eventually, the REIT can have excellent assets but mediocre incremental returns.

This is one of the biggest risks in thematic investing:

A great industry can still produce poor investment returns if investors pay too much for exposure to it.

Interest rates could expose the difference between a good asset and a good REIT

Data centre REITs are still REITs.

That means investors cannot analyse them purely through the AI growth story.

Higher interest rates increase borrowing costs and can reduce the relative attractiveness of REIT yields.

Japan’s financing environment has historically been supportive in some respects, particularly for yen-denominated borrowing, but investors should not assume that financing will remain permanently cheap.

A REIT acquiring assets at aggressive valuations while relying heavily on debt could find that rising financing costs absorb a meaningful portion of rental growth.

The correct metric is therefore not simply:

data centre demand growth.

It is:

data centre demand growth minus acquisition pricing, financing costs and required capital expenditure.

That is the spread that ultimately matters to unitholders.

What investors should monitor instead of simply watching AI demand

The next 12–24 months could reveal whether the Japanese data centre thesis is genuinely creating shareholder value.

Investors should track at least six things.

1. Acquisition yields

Are REITs still able to buy assets at yields that exceed their cost of capital?

If that spread disappears, expansion can become value-destructive.

2. DPU accretion

A larger portfolio is meaningless if distributions per unit do not rise.

Investors should focus on per-unit economics, not asset growth alone.

3. Power availability

This may become one of the most important qualitative metrics in the sector.

How much secured power does an asset have?

Can capacity be expanded?

How long will additional grid connections take?

4. AI-related capex

Are older assets requiring substantial investment to handle higher-density workloads?

A facility’s headline occupancy may conceal significant future capital requirements.

5. Lease structures

Long leases provide income visibility, but investors should also examine escalation mechanisms, renewal terms and tenant concentration.

Strong occupancy is only valuable if the underlying economics remain attractive.

6. Balance-sheet capacity

REITs need financial flexibility to take advantage of opportunities when markets become distressed.

The most valuable balance sheet may ultimately be the one that can wait.

The investment conclusion: Japan is attractive, but selectivity matters more than ever

Japan is increasingly becoming an important growth market for Singapore-listed data centre REITs.

The structural case is compelling.

Cloud adoption is increasing. AI workloads are accelerating. Hyperscalers require more capacity. Japan has a large established market and substantial additional demand potential.

But investors should avoid reducing the thesis to:

AI demand → more data centres → higher REIT returns.

The missing variable is power.

And the second missing variable is price.

Power scarcity could create a valuable moat around existing data centres. Yet that same scarcity can encourage investors to pay premium prices for power-ready assets, reducing future acquisition returns.

This means the next phase of the Japanese data centre boom may favour asset quality and capital discipline over simple portfolio expansion.

For investors, the most attractive REITs may not necessarily be those announcing the largest acquisitions.

They may be the ones that can demonstrate three things:

secured power, sustainable rental growth and disciplined acquisition pricing.

Over the next 12–24 months, watch whether Japanese acquisitions actually translate into DPU growth and attractive returns on incremental capital.

If they do, Japan could become a meaningful long-term growth engine for Singapore’s data centre REIT sector.

If acquisition yields continue falling while financing and AI-related upgrade costs rise, however, investors could discover an uncomfortable truth:

The data centre shortage can be fantastic for data centre owners without necessarily being fantastic for the investors buying those owners at peak prices.

That is why the key investment question is no longer whether Japan needs more data centres.

It is whether Singapore REITs can acquire the right power-ready assets at the right price.