Singapore’s artificial-intelligence boom is beginning to show up in a place investors may not immediately associate with AI: the office market.
OpenAI is reportedly in talks to lease about 100,000 square feet at Shaw Tower. Anthropic is establishing a Singapore office. Mistral AI has taken dedicated space at Asia Square, while other AI and technology companies including Databricks, Google DeepMind and Sierra are expanding their presence.
The timing matters.
Singapore’s prime office market was already tight before the latest AI expansion. CBRE reported Core CBD Grade A vacancy of just 3.3% in the second quarter of 2026, with rents rising 0.8% quarter-on-quarter to S$12.50 per square foot per month. It was the sixth consecutive quarter of rental growth. CBRE also said there were no significant new completions expected through 2027.
The Financial Times now reports that vacancy for the most sought-after office space has fallen to about 3.1%, with market participants expecting rents could rise by as much as 10% over the next year. That 10% figure is a market expectation rather than an established industry-wide forecast.
For investors, however, the important question is not whether AI companies are taking more office space.
It is:
Which listed landlords can actually turn that AI demand into higher rental income, net property income and DPU?
That distinction is crucial.
The investment chain is:
AI expansion → hiring → office demand → specific building → landlord → lease expiry → rental reversion → NPI → DPU
Until the chain reaches the income statement, “AI property exposure” is only a theme.
And that makes Singapore’s office market particularly interesting for Keppel REIT and CapitaLand Integrated Commercial Trust investors.
AI is creating a new source of office demand
The traditional office-demand equation is relatively straightforward:
Economic growth
→ employment
→ corporate expansion
→ office demand.
AI potentially adds another pathway:
AI investment
→ AI companies establish regional operations
→ hiring and R&D
→ headquarters and customer-facing teams
→ demand for high-quality offices.
There is already evidence that this is happening.
OpenAI announced in May that it would establish its first Applied AI Lab outside the US in Singapore and invest more than S$300 million in the country, with plans to hire more than 200 specialised professionals over the coming years.
Its reported 100,000-square-foot Shaw Tower requirement would represent a substantial expansion from its current flexible-office presence at CapitaSpring.
Anthropic is also establishing a Singapore presence, while Mistral has taken dedicated office space at Asia Square and has said it intends to increase its Singapore headcount.
Databricks is another example of the trend: the Financial Times reported that the company plans to expand its Singapore office footprint to 32,000 square feet and increase its local headcount.
The significance is not simply the number of square feet.
It is that some AI companies appear to be moving beyond temporary or flexible arrangements into dedicated corporate space.
CBRE specifically noted that AI occupiers that had spent the previous two to three years in coworking environments were beginning to move into traditional self-managed offices, suggesting greater operational commitment to Singapore.
That could create a new, incremental source of demand for premium office buildings.
But there is a catch.
Singapore’s office market was already tight
The AI story would be less interesting if Singapore had a large amount of vacant Grade A office space waiting to be occupied.
It does not.
CBRE reported Core CBD Grade A vacancy of 3.3% in Q2 2026, down from 7.8% in Q4 2024. Core CBD Grade A rents reached S$12.50 psf per month, representing six consecutive quarters of rental growth.
More importantly, CBRE said Shaw Tower’s completion effectively marked the end of meaningful new office supply for 2026, with no significant new completions projected through 2027.
This creates an important setup:
Demand is increasing
while
high-quality supply remains constrained.
That is the environment in which landlords can obtain pricing power.
The FT’s latest reporting suggests that AI companies are now competing for precisely the type of centrally located, high-quality office space that is becoming scarce.
But scarcity alone does not tell an investor which REIT will benefit.
For that, we have to go one level deeper.
The AI-to-DPU chain investors should follow
Imagine an AI company signs a large lease in Singapore.
That sounds bullish for property.
But there are at least five questions an investor should ask.
1. Which building?
A 100,000-square-foot lease is only relevant to a particular asset.
2. Who owns it?
The building could be privately owned, owned by a listed REIT or held through a joint venture.
3. What rent is being paid?
A new lease does not automatically mean a large increase in rental income.
4. When does the existing lease expire?
If a landlord has a long lease at a fixed rent, it may not capture higher market rents immediately.
5. How much reaches DPU?
Even higher property income can be diluted by financing costs, management fees, joint-venture structures or a larger unit base.
So the real investment equation is:
AI tenant
↓
specific property
↓
specific landlord
↓
lease expiry
↓
rental reversion
↓
NPI
↓
distributable income
↓
DPU
This is the framework that separates an AI-themed property story from an actual investment thesis.
OpenAI shows why the distinction matters
OpenAI’s reported Shaw Tower negotiations are an excellent example.
The company is reportedly discussing approximately 100,000 square feet across five floors of the new building.
At first glance, that looks like an obvious positive for Singapore office landlords.
But Shaw Tower is owned by Shaw Towers Realty. Lendlease was appointed to handle development, project, construction, asset and property management.
So:
OpenAI → Shaw Tower
does not automatically mean:
OpenAI → SGX-listed REIT earnings.
That may sound obvious, but it is exactly the type of distinction that can disappear when investors turn a thematic story into a stock thesis.
The lesson is simple:
Never stop at the tenant. Identify the landlord.
Mistral provides another warning
Mistral AI has taken dedicated office space at Asia Square and plans to expand its Singapore team.
But investors need to know exactly which building is involved.
This matters because CICT has been reshaping its office portfolio.
In April, CICT announced the divestment of Asia Square Tower 2 for an agreed property value of S$2.476 billion while acquiring Paragon for S$3.9 billion. The manager described the transaction as a capital-recycling exercise, selling the leasehold office asset at a 3.0% exit yield and redeploying capital into the freehold integrated development at a 3.9% net yield.
That means investors should not casually treat “Asia Square” as synonymous with CICT exposure.
The exact tower matters.
The owner matters.
And the lease matters.
This is why property-level mapping is more useful than simply counting AI companies entering Singapore.
Keppel REIT is where the thesis gets more interesting
If the question is which listed landlord has meaningful exposure to Singapore’s prime office market, Keppel REIT deserves close attention.
Its portfolio includes:
- Ocean Financial Centre
- Marina Bay Financial Centre
- One Raffles Quay
- Keppel Bay Tower
and, following its acquisition completed at the end of 2025, Keppel REIT owns a two-thirds interest in MBFC Tower 3.
Singapore accounted for 78.9% of Keppel REIT’s portfolio by value as at 31 March 2026.
That gives investors a relatively direct way to participate in Singapore’s prime-office rental cycle.
The operating numbers are also important.
For 1H 2026, Keppel REIT reported:
- 12.8% rental reversion
- portfolio WALE of approximately 4.5 years
- top-10 tenant WALE of approximately 8.0 years
- distributable income from operations up 25.2% year-on-year.
The 12.8% rental reversion is particularly relevant.
It demonstrates that landlords do not need AI companies to suddenly create an entirely new rental cycle.
The rental market is already showing pricing power.
AI could instead provide another source of demand that helps sustain that pricing power.
That is a more defensible thesis.
Keppel REIT also shows why lease expiry matters
There is another useful data point.
In 1Q 2026, Keppel REIT reported Singapore CBD office leases signed at a weighted average rent of approximately S$13.26 psf per month, compared with an average rent of S$11.98 psf per month for leases expiring in FY2026.
That spread helps illustrate the mechanism.
If existing leases are below prevailing market rents, the landlord has an opportunity to reprice space as leases expire.
Therefore, investors should not simply ask:
“Will Singapore office rents rise?”
They should ask:
“How much of the rent increase can this REIT capture, and when?”
That is a much more useful question.
CICT is a different AI-office proposition
CICT is also highly relevant, but its investment exposure is different.
The trust is substantially larger and more diversified across offices, retail and integrated developments.
Its 1H 2026 results showed:
- gross revenue up 7.5%
- NPI up 8.7%
- distributable income up 13.3%
- DPU up 7.1% to 6.02 cents.
That means CICT is already producing income growth.
But the AI-office thesis is less pure.
The Asia Square Tower 2 divestment and Paragon acquisition make this particularly important. CICT is deliberately reshaping its portfolio rather than simply maximising exposure to Singapore Grade A offices.
Paragon also brings a very different mix of retail, medical and office space.
The acquisition gives CICT exposure to a freehold integrated asset with 100% committed occupancy across its retail and medical/office components as at the stated dates.
So CICT’s investment proposition is broader:
Singapore commercial property
rather than simply:
Singapore AI-driven office rental growth.
That distinction matters when investors try to translate the AI theme into DPU expectations.
The AI tenant map is more important than the AI tenant list
The most interesting research exercise from here is not to compile every AI company entering Singapore.
It is to build an:
AI Tenant → Property → Landlord → Lease → Earnings map
Consider the difference.
| AI company | Reported Singapore expansion | Property / space | What investors need to establish |
|---|---|---|---|
| OpenAI | ~100,000 sq ft under negotiation | Shaw Tower | Private owner; not automatically an SGX REIT beneficiary |
| Anthropic | ~100 desks | Ocean Financial Centre via flexible-office operator | Direct landlord economics need to be distinguished from flexible-space occupancy |
| Mistral AI | Dedicated office | Asia Square Tower 1 | Confirm owner and lease economics |
| Databricks | 32,000 sq ft | Singapore office expansion | Establish property ownership and lease structure |
| Sierra | ~10,000 sq ft reported | Keppel South Central | Determine exact ownership/structure |
| Manus | Reported expansion | Keppel South Central / flexible workspace | Distinguish direct lease from flexible-office demand |
The table itself reveals something important.
Not every AI tenant produces the same economic benefit for a landlord.
A direct long-term lease to a building owner is different from a company occupying flexible workspace.
A new lease at market rent is different from an existing tenant renewing.
And an AI company occupying a building owned by a private landlord does nothing directly for an SGX-listed REIT.
There is another complication: AI companies may not need offices forever
The bull thesis is compelling.
But it should not be treated as permanent.
AI companies are currently expanding headcount aggressively in Singapore.
That creates near-term office demand.
But AI itself could eventually change the economics of office demand.
There are two opposing forces.
Near term
AI investment →
more companies →
more hiring →
more office demand.
Longer term
AI productivity →
more output per employee →
potentially fewer employees required for a given level of output →
potentially less office demand.
We do not yet have sufficient evidence to conclude which force will dominate over the long term.
There is also a more immediate question.
Some AI companies are moving from flexible offices into conventional offices. CBRE explicitly identified this transition in its Q2 research.
That means part of the apparent increase in conventional office demand could represent:
flexible office → dedicated office
rather than:
no office → entirely new office demand.
That distinction is important.
The bear case is not that AI companies leave Singapore
The more interesting bear case is that AI demand simply isn’t large enough to determine the entire office market.
Singapore’s office market contains thousands of companies.
AI companies are receiving considerable attention because they are fast-growing, well-funded and strategically important.
But if traditional occupiers begin reducing space, AI expansion could merely offset weakness elsewhere.
Likewise, today’s supply shortage will not necessarily last forever.
CBRE says there are no significant new completions through 2027, but the supply picture can change from 2028 onward.
If new Grade A supply arrives just as AI-related demand moderates, rental growth could slow.
That is why the thesis should not be:
AI means office rents will keep rising.
It should be:
AI is adding a potentially durable demand source to an office market that is already supply-constrained.
Those are very different claims.
What investors should watch now
For investors following Singapore office REITs, the next set of numbers matters more than the number of AI headlines.
1. Rental reversion
This is arguably the clearest evidence that market rents are flowing into landlord economics.
Keppel REIT’s 12.8% 1H26 rental reversion provides a useful current benchmark.
2. Lease expiries
A rising market rent only matters when leases can be repriced.
Watch:
2026 → 2027 → 2028 lease expiries
and compare passing rents with current market rents.
3. Occupancy
Rental growth accompanied by high occupancy is a different proposition from rental growth achieved by filling previously vacant space.
4. New supply
The shortage of Grade A space is supporting landlords today.
The arrival of new supply could change that equation.
5. AI tenant quality
Do not simply count AI companies.
Ask:
- Is the tenant taking dedicated space?
- Is it expanding?
- Is the lease direct?
- How long is the lease?
- Is the rent materially above the previous passing rent?
- Who owns the building?
6. Financing costs
Higher property income does not automatically translate one-for-one into DPU.
Interest costs, refinancing requirements and leverage remain important.
AI has created two Singapore property plays
This is where the story becomes broader than an office-REIT article.
Singapore’s AI boom can reach property through at least two distinct channels.
AI → computing
AI investment
→ data-centre capacity
→ power and cooling requirements
→ data-centre rents
→ property income
→ DPU.
That is the mechanism behind the AI/data-centre thesis surrounding Keppel DC REIT.
AI → people
AI investment
→ hiring
→ regional headquarters and R&D teams
→ premium office demand
→ office rents
→ rental reversion
→ NPI
→ DPU.
That is the emerging office thesis.
The two stories are related but not identical.
Data centres depend heavily on:
- power availability
- data-centre capacity
- hyperscaler demand
- colocation demand
- cooling infrastructure
- technology investment.
Offices depend more heavily on:
- hiring
- corporate expansion
- workplace strategy
- location
- lease commitments
- future headcount.
Both are ways for investors to gain exposure to AI-driven real-estate demand.
But they have different economic transmission mechanisms and different risks.
The bigger investment lesson
The market may eventually talk about “AI REITs” as though AI exposure were a property-sector classification.
That would be too simplistic.
The real question is:
Where does AI spending ultimately become rent?
An AI company can spend billions on technology without increasing an SGX-listed landlord’s earnings.
A hyperscaler can expand without increasing the DPU of every data-centre REIT.
And an AI company can hire hundreds of people without benefiting a listed landlord if it occupies privately owned property.
The investor therefore has to trace the money.
AI spending
↓
economic activity
↓
physical requirement
↓
specific asset
↓
specific landlord
↓
lease economics
↓
rental income
↓
NPI
↓
DPU
That is the difference between an AI narrative and an AI investment thesis.
So, which Singapore REITs can actually capture the rent?
The evidence today points to an interesting but still developing picture.
Keppel REIT has a substantial Singapore prime-CBD portfolio, 78.9% Singapore exposure by portfolio value as at March 2026, and strong recent rental reversion. Its Ocean Financial Centre, MBFC and One Raffles Quay exposure makes it particularly relevant to the office-rental mechanism.
CICT provides a broader Singapore commercial-property exposure. Its earnings are growing, but the trust’s diversification and recent capital recycling mean that the AI-office thesis is only one part of its investment equation.
Neither should be treated as an “AI winner” simply because AI companies are expanding in Singapore.
The more useful conclusion is narrower:
Singapore’s AI boom is creating an additional source of demand for premium office space at a time when supply is constrained. The landlords that can monetise this demand will be those with the right buildings, high occupancy, below-market passing rents and leases expiring early enough to capture higher market rents.
That puts the spotlight not on the number of AI companies entering Singapore, but on lease expiry schedules and rental reversion.
And that is where the next stage of the research should focus.
What could invalidate the thesis?
The thesis would weaken if:
- AI hiring in Singapore slows materially;
- AI companies shift toward smaller or more flexible workplaces;
- conventional office occupiers reduce their footprints;
- significant new Grade A supply arrives faster than expected;
- market rents stop rising;
- rental reversions compress;
- landlords’ financing costs rise enough to offset property-income growth;
- AI companies increasingly use productivity gains to reduce headcount requirements.
The most important warning sign would therefore not necessarily be a weaker AI headline.
It would be:
falling rental reversion + rising vacancy + increasing supply.
That combination would tell investors that the property market’s pricing power is weakening.
The investment question
The most useful question for investors is therefore:
Is Singapore’s AI boom creating a new secular source of office rental growth — and which listed landlords have the lease-expiry profile to capture it?
That question is more valuable than asking which REIT is the “AI office REIT”.
Because ultimately, AI does not pay the DPU. Tenants do.