Market · Asia Pacific
APAC. The constraint is rarely the building.
Power, land and capital are available somewhere in this region on almost any timeline you want. What decides your deployment is where your data is allowed to sit, where the silicon is allowed to go, and which operator can actually execute.
Where the capacity actually is
- —Singapore
The best connectivity, legal environment and operator quality in the region, and the most deliberately rationed supply. New capacity is awarded selectively against efficiency and sustainability criteria rather than simply sold to the highest bidder. Treat Singapore as a scarce premium allocation for latency-critical and regulated workloads, not as the regional default.
- —Johor and southern Malaysia
Where the demand Singapore cannot serve has gone. Land, power and cost are all materially better, and the latency penalty is negligible for almost every workload. The real diligence items are the pace of power delivery and water availability, both of which need documentation rather than assurance.
- —Japan — Tokyo and Osaka
Mature, reliable, well-operated and expensive. Power cost is high and seismic requirements add engineering, but delivery certainty and operator quality are the best in the region. Osaka has become the serious second market rather than a disaster-recovery afterthought.
- —India — Mumbai, Chennai, Hyderabad, Pune
The cheapest capacity in the region and the fastest-growing supply, with subsea landings at Mumbai and Chennai. Data localization obligations are real and increasing, grid quality varies by state, and the execution gap between the best and worst operators is wider here than anywhere. Operator selection is the mandate.
- —Australia — Sydney and Melbourne
Mature markets, good power, strong legal environment and a solid domestic demand base. Latency to the rest of APAC limits the regional hub case, so this is generally a domestic answer.
- —South Korea, Indonesia and Vietnam
Seoul is constrained on power and land in the metro area despite very strong demand. Indonesia and Vietnam are early but moving, with cost advantages and correspondingly higher execution and regulatory risk.
What gates an APAC deployment
The firm’s APAC presence runs through Strategic Supply Partners in Sydney, co-founded on the same model — no inventory, no listings, retained by both sides — so that the operator and offtaker relationships in this region are held directly rather than borrowed.
In North America the question is usually when. In APAC the question is usually whether — and it gets answered by regulators and by export control before it gets answered by a developer.
- Export control on advanced accelerators. Where the hardware may be sited, who may own it and who may access it downstream. This is the first question for any APAC deployment, not a late-stage compliance check, because getting it wrong does not delay a plan — it voids it.
- Data localization. Several markets impose residency obligations on categories of data, and the obligations differ by sector. This frequently decides the country before any commercial comparison begins.
- Power delivery pace, not power availability. Most APAC markets have the generation. What varies enormously is how quickly a specific site can be energised at the megawatt level you need, and how firm the utility’s commitment actually is.
- Water. In the tropical markets, water availability and discharge policy determine whether high-density liquid cooling is viable at all. This is where cooling method becomes a site-selection question rather than an engineering preference.
- Operator execution risk. The gap between the strongest and weakest operators in this region is larger than in North America or Europe. Track record at your scale, in that country, is the only predictor worth weighting, and it is verifiable.
- Currency and repatriation. Long-term contracts denominated in volatile currencies, with varying ease of moving money out, are a genuine commercial exposure and belong in the agreement rather than in a footnote.
Where the answer is a purpose-built cluster rather than existing space, the GPU cluster procurement mandate covers allocation and the contracts around it; GPUaaS contract negotiation covers the managed alternative.
What we source here
Asked in this market.
01Can you still get capacity in Singapore?
In limited volume and on a selective basis. Singapore rations new data center capacity deliberately, awarding it against efficiency and sustainability criteria rather than simply selling it. It remains the best connectivity and legal environment in the region, and it should be treated as a scarce, premium allocation rather than a default.
02Why does Johor come up in every Singapore conversation?
Because it is the release valve. Southern Malaysia sits close enough to Singapore that latency is negligible for most purposes, with land, power and cost that Singapore cannot offer. The genuine constraints are power delivery pace and water, and both need to be documented rather than assumed.
03Is Japan worth the premium?
For workloads serving Japanese users or Japanese enterprises, generally yes. Tokyo and Osaka offer excellent reliability, strong operators and a mature legal environment. You pay for power and you accept seismic engineering requirements, and in exchange you get a level of delivery certainty that is rare in the region.
04What makes India complicated?
Nothing about the economics — it is among the cheapest markets in the region and capacity is being built rapidly. The complications are data localization obligations, variable grid quality between states, and a wide gap in execution between the strongest and weakest operators. Operator selection matters more here than almost anywhere.
05How do export controls affect APAC deployments?
Considerably, and unevenly by country. The regime governing where advanced accelerators may be sited and who may access them is the first thing to establish for any APAC deployment, before site or platform selection. Getting this wrong invalidates the entire plan rather than merely delaying it.
06Is Australia a serious AI infrastructure market?
For domestic workloads, yes — Sydney and Melbourne are mature markets with good power and a strong regulatory environment. Latency to the rest of APAC limits its role as a regional hub, so it is usually a domestic answer rather than a regional one.
GPU cluster procurement
Allocation, platform selection and the contracts around a bare metal build in-region.
Open the mandate →ContractsGPUaaS contract negotiation
Term, commitment, SLA and exit — read properly before you sign a three-year agreement.
Open the mandate →MarketEMEA
The other region where regulation decides the site before economics get a vote.
Open the market →Establish what is allowed before you shortlist anything.
Tell us the workload, the data and the entity. We will tell you which APAC markets are actually open to you and which operators can execute at your scale.