Singapore data centres and the AI bubble
The essentials in 30 seconds
- Singapore data centre risk in an AI bubble comes down to 6 checks: leases, tenants, power, pipeline, debt, and pricing.
- A hall leased long term behaves like a bond, and a hall built on a forecast behaves like a bet.
- The EQOS Asia AI study scores the enterprise AI market at 0.337, a fragile zone (April 2026).
- About $200B a year is invested in AI, and 95% of AI projects fail in production (MIT 2025, cited by the study).
- Each check returns a number you can date and compare.
A Singapore data centre sells space, power, and cooling to customers who run AI and cloud workloads. This guide is part of the page on AI bubble burst and Singapore, and it gives 6 checks for any hall or campus, as of October 2026. It is general information, not investment advice.
On this page
Compare the EQOS Asia offers and request a 30-minute pre-diagnostic.
Why do Singapore data centres matter in an AI bubble debate?
A data centre turns AI demand into a physical asset with a long life. The building lasts decades, while the servers inside change every few years. That mismatch is the core of the debate: a hall is built on a forecast of demand, and the forecast can move faster than the concrete. Singapore concentrates this question because land and power are limited and every hall competes for them.
The EQOS Asia AI study frames the demand side. About $200B a year is invested in AI, valuations rose ×4 in 18 months, and 95% of AI projects fail in production (MIT 2025, cited by the study, April 2026). If many projects stall, the customers who rent halls may slow their orders, even while their ambitions stay high.
How long should leases run for a data centre to look sturdy?
Lease length is the first check. A hall leased for 5 years or more to a customer with a strong balance sheet produces income that survives a change in sentiment. A hall leased month to month produces income that follows sentiment. Ask for the weighted average lease term and the date of each large expiry, then chart the expiries on one timeline.
A cluster of expiries in the same year is a warning even when the average looks long. If 3 large leases end together, one tenant decision can empty a third of a campus. Spread expiries across years, and renewal talks become a routine event instead of a crisis.
Which tenant patterns raise AI bubble risk in a data centre?
Tenant concentration is the second check. A campus that depends on one cloud provider, or on one AI start-up, carries the risk of that single customer. A mix of enterprise, cloud, and network tenants spreads the risk. Check how much of the rent comes from the 3 largest customers, and how those customers finance their own growth.
Customers funded by new investor money behave differently from customers funded by revenue. The first group can cancel quickly when funding tightens. The second group follows its own sales. The funding source of the tenant is therefore a data point about the landlord, and it belongs in your file.
What do power and cooling say about data centre risk?
Power is the scarce input. A hall with secured power capacity holds value because a new tenant can move in. A hall waiting for a power allocation is a project, not an asset. Check the contracted capacity, the share already used, and the cost per unit of power compared with rivals in the region.
Cooling adds cost for AI racks, which draw more power per cabinet than older workloads. Halls designed for lighter loads may need costly upgrades to host AI customers. A budget line for retrofit tells you how much new spending sits behind a hall that looks finished.
How does new capacity change data centre risk in Singapore?
A pipeline of new halls adds supply. If demand grows as forecast, supply is absorbed. If demand slows, supply competes for the same tenants and pushes rents down. Compare the capacity under construction with the capacity leased in the last 12 months. A pipeline many times larger than recent leasing is a sign to ask more questions.
Neighbouring markets matter too. Campuses in the region can offer cheaper land and power, and tenants can split their needs between sites. Nearby alternatives, including Johor, shape what a Singapore hall can charge. Ask how many tenants have signed options on other sites, because an option is a quiet sign of how easily a customer can leave.
Which debt and pricing checks complete the data centre review?
Debt is the fifth check: loan size against value, interest cost against rent, and the date when each loan must be refinanced. A hall with moderate debt and distant maturities has time. A hall with heavy debt due next year has a deadline. Pricing is the sixth check: compare what investors pay for the asset with the rent it earns.
Score each of the 6 checks from 0 to 2 and add them. A total of 9 or more points to a sturdy asset. A total of 5 or less points to an asset that depends on the market staying calm. The score does not predict prices, and it ranks assets on a common scale.
How can you use these data centre checks in practice?
Run the checks once a quarter and keep the answers in one table with dates. Compare halls with each other and each hall with its own past. A hall that improves on 3 checks in a row earns more confidence than a hall that scores well once. Share the table with lenders and owners so the discussion stays on numbers. This page is general information, not investment advice, and a licensed adviser should confirm what fits your case.
For a first score of your own case, the Λ simulator takes 10 minutes. For a decision that involves AI spending in Singapore, a 30-minute pre-diagnostic applies the EQOS Asia scoring and returns a ranked set of options.
See also in this guide
Frequently asked questions
Are Singapore data centres exposed to an AI bubble?
They are exposed through demand from AI and cloud customers. Exposure differs by hall. Long leases, diverse tenants, secured power, moderate debt, and fair pricing reduce it. A forecast-based project with heavy debt carries more. This page is general information, not investment advice.
Which check matters most for a data centre?
Lease length and tenant strength come first, because they decide whether income is signed or forecast. Power comes second, since it decides whether a vacant hall can be re-let. Debt maturity comes third, as it sets the deadline for any problem.
How do I score a data centre with the 6 checks?
Give each check 0, 1, or 2 points, then add them. A total of 9 or more points to a sturdy asset, and a total of 5 or less points to an asset that depends on calm markets. The score ranks assets on a common scale and does not forecast prices.
Why does the tenant funding source matter?
Customers funded by new investor money can cancel quickly when funding tightens, while customers funded by their own revenue follow their sales. The funding source of a tenant therefore tells you how stable the rent is.
Test your knowledge
1. How many checks does this page give?
- 3
- 6
- 10
Show the answer
6.
2. What does a total of 9 or more points mean?
- A sturdy asset
- A crash date
- An empty hall
Show the answer
A sturdy asset.
3. Which input does the page call scarce?
- Land titles
- Power
- Staff
Show the answer
Power.
4. What does a month-to-month lease produce?
- Income that follows sentiment
- Income that survives any shock
- No income
Show the answer
Income that follows sentiment.
5. Which funding source makes a tenant steadier?
- New investor money
- Its own revenue
- A rising share price
Show the answer
Its own revenue.
About the author
Stéphane Jambu
President, EQOS Forecast Asia
Stéphane Jambu leads EQOS Forecast Asia and works as a consultant in e-commerce, search visibility, and AI. LinkedIn profile
Sources: EQOS Asia AI impact study (eqos-asia.com/case-studies/ia-2026.html), EQOS Asia simulator page (eqos-asia.com/simulator.html) and contact page (eqos-asia.com/contact.html), consulted on 22 October 2026. Page updated on 22 October 2026.