Singapore’s Data-Centre Test Is No Longer Scarcity. It Is Operating Discipline

Singapore’s Data-Centre Test Is No Longer Scarcity. It Is Operating Discipline

*By Matthieu Gallego — 7 November 2026*

Singapore’s data-centre market has spent several years proving that scarcity can be governed. The opening of Equinix SG6 on 6 November shows the harder part: turning a selective capacity award into an operating facility without losing the distinctions between a first phase, a full-build design and capacity that customers can actually use.

Equinix says SG6 opened with 8 MW in its first phase and approximately 1,550 cabinets. The company describes up to 20 MW of IT capacity at full build-out, an initial investment of about US$290 million, a target power usage effectiveness of 1.20, liquid-cooling capability and a 2027 co-innovation facility with the National University of Singapore. The operator also says the building is covered by 100% clean and renewable energy and incorporates rooftop and facade solar generation.

Those are meaningful delivery facts. They do not make all 20 MW simultaneously live or available. The 8 MW first phase is the operating milestone; the remaining design capacity is a development envelope. That distinction is more than database hygiene. It is the economic logic of Singapore’s market.

My view is that Singapore is moving from a market defined by constrained supply to one defined by operating discipline. A developer no longer wins merely by obtaining permission to add megawatts. It must show how each allocated phase will support high-value digital activity, consume less energy per unit of compute, integrate with a tight power system and produce evidence that the model can be repeated.

SG6 converts a policy award into a delivery record

SG6 is among the first facilities built through Singapore’s original Data Centre Call for Application process. That matters because the framework did not treat electricity capacity as an ordinary input available to any credible developer. Capacity was selectively allocated under policy conditions.

The facility’s opening therefore closes one part of the evidence chain:

• a capacity pathway was awarded;
• the project was financed and built;
• phase one is now open;
• technical characteristics can be tested against operating performance.

The last point is the most important. A target PUE of 1.20 is a design and certification objective until measured over a representative operating period. Liquid cooling is a capability until customers deploy compatible high-density systems. Renewable coverage is a procurement claim whose annual and hourly basis matters. The co-innovation facility is a 2027 commitment, not a current operating result.

None of these qualifications diminishes SG6. They identify the next facts that will determine whether the project becomes a reference design rather than simply another opening.

What PowerlandMap’s Singapore evidence shows

At the 7 November review point, PowerlandMap contained 55 verified public Singapore supply records, excluding demo records. Forty-eight were classified as live and seven remained in the active pipeline. Only ten records carried an explicit operator-sourced IT-capacity figure, with different bases and levels of completeness.

These counts describe the evidence currently held in the database. They are not a claim that Singapore has only 55 facilities, and the ten IT-capacity figures should not be treated as a complete national total. Many records have verified identity, operator and location but no comparable IT-load value. Others publish campus or electrical capacity without defining the IT boundary.

SG6 illustrates why that restraint matters. The record now carries 20 MW because Equinix explicitly defines the full-build figure as IT capacity. Its capacity basis simultaneously states that only 8 MW opened in phase one. PowerlandMap does not convert the unused portion of the design envelope into available capacity.

That method produces a more useful commercial record. An investor can distinguish the facility’s maximum design from the capacity associated with the opening milestone. A tenant can ask what part of the first phase is commissioned, contracted, occupied or technically available. A public authority can compare the awarded capacity with delivery and efficiency outcomes.

The first sensitivity is phase, not PUE

PUE receives attention because it is easy to compare. Phase discipline is more fundamental.

If the full 20 MW IT design were eventually operated at a PUE of 1.20, the corresponding facility input would be approximately 24 MW before considering the limitations of a simplified annual model. At 90% average IT utilisation, the annual energy associated with the IT load alone would be about 157.7 GWh; applying a 1.20 PUE would bring total facility energy to about 189.2 GWh.

For the current 8 MW first phase, the same illustrative assumptions produce about 63.1 GWh of annual IT energy and 75.7 GWh of facility energy.

The calculation is transparent: facility input equals IT load multiplied by PUE, and annual energy equals facility input multiplied by 8,760 hours and 90% utilisation.

This is not a forecast of SG6 consumption. Actual utilisation, redundancy, cooling performance, commissioning sequence and customer mix are unknown. The sensitivity simply shows why “opened” and “20 MW” cannot be collapsed into one status. The difference between the first phase and the full design represents 14.4 MW of facility input under the assumed conditions—material in a system where every increment of demand competes with other industrial and household needs.

Singapore is allocating more than electricity

The second Data Centre Call for Application makes the policy direction clearer. Singapore provisionally allocated 200 MW to four proposals in August 2026. The assessment was not based only on technical efficiency. The government also evaluated strategic value, economic contribution, innovation, talent, international connectivity, resilience and green-energy commitments.

This means a megawatt of data-centre demand is being treated as a scarce development right with attached obligations.

For operators, that changes site economics. The cost of winning capacity includes the capability to document outcomes beyond the building: research partnerships, workforce development, connectivity, energy procurement and the type of digital workloads supported. The return on that investment depends on attracting customers whose value density justifies Singapore’s land and power constraints.

For customers, the framework can improve quality but narrow flexibility. Capacity inside a selected facility may carry stronger sustainability and resilience credentials, yet it may also be more expensive and subject to stricter deployment requirements. The most valuable capacity will not necessarily be the lowest-cost rack. It will be the rack that combines network density, power reliability, high-density cooling, compliance and access to the regional ecosystem.

The grid plan changes the denominator

Singapore’s Energy Market Authority published a five-year infrastructure plan on 26 October. Peak electricity demand was 8.6 GW in August 2026 and could reach between 9.9 GW and 13 GW by 2031, with data centres and semiconductor manufacturing among the main drivers. EMA says five new gas-fired plants will expand generation capacity by around 26% between 2026 and 2031, while up to S$5.3 billion is projected for grid investment from 2025 to 2029.

This broader system context is essential. Singapore can allocate additional data-centre capacity because generation, gas infrastructure, electricity networks, storage, solar and low-carbon imports are being planned together. But the system remains exposed to fuel and infrastructure choices. Natural gas currently supplies around 95% of electricity needs.

A facility can achieve an efficient PUE and still operate within a carbon-intensive or fuel-import-dependent system. Conversely, a data centre can support system value through efficiency, flexible operation, storage and credible clean-energy procurement even when the national grid is not fully decarbonised.

The market should therefore resist a single-metric ranking. PUE measures the relationship between facility energy and IT energy. It does not measure carbon intensity, water use, compute productivity, hourly renewable matching or the value of the workload. Singapore’s policy is beginning to price these dimensions through selection, but operators will still need to publish enough operating data for customers and investors to compare them.

The serious objection: selection can protect incumbents

The strongest objection to Singapore’s model is that selective allocation can favour large operators with the capital and partnerships to satisfy complex criteria. That may reduce speculative development, but it can also make market entry harder and concentrate capacity among incumbents.

SG6 reinforces both sides of the argument. Equinix can combine a US$290 million first investment, an established interconnection ecosystem, a university partnership, specialised cooling and renewable procurement. Those capabilities improve the probability of delivery. They also create a threshold that smaller or newer operators may struggle to cross.

A tightly managed market can therefore become efficient but less contestable. If allocation criteria are opaque, or if operating outcomes are not compared with commitments, capacity can remain scarce without necessarily becoming more productive.

The answer is not to return to unrestricted growth. It is to make the allocation cycle auditable. Award criteria, capacity basis, delivery dates, actual efficiency, customer use and economic contributions should be reported in comparable form. Facilities that meet or exceed their commitments should build a stronger case for subsequent phases. Projects that remain late or underused should not retain the same claim on future capacity.

What investors and customers should verify next

SG6 now moves from construction diligence to operating diligence.

First, confirm the precise status of the 8 MW first phase: commissioned capacity, contracted capacity and available capacity are different figures. Second, track measured PUE over time rather than relying only on the 1.20 target. Third, examine the basis of the 100% renewable-coverage claim, including instruments, geography and temporal matching. Fourth, verify how liquid cooling is offered in commercial deployments and what rack densities are supported. Fifth, follow the 2027 co-innovation facility and whether its work produces deployable efficiency or storage outcomes.

For the wider Singapore market, the next proof is equally concrete. The four DC-CFA2 awards should progress from provisional capacity to named sites, defined capacity boundaries, construction milestones and customer-ready phases. The at-least-200-MW programme will be credible when those records can be reconciled without adding provisional allocations to operating supply.

Singapore’s advantage is the feedback loop

Singapore cannot compete by offering unlimited land, water or electricity. Its advantage is the ability to create a fast feedback loop between policy, infrastructure allocation, facility delivery and operating evidence.

SG6 is useful because it moves that loop forward. It turns an earlier capacity award into a real first phase and gives the market a specific facility against which design targets can be tested. It also shows why every number needs a boundary: 8 MW opened, 20 MW designed, 1.20 PUE targeted, and a 2027 innovation programme planned.

The next phase of Singapore’s data-centre market will be won by operators that convert those qualifiers into measured results. Scarcity created the selection mechanism. Operating discipline will determine whether the allocated capacity produces durable economic and infrastructure value.

Sources and method

Primary sources reviewed: Equinix’s 6 November 2026 SG6 opening announcement; Singapore EDB and IMDA materials for the second Data Centre Call for Application and the 21 August 2026 provisional allocation; IMDA’s Green Data Centre Roadmap; and EMA’s 26 October 2026 Energy Infrastructure Plan. PowerlandMap figures were calculated from verified public Singapore supply records on 7 November 2026, excluding demo records. Counts measure current database coverage, not market completeness. The energy table is an illustrative sensitivity using stated capacity, an assumed PUE of 1.20 and 90% utilisation; it is not a forecast of SG6 consumption.