Singapore stopped approving new data centre land in 2019. The moratorium lifted in 2022 with a conditional pipeline attached — new capacity has to meet strict power efficiency and green energy targets before it gets built. The result, four years on, is a colocation market where the constraint was never really rack space. It was power. Most buyers still shop for colocation the way they'd shop for self-storage: how many square feet, how many U of rack, what's the monthly rent. In Singapore, that's the wrong question. The right question is how many kilowatts your equipment actually draws, because that number — not your floor footprint — is what a Singapore colocation provider is really selling you, and it's what will make your quote three times higher than the SME down the road with an identical-looking rack.
This guide explains how colocation pricing actually works here, what tier certifications mean for uptime, who the major providers are, how colocation stacks up against cloud and on-premise, and the questions to ask before you sign a contract.
Why Singapore's Land Moratorium Changed How Colocation Is Priced
Land scarcity is old news in Singapore — it's been the default constraint on every kind of property since independence. What's different about data centres is that the 2019 moratorium wasn't really about land at all. It was about electricity. Singapore's power grid, and the national ambition to hit long-term carbon targets, couldn't absorb an unlimited number of new data halls each drawing tens of megawatts. The Economic Development Board's 2022 Data Centre Call for Application reopened the pipeline, but only for operators who could commit to Power Usage Effectiveness (PUE) targets around 1.3 or better and demonstrate green energy sourcing.
That policy shift did two things to the market. First, it capped how much new colocation capacity can come online in any given window, which keeps utilisation at existing facilities high and gives operators little incentive to compete hard on price. Second — and this is the part most buyers miss — it made power density, not floor space, the currency colocation providers actually trade in. A facility's entire business case is built around how many megawatts it can deliver and how efficiently it can cool that load. When you ask for a rack, you're not really renting four square feet of floor. You're renting a slice of a finite, regulator-constrained power and cooling budget, and that's reflected in how the pricing is structured.
Businesses that come from markets like the US or Australia, where land is more available and power is comparatively cheap, often quote colocation costs to their Singapore finance team based on assumptions from those markets. The numbers don't transfer. Singapore colocation runs among the highest per-kilowatt costs in Asia Pacific, and the reason isn't greed — it's the physics and policy of a small island with a capped grid.
How Colocation Pricing Actually Works in Singapore
Once you understand that power is the scarce resource, the pricing structure makes sense. Providers typically quote in one of two ways, and knowing which one you're being quoted changes how you should compare vendors.
Per-rack pricing (space-based). You pay a flat monthly rate for a full or partial rack — commonly SGD 1,000 to SGD 3,000 for a full rack depending on facility tier and location. This model looks simple, but it usually comes with a power cap (say, 3kW to 5kW per rack) baked in. Exceed that draw and you either get billed overage or told to spread your equipment across additional racks — which means you're paying for space you don't need just to get the power you do.
Per-kilowatt pricing (power-based). You pay based on committed power draw regardless of how much physical rack space you actually use. This is the more transparent model for compute-dense deployments — GPU servers, dense virtualization hosts, high-performance storage arrays — where a single rack unit can draw more power than an entire rack of standard 2020-era servers. Singapore colocation power costs run in the region of USD 12 to USD 15 per watt of committed capacity in many facilities, among the higher end globally, a direct consequence of land scarcity and the power infrastructure investment required to serve it.
The mistake we see constantly: a business quotes "per rack" from one provider and "per kilowatt" from another and compares the headline numbers directly. They're not the same unit of measurement. A cheap-looking per-rack quote from a facility with a low power cap can end up costing more once you've paid for the extra racks needed to house equipment that couldn't fit within the power allocation of a single rack.
What actually drives your bill, beyond the base rate:
- Power draw and redundancy configuration. N+1 or 2N power redundancy (dual, independent power feeds to your equipment) costs meaningfully more than single-feed power, but it's the difference between riding out a feed failure and going dark.
- Cooling load. High-density racks need supplemental cooling beyond standard hot-aisle/cold-aisle containment. Some facilities charge a density surcharge above a certain kW-per-rack threshold.
- Cross-connects. Every physical cable running from your rack to another tenant, an ISP, or a cloud on-ramp inside the facility is a separate line-item, typically SGD 200 to SGD 800 per connection per month. This is the cost category almost every first-time colocation buyer underestimates — more on that below.
- Remote hands. Any physical task you need facility staff to perform on your behalf (swapping a failed drive, rebooting a server, running a cable) is billed hourly, usually SGD 100 to SGD 250 per hour with a minimum call-out.
- Bandwidth and IP allocation. Carrier-neutral facilities let you bring your own ISP or choose from several on-site; single-carrier facilities may bundle bandwidth into the rack rate at a markup.
For a mid-sized deployment — say, a half-rack running general-purpose virtualized servers at 2 to 3kW — an all-in monthly figure in Singapore typically lands between SGD 1,500 and SGD 3,500 once power, one or two cross-connects, and basic remote-hands allowance are included. Push into GPU or high-density compute and the power line alone can exceed that entire budget.
Tier Certification and What It Actually Means for Uptime
Colocation providers market their facilities using the Uptime Institute's Tier system (Tier I through Tier IV), and it's worth understanding what each tier actually guarantees before you pay a premium for a higher number.
| Tier | Redundancy | Expected Annual Downtime | Typical Use Case |
|---|---|---|---|
| Tier I | Single path for power and cooling, no redundancy | ~28.8 hours | Rarely offered commercially in Singapore |
| Tier II | Some redundant components, single distribution path | ~22 hours | Budget colocation, non-critical workloads |
| Tier III | N+1 redundancy, concurrently maintainable (no shutdown needed for maintenance) | ~1.6 hours (99.982% uptime) | Standard for Singapore enterprise colocation |
| Tier IV | 2N (fully redundant, fault-tolerant) | ~26 minutes (99.995% uptime) | Financial services, mission-critical systems |
Most commercial colocation facilities in Singapore operate at Tier III or claim "Tier III-equivalent" design, which means the facility can perform maintenance on any single component — a generator, a UPS bank, a cooling unit — without taking your equipment offline. Tier IV, with full fault tolerance across two independent paths, exists in Singapore but commands a meaningful price premium and is generally reserved for financial institutions and government-adjacent tenants under MAS TRM-driven requirements.
The distinction that matters in practice: Uptime Institute certification comes in two forms — a Design Certification (the blueprint meets the tier standard) and a Constructed Facility Certification (the built, operating facility was tested and verified against that standard). Some Singapore facilities market "Tier III" based only on design certification. Ask specifically which certification the provider holds, and ask for the certificate — not a marketing brochure that says "Tier III equivalent," which is not a certified standard at all and can mean almost anything the operator wants it to.
For MAS-regulated financial institutions, the practical uptime number matters less than the audit trail. MAS TRM guidelines require financial institutions to retain accountability for technology risk regardless of where systems are hosted, which means your due diligence needs to cover the facility's maintenance history, incident reports, and physical security audit logs — not just the tier badge on their website.
Major Providers at a Glance
This is not an endorsement of any of these operators — Aggasys is not a colocation facility and has no commercial relationship steering you toward one provider over another. This is simply who you'd typically be evaluating if you're shopping for colocation in Singapore in 2026.
Equinix. Operates the SG1 through SG5 campuses concentrated in the Jurong area, positioned heavily around interconnection — Equinix's core business model is being a neutral meeting point where enterprises, ISPs, and cloud providers cross-connect directly. If your architecture depends on low-latency links to multiple cloud providers from a single physical location, Equinix's ecosystem density is the primary reason to shortlist them, and it comes with pricing to match.
ST Telemedia Global Data Centres (STT GDC). A Singapore-headquartered operator with a substantial regional footprint across Southeast Asia and India as well as domestic facilities. STT GDC positions itself as a carrier-neutral operator serving both hyperscale cloud tenants and enterprise colocation customers, with facilities generally considered a strong fit for businesses wanting a Singapore-rooted provider with regional expansion options if they later need capacity in Jakarta, Bangkok, or Mumbai.
Digital Realty. A global colocation and interconnection operator with Singapore capacity, typically appealing to multinational tenants who already run Digital Realty facilities in other markets and want consistency in contract terms, service levels, and global account management across regions.
Keppel Data Centres. A Singapore-based operator notable for pioneering seawater cooling technology at its Keppel DC Singapore 7 facility — a closed-loop system using seawater rather than potable water for heat rejection, reported to deliver PUE improvements of up to roughly 80% in cooling energy efficiency compared to conventional air-cooled systems. Given Singapore's power-cost-driven pricing model discussed above, an operator with materially better cooling efficiency has a real structural cost advantage to potentially pass through — worth specifically asking about when you're comparing power-based quotes.
Evaluate all of them (and any others you're considering) against the same criteria: certified tier level, actual power cost per kW at your target density, cross-connect ecosystem relevant to your architecture, and physical proximity to your team if remote-hands visits will be frequent.
Colocation vs Cloud vs On-Premise
The three models solve different problems, and the right answer depends on workload shape more than company size.
| Factor | On-Premise | Colocation | Public Cloud |
|---|---|---|---|
| Who owns the hardware | You | You | Provider |
| Upfront cost | High (SGD 60K–130K+ typical SME buildout) | Medium (hardware only, no facility build) | None |
| Monthly cost for steady, predictable workloads | Low (after capital outlay) | Medium, predictable | Often highest at scale |
| Data sovereignty certainty | Full — it's in your office | Full — you control the hardware and access logs | Depends on provider region and configuration |
| Uptime standard achievable | Depends entirely on your setup | Tier III/IV (99.98%+) | Provider SLA, typically 99.9%–99.99% |
| Scalability | Slow — requires new hardware purchase | Moderate — add racks/power as needed | Near-instant |
| Best fit | Regulated firms with cheap office space and in-house IT | Predictable, 24/7 workloads at scale; hardware you already own; latency-sensitive systems | Bursty demand, rapid development, globally distributed apps |
The workload profile that most consistently favours colocation: predictable, continuously-running compute where you already own (or plan to own) the hardware, and where either the monthly cloud bill for an equivalent workload has climbed past roughly SGD 8,000–10,000, or data sovereignty and audit certainty are non-negotiable for regulatory reasons. Singapore colocation facilities with direct interconnects to regional network hubs typically deliver sub-50 millisecond latency to other Southeast Asian markets, which is more than adequate for the vast majority of regional applications and often materially better than routing through cloud regions elsewhere.
Where colocation loses: workloads that spike unpredictably, applications still in active development where provisioning speed matters more than unit cost, and anything genuinely global in user base where cloud's multi-region footprint does work colocation cannot replicate without building or renting capacity in every region you serve.
What We Typically See in Practice
When Aggasys works with clients evaluating or already running colocation, two mistakes show up more often than any others, and both are expensive in ways that don't appear until the second or third invoice.
Over-provisioned rack space, under-provisioned power planning. Clients frequently sign for a full rack "to be safe" and end up using 30% of the physical space while hitting the power cap on the remaining allocation within the first year of growth. The instinct to buy space as a buffer is backwards in a market where power, not space, is the constraint. We consistently recommend clients start with an accurate power draw calculation — actual measured wattage of the equipment going in, not nameplate maximum rating, which overstates real draw by a wide margin — and negotiate the power allocation first, letting the rack count follow from that number rather than the other way around.
Cross-connect costs treated as an afterthought. This is the single most underestimated line item in every colocation budget we review. A business plans for the rack rental and the equipment cost, then discovers mid-negotiation that reaching their cloud provider's on-ramp, their ISP, a backup replication target at a second facility, and a partner's network all require separate cross-connects, each billed monthly, each requiring its own provisioning lead time (often two to four weeks, not instant). We've seen clients budget for one cross-connect and need five once the actual network architecture is mapped out. The fix is straightforward but rarely done: diagram every external connection your deployment actually needs — cloud, ISP, DR site, partner networks — before signing, and price every one of them individually rather than assuming they're bundled.
Two secondary patterns worth flagging. Clients regularly assume "Tier III" marketing language equals certified performance without asking whether the facility holds Design or Constructed Facility certification — a distinction that matters if you ever need to demonstrate compliance to an auditor or regulator. And remote-hands costs are consistently underestimated for businesses without a strong on-site or nearby IT presence — if your team can't drive to the facility to swap a drive themselves, budget realistically for the hourly remote-hands rate, because it adds up faster than clients expect when a facility is 40 minutes from the office instead of 10.
Colocation for Startups and Smaller Deployments
Colocation has a reputation as an enterprise-only product, and for full-rack, dedicated-power deployments, the reputation is largely earned — the economics only make sense at a certain scale. But Singapore facilities increasingly offer entry points that fit smaller businesses and growth-stage companies:
- Shared or "cage-lite" rack space — quarter-rack and half-rack options exist at several facilities, letting a startup with a handful of physical servers get Tier III uptime without committing to a full rack's power allocation.
- Hybrid deployments — a small colocated footprint for the systems that genuinely need physical control or predictable low latency (a core database, a licensing server, specialised hardware that doesn't virtualize well), with everything else running in public cloud. This is by far the most common pattern we see among growth-stage companies rather than an all-or-nothing choice.
- Managed colocation — some providers, and increasingly managed service partners like Aggasys, offer to handle the actual server management, monitoring, and remote-hands coordination on top of the facility relationship, which removes the need for the client to have in-house data centre expertise at all.
The threshold question for a smaller business isn't "are we big enough for colocation" — it's "do we have specific hardware or specific compliance requirements that public cloud can't satisfy as cleanly." If the answer is yes for even one system, a small colocation footprint for that system alone, paired with cloud for everything else, is often the most cost-effective and lowest-risk structure available.
Checklist: 8 Questions to Ask a Colocation Provider
- What is the certified tier — and is it Design Certified or Constructed Facility Certified? Ask for the actual Uptime Institute certificate, not marketing language like "Tier III equivalent."
- Is pricing quoted per rack, per kilowatt, or both — and what happens if I exceed my power allocation? Get the overage rate in writing before you sign.
- What redundancy configuration am I actually getting — single feed, N+1, or 2N? This directly determines your realistic downtime exposure, independent of the tier marketing.
- What does every cross-connect cost, and what's the provisioning lead time? Map every external connection you need (cloud on-ramp, ISP, DR site, partners) and price each one individually.
- What's the remote-hands rate, minimum call-out charge, and typical response time? Especially important if the facility isn't within easy driving distance of your team.
- What's the facility's actual PUE, and how is cooling handled at higher rack density? A materially better PUE (as with seawater or other advanced cooling approaches) can be a genuine cost advantage given how power-cost-driven Singapore colocation pricing is.
- What audit and compliance documentation can you provide? For MAS-regulated or PDPA-sensitive workloads, ask specifically for physical security audit logs, access control documentation, and incident history — not just a compliance badge.
- What's the contract term, exit process, and physical equipment removal procedure? Understand what happens to your hardware and your data if you need to migrate out, including notice periods and any decommissioning fees.
Frequently Asked Questions
Why is colocation priced by power in Singapore instead of by rack space?
Singapore's 2019 data centre land moratorium (lifted in 2022 under strict power efficiency and green energy conditions) capped how much new colocation capacity can come online, and the underlying constraint was never floor space — it was the national power grid's ability to absorb new load. As a result, a facility's business case is built around how many megawatts it can deliver and cool, not square footage. When you rent a rack, you're really renting a slice of a finite, regulator-constrained power and cooling budget, which is why Singapore colocation pricing needs to be compared on a per-kilowatt basis, not just headline rack rental rates.
What's the difference between per-rack and per-kilowatt colocation pricing?
Per-rack pricing charges a flat monthly rate (commonly SGD 1,000–3,000 for a full rack) that usually includes a power cap of 3–5kW — exceed it and you pay overage or need additional racks. Per-kilowatt pricing charges based on committed power draw regardless of physical space used, which is more transparent for compute-dense deployments like GPU servers where a single rack unit can draw more power than an entire rack of older equipment. Singapore power costs run roughly USD 12–15 per watt of committed capacity. The two models aren't directly comparable on headline price alone — a cheap per-rack quote with a low power cap can end up costing more once you've paid for extra racks to house equipment that didn't fit the power allocation.
What do Tier III and Tier IV certification actually guarantee?
Tier III facilities offer N+1 redundancy and are concurrently maintainable, meaning any single component (a generator, UPS bank, cooling unit) can be serviced without taking your equipment offline — expected annual downtime is around 1.6 hours (99.982% uptime), and this is the standard for most Singapore enterprise colocation. Tier IV adds full 2N fault tolerance across two independent paths, cutting expected downtime to roughly 26 minutes annually (99.995%), at a meaningful price premium generally reserved for financial institutions and government-adjacent tenants. Critically, ask whether the certification is Design Certified (blueprint meets the standard) or Constructed Facility Certified (the built facility was tested and verified) — some Singapore facilities market "Tier III" based only on the former.
What colocation costs do first-time buyers typically underestimate?
Cross-connects are the most underestimated line item — every physical cable to a cloud on-ramp, ISP, DR site, or partner network is billed separately, typically SGD 200–800 per connection per month, and provisioning can take two to four weeks. Remote-hands charges (SGD 100–250 per hour with a minimum call-out) are also frequently missed by businesses without staff near the facility. On the space side, clients often over-provision rack space "to be safe" while under-planning power, since power — not floor space — is the actual constraint in the Singapore market.
Does colocation make sense for a startup or small deployment?
Yes, in the right shape. Quarter-rack and half-rack options let smaller businesses get Tier III uptime without committing to a full rack's power allocation, and the most common pattern among growth-stage companies is a hybrid model — a small colocated footprint for systems that need physical control or predictable low latency (a core database, licensing server, or specialised hardware), with everything else in public cloud. The right question isn't company size, it's whether you have even one system with hardware or compliance requirements public cloud can't satisfy as cleanly.
Talk to Aggasys About Whether Colocation Fits Your Infrastructure Roadmap
Aggasys is not a colocation facility — we're an infrastructure and managed services partner who helps Singapore businesses figure out whether colocation, cloud, on-premise, or a hybrid of the three actually fits their workloads, compliance obligations, and growth plans, and then manages the hardware once it's in place. We've run this assessment for organisations across education, aviation, maritime, and financial services since 2006, and the recommendation is never the same twice because the right answer depends entirely on what you're actually running.
If you're evaluating a colocation move, revisiting an existing facility contract, or trying to work out whether your current cloud spend would be better spent on colocated hardware, we'll map your actual power draw, network requirements, and compliance obligations against the real Singapore market before you sign anything.
Book your free infrastructure assessment: aggasys.com/contact or call (+65) 6250 0045.
Written by Lee Yang Sean, Aggasys Solutions | sean@aggasys.com | LinkedIn