Data Center Tiers: What Tier III and IV Prove to a Buyer

A data center tier tells you one thing: how the building behaves when a piece of its power or cooling plant has to come out of service. It does not tell you whether the operator is competent, whether the network is diverse or whether your own rack will stay up. And the word “tier” on a sales deck tells you even less, because anyone can print it.

That gap between the label and the proof is what this guide is about. We read the certifications listed in our own facility records, 6,018 data centers across 183 countries, and sorted every tier or rating claim by the strongest evidence behind it. The short version: 1,490 facilities, about 25% of the records, state a tier or rating at all, and only 618 of them show an Uptime Institute award for the finished building or its operations.

The rest of the guide explains what each tier means in engineering terms, which bodies certify what, how to read the evidence ladder, where certificates are common and where nobody lists them, what a tier does and does not do for your own rack, and the checks and contract clauses that turn a tier claim into something you can rely on.

Tier I to Tier IV in plain engineering terms

Uptime Institute’s Tier Classification System grades site infrastructure, meaning the power and cooling plant that keeps IT equipment running. Each tier includes everything in the tier below it. Uptime describes the four levels by what the plant can survive, not by a percentage, and that framing is the useful one for a buyer.

Tier I: basic capacity

A Tier I site has dedicated infrastructure for IT beyond what an office building offers: an uninterruptible power supply to filter spikes, sags and short outages, cooling that is not switched off at the end of the working day, and an engine generator for longer utility failures. There is one path for power and one for cooling, and no spare components. Any maintenance on the plant means shutting the IT load down. A small server room with a UPS and a generator behind it is a Tier I design.

Tier II: redundant capacity components

Tier II adds spare critical components: extra UPS modules, chillers or pumps, and generators beyond what the load needs. A failed UPS module or chiller can be swapped for its spare without an outage. The distribution path is still single, so work on the switchgear, the pipework or anything else on that one path still takes the IT load down. Uptime describes Tier II as giving the operator “select maintenance opportunities”, which is a careful way of saying not all of them.

Tier III: concurrently maintainable

Tier III is the threshold that most colocation buyers ask for, and for good reason. A Tier III data center is concurrently maintainable: every component that supports the IT load can be shut down for planned maintenance or replacement without shutting the IT load down, because power and cooling have a second delivery path. In Uptime’s wording, Tier III requires “no shutdowns for equipment replacement and maintenance.”

What Tier III does not promise is immunity to the unplanned. A Tier III design can still have single points of failure for unplanned events; it guarantees that the operator never has to choose between maintaining the plant and keeping you online. For a buyer that matters more than it sounds. Maintenance that has to wait for an outage window tends to be postponed, and a building where every breaker and pump can be serviced on schedule removes that excuse.

Tier IV: fault tolerant

Tier IV builds on Tier III and adds fault tolerance: an equipment failure or a distribution path interruption is stopped short of the IT operations, without a person having to act. In practice that means independent, simultaneously active power and cooling systems, kept apart so that one event does not reach both, each with enough capacity to carry the full load alone.

The difference between the two top tiers, in one line each: Tier III lets the operator take any component out for planned work without an outage. Tier IV adds fault tolerance: a failure of any single component or distribution path is stopped short of the IT equipment, with no human action needed.

Tier IV is not automatically the better buy. Uptime itself says the right tier depends on the business value of the application, and a fault tolerant plant costs more to build, more to run and more to rent. Two Tier III sites with your application replicated between them can give you a stronger position against a site-wide event than one Tier IV building, because a tier rating says nothing about events that take out the whole site.

The availability percentages, and why they were dropped

Search for data center tiers and you will meet a set of availability percentages, one per tier, usually followed by hours of downtime per year. They look precise and they are everywhere: in operator brochures, in procurement templates and in generated summaries.

They are not part of the Tier Standard. Uptime removed all references to expected downtime per year in 2009, and its current explanation of the system states that the standard does not assign availability predictions to tiers. The reason it gives is the one every outage investigator would give: how a site is operated affects availability at least as much as how it was designed. A Tier IV plant run by an understaffed team with poor procedures can lose a hall; a Tier II room run with discipline can go years without an incident.

So if a proposal promises an availability figure “because the site is Tier III”, treat the number as the provider’s marketing, not as a property of the tier. The only availability number that binds anyone is the one in the service level agreement, with a definition of what counts as downtime and a credit schedule attached. Ask for that, and ask for the provider’s actual incident history in the hall you would occupy.

Three bodies, three scales: Uptime, TIA-942 and EN 50600

Three systems dominate what you will find on data sheets, and they are not interchangeable. Only Uptime Institute certifies against its own Tier Classification System, at three stages: design documents, constructed facility and operational sustainability. TIA-942 ratings are audited by certification bodies licensed by TIA, and EN 50600 classes by certification bodies such as TÜV. A tier stated without any of these is a claim, not a certification. All three ladder up from basic to fault tolerant, which makes them look alike, but they differ in who audits, what is audited and how long a certificate lasts.

	Uptime Institute Tiers	ANSI/TIA-942 Ratings	EN 50600 / ISO/IEC 22237
Levels	Tier I–IV	Rated-1 to Rated-4	Availability Class 1–4 (plus protection classes and granularity levels)
Who certifies	Uptime Institute only	Certification bodies licensed by TIA	Accredited certification bodies (for example TÜV)
Stages	Design documents, constructed facility, operational sustainability	Design, facilities (on-site), “Ready” for modular units	Design and facility audits by the certifier
Validity	Design: 2 years; facility: until modified; operations: 1–3 years	Design: 1 year; facilities: 3 years with yearly surveillance	Per certifier’s scheme

Uptime Institute

Uptime describes itself as the only organisation that can formally certify data centers against its Tier Classification System, and it does not design, build or operate them. Its certification has three stages. The Tier Certification of Design Documents reviews the drawings. The Tier Certification of Constructed Facility is assessed on site, in the finished building, against the tier the design promised. The Tier Certification of Operational Sustainability looks at how the site is run: staffing and organisation, training, preventive maintenance, operating conditions, planning and coordination. Sites without a tier award can apply for the Management and Operations Stamp of Approval, which applies the same operational criteria.

Uptime reports over 4,300 awards issued in more than 120 countries. That figure counts awards, not buildings, because one building can collect a design award, a facility award and successive operations awards.

TIA-942

ANSI/TIA-942 is a standard for data center infrastructure published by the Telecommunications Industry Association. Its current revision, TIA-942-C, was published in May 2024 and covers telecommunications cabling, power, cooling, architecture, fire protection, physical security and safety, which is a broader scope than Uptime’s tiers. Its four ratings read much like the tiers: Rated-1 basic site infrastructure with single capacity components and a single non-redundant path; Rated-2 with redundant capacity components on a single path; Rated-3 concurrently maintainable with multiple independent distribution paths; Rated-4 fault tolerant, allowing concurrent maintenance and one fault anywhere in the installation without downtime.

Audits are done by companies licensed by TIA as certification bodies, not by TIA itself. A TIA-942 design certificate is valid for one year and can be extended while a site is not yet operational; a facilities certificate includes an on-site inspection, lasts three years, and requires surveillance audits at the end of years one and two. In March 2025 TIA reported more than 600 data centers certified under the standard in 51 countries.

EN 50600 and ISO/IEC 22237

The European standard EN 50600, mirrored internationally by ISO/IEC 22237, does not produce a single number. A certificate states an availability class from 1 to 4 for power, cooling and cabling, a set of protection classes for physical security, and a granularity level for energy-efficiency monitoring. Availability Class 3 corresponds to multiple supply paths that allow maintenance during operation; Class 4 adds fault tolerance. Treat a certificate as “EN 50600 Class 3” rather than “Tier III”, because the two are close in spirit but not equivalent clause by clause.

What a tier claim proves: the evidence ladder

Put the three systems together with the way operators actually describe their buildings and you get a ladder of evidence. At the top, an independent engineer has stood in the finished building and tested it. At the bottom, a marketing page uses the word “tier” with nothing behind it. Here is how the tier and rating claims in our facility records spread over that ladder.

Reading the ladder from the top:

Constructed facility and operational awards are the evidence that counts. Between them they cover 618 facilities, 41% of those that state any tier, and 10.3% of all the records. Of those, 86 also hold an Operational Sustainability award, which is the closest thing the industry has to proof that the plant is run the way it was designed to be run. A constructed facility award stays valid until the facility is modified, meaning changes to the capacity components or distribution paths from the certified design. A building that has been expanded or re-engineered since its award may no longer match it, so the award date matters.

Design-only awards are a checkpoint, not a result. 283 facilities show an Uptime design award and nothing later, which is 30% of the facilities that name Uptime at all. A design certificate proves the drawings met the tier, not that the finished building did. Uptime calls the design award provisional, a checkpoint on the way to the constructed facility award, and design awards issued since 1 January 2014 expire two years after the award date. It introduced that two-year expiry after finding that design certificates were being used to market facilities that were never certified as built. In July 2015 it went further and stopped issuing design certification to commercial data center providers in North America, from colocation to cloud. A design award that is older than two years and has not been followed by a facility award has, in Uptime’s own terms, expired.

“Uptime” with no stage is a question, not an answer. 45 facilities name Uptime without saying which award they hold. Some are certified as built and simply describe it loosely; some hold a design award. The award foil settles it in a minute, because it names the stage and the part of the building it covers.

TIA-942 facility certificates are real audits. 317 facilities list a TIA-942 rating that reads as a facility certificate, and 27 a design-only one. A facility certificate means a licensed certification body inspected the site. Check the certificate’s date: it runs three years with surveillance audits, so an undated rating, or one issued more than three years ago with no recertification, is not current evidence.

Other schemes are different measures. 30 facilities cite EN 50600, the Japan Data Center Council standard or a TÜV scheme instead. These are legitimate frameworks with their own classes; read them on their own terms.

Claims without a certifier are marketing. 170 facilities, 11% of those stating a tier, describe themselves with phrases such as “Tier III design”, “Tier 3+”, “built to Tier III standards” or “Tier III compliant”, with no certifying body named. Uptime’s system has four tiers and no plus levels. Labels such as Tier III+, Tier 3 ready or Tier III compliant come from operators, not from a certifier, and should be read as marketing until the operator shows a certificate. Some of these buildings may be engineered very well. The point is that the label proves nothing, so the burden of proof sits with the operator.

Add it up and 37% of the facilities that state a tier or rating show no on-site audit by Uptime or a TIA-licensed body in our records. That is not a verdict on the buildings; it is a measure of how often a buyer has to ask for the paper.

Tier III is the market standard, Tier IV is rare

The level being claimed matters as much as the evidence behind it. Splitting the same facilities by the highest level named in their strongest evidence shows how the market is built.

Tier III dominates every column: 498 facilities hold an Uptime award at Tier III for the constructed building or its operations, against 107 at Tier IV. Tier IV costs more to build and run, and certified Tier IV buildings are far rarer than certified Tier III ones; it is a specialist product, priced as one. In the colocation market a certified Tier III hall is the normal high-grade offer; a provider that claims Tier IV should be able to show the award for exactly the hall it is quoting, because a campus can contain halls certified at different levels and halls that were never certified at all.

The table also shows how often a level is claimed without the paperwork to match. Tier IV claims are especially worth checking: Uptime has said publicly that it is highly irregular for a design award and a facility award for the same site to show different levels, and its 2015 statement on design certificates was written in response to a press report that a facility had been marketed as Tier IV without being built to that standard.

Where certificates are common, and where nobody lists them

Tier claims are not spread evenly across the world. The table below lists the countries with the most facilities stating a tier or rating in our records, next to the total number of records for each country.

Two patterns stand out. In several fast-growing markets, India, Indonesia, Brazil and much of Latin America and Southern Europe, a large share of facilities list a tier or rating, and TIA-942 is especially common in India. In the United States, by contrast, only a small fraction of a very large number of records lists any tier at all. That reflects what operators choose to publish more than how their buildings are engineered: a record without a certification means the operator did not list one, not that the building would fail an audit. Where awards are not listed, operators tend to describe the design in their own words instead (“concurrently maintainable”, “N+1 UPS”, “2N power”), and those words carry no third-party check.

For a buyer this changes the conversation by market. Where certificates are common, ask for the certificate and expect to get it. Where they are rare, ask for the design description in writing, the one-line diagrams for the hall, the maintenance procedures and the incident history, and make the service level agreement carry the weight a certificate would otherwise carry.

The same split shows up in the markets where Watt & Rack publishes colocation prices. Follow a market name to its published rate cards.

Tier and price: what the published rate cards say

No rate card in our data prices a tier. Of the 67 published retail prices we track, from 37 operators in 16 currencies, none lists a separate line for Tier III or Tier IV space; the tier is a property of the building the space sits in, and the price reflects it only through the overall rate.

There is a more practical finding in the same data. Of those 37 operators, only 3 appear as the operator of a building in our facility records. The others are hosting companies and resellers selling space in a building someone else runs. Many of them are honest about it, and reselling is a legitimate model for small deployments. But it means the tier claim on a reseller’s page belongs to the building owner, not to the company you sign with. Ask which building and which hall, then check that building’s certificate, not the reseller’s description of it.

At the wholesale end, where deployments are leased per committed kilowatt, CBRE’s asking rents in the 22 markets we track run from USD 160 per kW per month in São Paulo to USD 403 in Singapore at the midpoint (H2 2025 – Q2 2026). CBRE reports those rents by market and requirement size, not by tier, so the certificate is something you verify alongside the price rather than something you can read off it. On the pricing sheet you can see both ends side by side.

A Tier III building does not make your rack Tier III

A tier describes the building’s plant up to the point where power reaches your cabinet. What happens after that point is your design, and it can undo the building’s.

Dual feeds only help dual-corded equipment. A concurrently maintainable building delivers power to your cabinet over two independent paths, usually called A and B feeds. During planned maintenance one of them goes dark. Equipment with two power supplies, one on each feed, keeps running. Equipment with a single power cord, plugged into whichever feed was nearer, goes down with it. Uptime’s 2009 Tier Standard included a requirement for dual-corded devices in Tier III and Tier IV data centers, and a 2014 Uptime Journal article noted that single-corded devices are found in more installations than should be expected. Its recommendation for single-corded equipment is a rack-mounted or point-of-use transfer switch fed from both A and B, and it suggests that colocation providers require dual-corded devices for high-availability deployments.

Each feed must be able to carry the whole rack. With A and B feeds, the normal state is that both share the load. When one goes away the other carries all of it. So the combined draw of the rack must fit within what a single circuit can safely supply, not within the sum of the two. Buyers who size a rack as “two 16-amp feeds, so 32 amps available” discover the arithmetic during the first maintenance window. Size the rack so that one feed can carry everything, and read the provider’s rule for how much of a breaker’s rating you may use continuously; the pricing sheet’s section on amps versus kW shows how published cards state it.

The network is outside the tier. The Tier Standard grades power and cooling. It says nothing about how many carriers are in the building, whether their fibre enters through separate paths, or whether the meet-me room has its own redundant power. A Tier III building with one carrier and one fibre entry is a single point of failure for your traffic. Ask for carrier diversity and entry diversity separately.

Cooling is per hall, and per density. A building certified at Tier III was certified for the loads and the layout in its design. A hall that has since been filled with high-density racks may be running closer to its cooling limits than the certificate’s engineers saw. Ask what density the hall was designed for and what it runs at today; our provider comparison lists the published density figures we hold, and in our records the median published maximum density is 15 kW per rack across the 90 facilities that state one.

Operations cause more outages than topology

The reason Uptime dropped availability percentages in 2009 is visible in its own outage research. Its Annual Outage Analysis 2025 found that outages are becoming less frequent and less severe relative to the growth of digital infrastructure, but the causes it describes are a reminder of what a tier certificate cannot control.

Power remained the leading cause of impactful outages in 2024, behind 54% of them according to Data Center Knowledge’s summary of the report. In the same survey, 54% of respondents said their most recent significant outage cost more than USD 100,000, and 20% said it cost more than USD 1 million. Nearly 40% of organisations had suffered a major outage caused by human error over the previous three years, and 85% of those incidents came from staff failing to follow procedures or from flaws in the procedures themselves. Failures to follow procedure rose by ten percentage points. And over nine years of tracking publicly reported outages, Uptime found that third-party IT and data center service providers accounted for about two-thirds of them.

That last figure is the one a colocation buyer should sit with. When you move into colocation you are choosing a third party to run the plant your systems depend on. The topology on the certificate sets the ceiling; the operator’s discipline decides how close the building gets to it. That is why the Operational Sustainability award, held by 86 facilities in our records, and the Management and Operations stamp, held by 45, are worth more weight than their numbers suggest. Where neither exists, ask the operational questions directly:

• How many engineers are on site at night and at weekends, and are they employees or contractors?
• How is planned maintenance on the A and B feeds scheduled, and how much notice do customers get?
• When was the last full load transfer to generators, and how long did it run?
• Which incidents in the last three years affected customer power or cooling, and what changed afterwards?
• Who holds the maintenance contracts for UPS, generators and cooling, and how often are they serviced?

How to verify a tier claim in fifteen minutes

Verification takes less time than most buyers expect, and it is the single most useful thing you can do before signing.

1. Ask for the award foil or certificate, not a logo. A foil names the facility, the stage (design documents, constructed facility, operational sustainability) and the level. A TIA-942 certificate names the certification body, the rating, the scope and the dates.
2. Read which part of the building it covers. Uptime’s foil specifies the phase, room, pod or module that was certified, precisely so that certified infrastructure can be told apart from uncertified infrastructure on the same campus. Make sure the hall you are being offered is the hall on the foil.
3. Check the date against the validity rules. An Uptime design award issued since 1 January 2014 expires two years after the award date. A constructed facility award is valid until the facility is modified. Operational Sustainability awards are valid for three years at Gold, two at Silver and one at Bronze. A TIA-942 facilities certificate lasts three years with surveillance audits in years one and two; a design certificate, one year.
4. Look it up independently. Uptime lists its awards publicly on its website for confirmation. For TIA-942, contact the licensed certification body named on the certificate. For EN 50600, the certifier’s certificate register.
5. Match the level on every document. The design award, the facility award, the sales deck and the contract should all say the same tier. If the design award says Tier IV and the facility award says Tier III, the building is Tier III.
6. Translate operator language. “Tier III design”, “built to Tier III standards”, “Tier III ready”, “Tier 3+” and “Tier III compliant” are not awards. Ask whether a certificate exists; if not, ask for the design description and the maintenance procedures instead.

Put plainly: ask for the award foil or certificate, read which stage it covers and which building, phase or data hall it names, check its date against the validity rules, and look the site up on Uptime Institute’s public list of awards or with the TIA-licensed certification body that issued it. Then write the certificate into the contract.

What to put in the RFP and the contract

A tier only protects you if it is written down where it can be enforced. In the request for proposal, state the evidence you need rather than the word: “Uptime Tier III Certification of Constructed Facility, or TIA-942 Rated-3 facilities certificate, or EN 50600 Availability Class 3, for the data hall offered; certificate to be supplied with the proposal.” That single sentence filters out every claim on the lower half of the evidence ladder. Our request-for-quotes form asks for the same evidence when you tick a tier.

In the contract, the useful clauses are short:

• Representation. The provider represents that the hall holds the named certificate at signature, with the award number and date.
• Notification. The provider notifies you within a fixed period if the certificate lapses, is withdrawn, or if the plant serving your space is modified in a way that affects it.
• Remedy. If the certificate lapses and is not restored within an agreed period, you may terminate without penalty or move to a certified hall at the provider’s cost.
• Maintenance notice. Planned work on either power feed to your cabinets is announced in advance, with the window and the expected duration.
• SLA definitions. Power availability is measured at your cabinet, per feed, and the credit schedule says what happens when both feeds fail and when one fails outside a notified window.

None of these clauses is exotic, and providers with genuine certificates rarely resist them. Resistance is itself information.

Which tier do you actually need?

Start from the cost of an hour of downtime for the systems you are moving, not from the tier. If your application already runs in two sites and fails over between them, a certified Tier III hall in each is usually the efficient answer, and the money a Tier IV premium would cost buys more resilience as a second site or as better network diversity. If the systems run in one place only and an outage stops revenue or breaks a regulatory obligation, concurrent maintainability is the floor, and fault tolerance is worth pricing.

For development, test and backup targets, a well-run building without any certificate can be perfectly adequate, provided the price reflects it and the contract does not pretend otherwise. What matters at every level is that the claim on the quote is the claim you can prove: the right stage of award, for the right hall, still in date and written into the contract.

When you are ready to compare, our colocation explainer covers the rest of the buying checklist, the city price sheets show what published space costs in each market, and the form below takes your kW, space, location and tier requirement to providers who can show the paper.