The Uptime Institute Tier level describes the topology of a facility’s engineering systems — it is not a marketing badge. For a rack tenant it turns into a set of very specific contract terms. Here is what actually changes between Tier III and Tier IV.

Formally: two different properties

LevelKey propertyDesigned availability
Tier IIbasic component redundancy99.741%
Tier IIIconcurrently maintainable — maintenance without stopping IT99.982%
Tier IVfault tolerant — survives the failure of any single element99.995%

The difference between III and IV is not the amount of hardware but the scenario. Tier III guarantees you can service any element without stopping the load. Tier IV adds a second scenario: an element fails suddenly, with no preparation — and the IT load does not notice.

In hours: 1.6 hours of designed downtime per year versus 26 minutes.

What that means in practice

Maintenance windows. On Tier III, planned works run without stopping the load, but some of them put systems into an unverified mode — and for that period the facility runs without redundancy. On Tier IV, maintenance happens while fault tolerance is preserved.

Behaviour under failure. The key question for an operator is: “what happens to my rack if one power feed or one chiller fails right now?” On Tier IV the correct answer is “nothing”.

Single points of failure. Tier IV requires physical separation of independent paths: not just two feeds, but different routes, different rooms, no shared nodes. The TCCF audit verifies this through live testing.

Certification: three different things

Distinguish three statuses that are often blurred together:

  • TCDD — certification of design documents. The design matches the level.
  • TCCF — certification of the constructed facility. Assessors come on site and test the real systems.
  • TCOS — operational sustainability assessment, i.e. the quality of running an already operating facility.

The phrase “a Tier IV data center” without a status tells you nothing. The right question is which certificate is already held, which is planned, and for what date.

When Tier III is enough

The honest answer: often. Internal services, test and pre-production environments, workloads with a rehearsed failover to a backup site — all of these live comfortably on Tier III. The price difference is justified where the cost of downtime exceeds it: financial services, payment and billing systems, regulated workloads, personal-data processing, and infrastructure with no second site in the region.

What to put in the SLA

The Tier level itself does not go into the contract — the terms derived from it do:

  • guaranteed availability and how it is measured;
  • the planned-works procedure: are windows agreed with the client;
  • compensation mechanics for unavailability;
  • committed power per rack and the cooling scheme;
  • remote hands response times and the escalation path.

How this looks at Akashi

Akashi Data Center is designed as Central Asia’s first commercial Tier IV facility: two independent 110 kV feeds in an active/active scheme, UPS with 5/4N redundancy, 12 hours of generator autonomy at full load, four independent optical entries. Designed availability is 99.995%.

Certification status: the design documentation is validated (TCDD), the constructed-facility audit (TCCF) is scheduled for April 2027, and the first phase launches in 2027.

Frequently asked questions

What is the difference between Tier III and Tier IV? Tier III means maintenance without stopping IT. Tier IV additionally survives a sudden failure of any single element with no impact on the load.

What is the designed availability of each? 99.982% (about 1.6 hours per year) versus 99.995% (about 26 minutes per year) — topology figures, not an operator’s promise.

What do TCDD and TCCF mean? TCDD validates the design documents; TCCF validates the built facility after live testing. TCDD without TCCF means the design, not the building.

When is Tier III enough? For internal services, test environments and workloads with a rehearsed failover elsewhere. Tier IV pays off where downtime costs more than the price gap.

What part of the Tier level ends up in the contract? Not the number, but the derived terms: availability, planned-works procedure, compensation, power per rack, remote hands response times.


Need to work out which level your workload requires? Contact sales — we will go through the profile and the SLA requirements.