Data centres in Iran: documented sites, operators and capacity

Data centres in Iran: documented sites, operators and capacity

For anyone assessing data-centre supply in Iran, the first decision is not which operator has the largest footprint. It is whether the evidence supports a site-level comparison at all.

A current review of public operator material shows that Asiatech's Cloud.ir service continues to advertise hosting in three Iranian data-centre locations: Milad Tower, Azadegan and Mir Emad. Its detailed Milad Tower page describes a physical facility in Tehran, refers to approximately 246 racks across 7,000 square metres, and lists cooling, backup power, fire protection, physical security and network infrastructure. The same page also markets a 10-gigabit port for the service.

Those facts are useful, but they are not a published IT-capacity figure. Rack count, floor area and network-port speed describe different parts of the facility. None can be converted directly into megawatts without documented rack loading, utilisation, electrical topology and cooling design.

That distinction matters because a market can look measurable while still lacking comparable capacity disclosure.

What is documented today

PowerlandMap currently has one Iranian facility that passes its public site-level verification threshold: the Asiatech Milad Tower Data Center in Tehran. The record is supported by the operator's own Cloud.ir material and was last rechecked on 9 November 2026.

The operator material establishes several points:

• Milad Tower is identified as a physical hosting location in Tehran.
• Cloud.ir describes active cloud, virtual-server and hosting services supported by the site.
• The detailed page reports approximately 246 racks and 7,000 square metres.
• The operator describes cooling, backup power, fire protection, network systems and physical security.
• Cloud.ir also names Azadegan and Mir Emad as other Iranian hosting locations used by its platform.

The evidence does not establish site-level IT MW, total electrical MW, current occupied rack count, usable colocation capacity, power density by rack, opening date or an audited uptime result.

PowerlandMap therefore records Milad Tower as live and publicly verified at city precision, but leaves capacity blank. Azadegan and Mir Emad remain operator-named locations rather than separate verified site records until their identities, addresses and site-level characteristics can be matched unambiguously.

That is the correct boundary between a service statement and an infrastructure inventory.

Why 246 racks is not an MW figure

The Milad Tower page provides the most concrete physical number in the reviewed material: approximately 246 racks. It is tempting to turn that into capacity, but any result would depend almost entirely on an assumed rack density.

The table below is a sensitivity calculation, not a statement of installed capacity.

The formula is simple: rack count multiplied by average IT load per rack. But the missing inputs are decisive. The public page does not state how many racks are installed, energised, occupied or available. It does not disclose a contracted power envelope, a measured peak, a design density or a power usage effectiveness value.

Even if every rack were installed and energised, the electrical input would still be higher than the IT load because cooling, power conversion, lighting and other systems consume energy. Without a documented PUE or site electrical boundary, total electrical MW cannot be calculated responsibly.

The useful conclusion is not that Milad Tower has between 1.23 MW and 4.92 MW of IT capacity. It is that a single published rack count produces a fourfold range under reasonable illustrative densities. Capacity should remain unknown until the operator or a technical filing supplies the missing evidence.

The operator market is broader than the verified site inventory

RIPE NCC's current membership directory lists a wide set of organisations offering Internet registry services in Iran. The roster includes Asiatech Data Transmission Company and many other network and technology entities. Individual entries also identify organisations such as Fanap and Datak with Iranian service areas and Tehran addresses.

This is relevant market context, but it is not a data-centre census.

A RIPE membership entry can help establish that a company operates network resources or serves the country. It does not prove that the company owns a data centre, identify a specific facility, or disclose racks, electrical design or available capacity. The same caution applies to cloud-product pages, hosting offers and CDN locations: they can reveal service availability without proving the physical site behind every service endpoint.

For a buyer, developer or enterprise customer, this creates two separate research tracks:

1. Network and service presence. Which operators provide cloud, hosting, connectivity, IP resources or managed services? 2. Physical infrastructure. Which named facilities can be tied to an address, operator, operational status, rack or power disclosure and dated source?

Combining those tracks too early creates duplicate sites and inflated capacity. Keeping them separate produces a smaller inventory, but one that can be defended.

What a bankable comparison would require

A reliable comparison between Iranian data centres would need the same fields for each facility and the same measurement boundary.

At minimum, the evidence pack should include:

• the legal operator and, where different, the asset owner;
• a stable facility name and physical location;
• operational status with a dated source;
• installed and available racks, clearly distinguished;
• contracted or installed IT power;
• total electrical intake or transformer capacity, separately labelled;
• rack-density limits and any high-density zones;
• redundancy description for utility feeds, UPS and generators;
• cooling architecture and design conditions;
• network-carrier availability and meet-me-room arrangements;
• certification scope, including the exact site and validity period.

A service brochure can support part of this list, but it rarely supports all of it. Technical specifications, tenders, certification records, operator engineering documents and customer due-diligence material are usually needed to close the gaps.

For Milad Tower, the next decisive evidence would be a dated technical document that links the approximately 246 racks to installed electrical infrastructure and current usable capacity. For Azadegan and Mir Emad, the first requirement is more basic: an unambiguous site identity with operator confirmation.

The commercial implication

The lack of comparable MW disclosure does not mean the market lacks operating infrastructure. The reviewed material clearly supports active cloud and hosting services in Tehran and names multiple hosting locations. It means that public evidence is better at proving service availability than at proving capacity.

That changes how the market should be approached.

An enterprise customer should not begin with a ranking based on estimated MW. It should begin with workload requirements, network reach, location, resilience, support model and the documents each operator can provide. A developer or investor should separate service demand from physical expansion capacity and avoid using a broad network-operator list as a proxy for buildable supply.

The objection is straightforward: waiting for perfect disclosure can exclude real facilities from an early market scan. That is true. The practical answer is to keep two layers. A broad candidate layer can contain operator-named locations and service evidence. A verified layer should contain only facilities with a unique identity and facts supported at site level.

This preserves coverage without presenting inference as measurement.

What to watch next

Three developments would materially improve the picture.

First, operator publications that disclose installed racks, occupied racks and power at the same date would make site comparisons possible. Second, certifications or technical tenders tied to a named facility would strengthen evidence on resilience and operational scope. Third, clear operator pages for Azadegan, Mir Emad and other named locations would allow service references to become verified physical records.

Until then, the defensible statement is narrow: Asiatech's Milad Tower facility is a documented live data-centre site in Tehran; the operator reports approximately 246 racks and 7,000 square metres; Cloud.ir names two additional Iranian hosting locations; and no reviewed primary source establishes site-level IT MW for these locations.

That is enough to map a verified starting point. It is not enough to claim a complete national inventory or an MW market total.

Sources and method

Primary operator sources reviewed on 9 November 2026:

• Cloud.ir — Milad Tower virtual server and data-centre features
• Cloud.ir — Asiatech cloud demo and hosting locations

Network-market context:

• RIPE NCC — members offering services in Iran
• RIPE NCC — Fanap member entry
• RIPE NCC — Datak member entry

PowerlandMap method: site identity, operator, status and physical facts are published only when the source supports them at facility level. Rack count, floor area, network speed, IT MW and total electrical MW are kept as separate fields. The sensitivity table is an illustrative calculation and is not stored as actual site capacity.