A 15 MW AI reservation puts Czechia’s largest data centre on the European compute map

German AI infrastructure provider Polarise has reserved 15 MW at Prague Gateway near Prague. The agreement is an important commercial signal for Czechia’s emerging role in sovereign AI cloud and high-density computing.

Czechia’s largest planned data centre has secured its clearest commercial signal so far. German AI infrastructure provider Polarise has signed a letter of intent with České Radiokomunikace, or CRA, reserving 15 MW of capacity at the Prague Gateway data centre in Jíloviště near Prague.

Polarise says the capacity is intended for a customer in the European AI cloud market. The company plans to operate sovereign AI infrastructure and cloud services at the site, while CRA will provide the data-centre environment, connectivity and local infrastructure.

Fifteen megawatts is not a small test deployment. Prague Gateway is currently designed for a total power capacity of 26 MW, meaning the reservation represents approximately 58 percent of the planned base capacity.

The agreement does not yet make Prague a European hyperscale capital. It does show that Czechia is beginning to attract the type of high-density computing demand associated with GPU clusters, AI training and inference, sovereign cloud services and liquid-cooled data-centre infrastructure.

What has actually been signed
CRA and Polarise have signed a letter of intent under which Polarise reserves 15 MW at Prague Gateway and obtains exclusivity for further expansion. This is a substantial commercial commitment, but it should not be described as a completed 15 MW deployment or a final long-term service contract with the unnamed end customer.

Prague Gateway is moving from plan to commercial project

Prague Gateway has been presented as the largest data centre in Czechia and one of the more significant new facilities in Central and Eastern Europe.

CRA’s published plans describe:

more than 2,000 racks,
twelve data halls,
more than 4,000 square metres of data-hall space,
an initial power supply of 26 MW,
infrastructure designed for high-density and liquid-cooled computing,
further expansion potential linked to the Czech AI Gigafactory initiative.

Construction began in August 2026. Skanska signed a contract worth CZK 2.1 billion to deliver the building and non-IT infrastructure. Its announcement gives 2028 as the completion date for the contracted construction work, while Polarise says the facility is expected to be ready for service in phases from mid-2027.

Those dates are not necessarily contradictory. Large data centres are commonly commissioned in stages, with initial halls and power becoming operational before the entire development is complete. Foreign customers should nevertheless confirm which capacity will be available, tested and contractually guaranteed at each stage.

Why the 15 MW figure matters

Data-centre capacity is often discussed through rack counts or building size. For AI infrastructure, power is the more revealing figure.

GPU servers used for training and operating advanced AI models consume far more electricity per rack than conventional enterprise servers. They also produce much more heat. A site may have physical room for additional racks but still be unable to support AI workloads if it lacks sufficient power, cooling and grid capacity.

A 15 MW reservation can support a substantial AI computing environment, depending on the hardware generation, cooling design, redundancy, utilisation and proportion of total facility power available to IT equipment.

The figure should not be converted directly into a promised number of GPUs. Hardware configurations change quickly, and data-centre power includes more than processors alone. But the scale is large enough to move Prague Gateway beyond ordinary regional colocation.

The reservation also reduces one of the central risks for a new data-centre project: building expensive capacity without an anchor user. A credible first customer can support financing, phased construction and negotiations with technology and energy suppliers.

Who Polarise is and what it plans to build

Polarise is a German provider of sovereign AI cloud infrastructure. It builds and operates what it calls AI factories: high-density computing environments designed for GPU-based workloads.

The company says the Prague capacity will become part of its GigaGrid concept, a distributed network of interconnected European AI sites. Individual nodes are typically planned in the range of 15 to 50 MW, with workloads able to move between locations.

This differs from the idea that Europe needs one enormous central campus to compete in AI. Polarise argues for a network of smaller sovereign facilities distributed across European markets.

The Prague Gateway agreement is described as the first concrete site partnership in that expanding network. Polarise already has projects in Germany and has used Nvidia GPU infrastructure in its existing AI operations.

For CRA, the partnership provides an operator that can bring computing equipment, customers and a cloud-service layer. For Polarise, it provides access to a purpose-built Central European site, local connectivity and potential expansion.

Sovereign AI is becoming a commercial infrastructure category

The phrase sovereign AI is sometimes used loosely. In infrastructure decisions, it usually refers to control over where computing takes place, which legal jurisdiction applies, who operates the platform and how data and models are protected.

European companies and public institutions increasingly ask whether AI workloads can run:

inside the European Union,
under European data-protection and cybersecurity rules,
without depending entirely on non-European hyperscale platforms,
with clear control over sensitive training data and model outputs,
on infrastructure that can be audited and contractually governed.

A Czech data centre does not create sovereignty automatically. Hardware may come from the United States or Asia, software can depend on global providers and ownership structures must still be examined.

However, local European infrastructure gives customers more options. It can support private AI clouds, dedicated GPU clusters, regulated workloads, national research and services requiring predictable data location.

Why Czechia can become a Central European AI node

Czechia is unlikely to compete with the largest established data-centre markets only through scale. Its stronger position may be as a connected Central European node serving regional businesses, public institutions, industrial companies and specialised AI providers.

The country offers several relevant advantages:

a central location between Germany, Poland, Slovakia and Austria,
strong fibre routes and regional connectivity,
an industrial customer base using automation, simulation and computer vision,
technical universities and research infrastructure,
demand for Czech and European cloud environments,
lower concentration than the largest Western European data-centre hubs.

Prague Gateway also benefits from CRA’s existing optical backbone and infrastructure experience. CRA already operates eight Czech data centres, while its current AI-server capacity in Lužice has reportedly attracted demand quickly.

The opportunity is not limited to Prague. MasterDC is developing an AI-focused data centre in Kanice near Brno, with an initial 4 MW phase and a total planned site capacity of up to 25 MW. Czechia is also developing the Czech AI Factory around national supercomputing infrastructure in Ostrava.

These projects serve different users and should not be treated as interchangeable. Together, however, they show that Czech AI infrastructure is becoming a multi-location market rather than a single isolated investment.

Prague Gateway and the AI Gigafactory are not the same project

CRA is also involved in Czechia’s attempt to host a European AI Gigafactory. The Czech government has approved preparatory infrastructure work for a proposed project that could require investment of approximately CZK 100 billion.

If the wider gigafactory plan succeeds, published reports suggest that Prague Gateway’s potential capacity could increase from 26 MW to approximately 77 MW.

CRA has also signed cooperation with Spanish quantum and AI software company Multiverse Computing. The company has discussed an investment of CZK 1.2 billion and potential demand for 30 MW if the gigafactory project proceeds.

These numbers should be kept separate:

26 MW is the planned base capacity of Prague Gateway,
15 MW is the capacity reserved by Polarise under the letter of intent,
30 MW is potential demand connected to Multiverse Computing and the gigafactory scenario,
77 MW is an expanded capacity scenario, not the currently completed site,
the European AI Gigafactory remains a separate selection and investment process.

Combining all of these figures as if they were already built or contracted would give a misleading picture.

What the agreement means for suppliers

A high-density AI data centre creates a different supply chain from a conventional office or hosting facility.

Potential opportunities include:

Power infrastructure

Grid connections, substations, transformers, switchgear, backup systems and power-quality management become strategic parts of the project. AI demand can arrive faster than new grid capacity.

Liquid cooling

Modern GPU racks can exceed the cooling requirements of traditional air-cooled facilities. Suppliers of direct-to-chip cooling, heat exchangers, pumps, monitoring and water-management systems should watch the technical specifications closely.

Computing hardware

GPU servers, networking, storage and high-speed interconnects account for a large part of the investment above the building itself. Delivery schedules and access to current hardware generations can determine when reserved capacity becomes useful.

Connectivity and cybersecurity

Distributed AI infrastructure requires low-latency links, resilient networks, workload orchestration and strong security. Prague’s role in a European grid depends on more than the physical data hall.

Operations and maintenance

High-density sites need specialised technicians, remote monitoring, spare parts, hardware servicing and incident procedures. The local skills requirement will grow as the site fills.

Software and AI services

Infrastructure creates value only when customers can use it. Model deployment, managed inference, private AI environments, data engineering and industry-specific solutions will determine whether the computing capacity generates long-term demand.

What foreign AI companies should verify

The announcement makes Czechia more relevant to foreign AI providers, but capacity should not be selected on location and price alone.

Before committing to a Czech site, a company should verify:

the date when its specific hall and power allocation will be operational,
whether the reserved figure represents utility power or usable IT load,
supported rack density and cooling method,
redundancy and service-level commitments,
electricity sourcing and price structure,
network routes and latency to target markets,
physical and cybersecurity standards,
data-location and subcontractor terms,
access to hardware maintenance and spare parts,
options for staged expansion,
responsibility for installing and operating the compute layer.

For sovereign AI services, the legal and operational chain also matters. Customers should know who owns the hardware, who administers the platform, where support teams are located and which entities can access data or models.

The main risks are power, delivery and customer concentration

The 15 MW reservation is commercially important, but several risks remain.

A letter of intent is not final utilisation

The agreement reserves capacity and creates a framework for cooperation. The market should still watch for final contracts, equipment orders and the identity or profile of the end customer.

Power must arrive on schedule

A data centre can be structurally complete while waiting for grid infrastructure. The relevant milestone is tested, energised and usable capacity, not only construction progress.

AI hardware changes quickly

GPU generations, rack density and cooling requirements evolve faster than ordinary buildings. Technical design must remain flexible enough for hardware delivered in 2027 or 2028.

One large customer creates exposure

An anchor customer improves project credibility but can also create concentration risk. CRA will need a broader customer mix if a large share of base capacity is allocated to one infrastructure partner.

Expansion depends on separate decisions

The proposed 77 MW scenario and European AI Gigafactory should not be treated as guaranteed. They depend on policy, financing, consortium structure, energy availability and European selection processes.

What this signals for Czech market entry

For international technology companies, the Prague Gateway agreement changes the Czech market in three practical ways.

First, it provides stronger evidence that large AI workloads can be commercially planned in Czechia. The country is not only discussing AI policy; private infrastructure providers are reserving significant power capacity.

Second, it creates a potential local platform for companies that need European AI computing without building their own data centre. This may be relevant to German, Austrian, Polish and other Central European businesses as well as Czech customers.

Third, it expands the supplier opportunity around AI. The market includes not only software companies, but also power engineering, cooling, construction, networking, cybersecurity, operations and specialised recruitment.

The opportunity should still be approached with realistic milestones. A company entering around Prague Gateway should distinguish announced capacity, reserved capacity, commissioned capacity and actively used computing power.

What to watch next

The strongest evidence of progress will be:

conversion of the Polarise letter of intent into definitive commercial agreements,
commissioning of the first Prague Gateway phase,
confirmed grid and IT power availability,
orders for GPU servers and liquid-cooling infrastructure,
disclosure of additional customers,
technical details for Polarise’s Prague AI Pod,
progress in Czechia’s European AI Gigafactory application,
a final plan for expansion beyond 26 MW,
evidence that Czech companies and institutions are buying the new AI services.

The main conclusion

The 15 MW Polarise reservation is one of the clearest signs that Czechia is entering the European market for high-density AI infrastructure.

It gives Prague Gateway an anchor partner before the data centre is fully operational and connects the Czech project to a wider European sovereign-cloud strategy. At approximately 58 percent of the planned base capacity, the reservation is large enough to influence the technical and commercial shape of the facility.

It is not yet proof that 15 MW of AI servers are installed, that the unnamed end customer has signed a final contract or that Prague Gateway will expand to 77 MW. Those steps require further agreements, power delivery, hardware and commissioning.

But the direction is clear. Czechia is moving from conventional data hosting towards AI factories, GPU infrastructure and regionally distributed European compute. For investors and suppliers, the relevant question is no longer whether this market will appear. It is which part of the infrastructure and service chain they can credibly provide.

How Kodo can help

Kodo helps international technology and infrastructure companies understand and enter the Czech market. We support market validation, partner mapping, Czech and English positioning, local web content, recruitment communication and coordination of the digital and commercial steps before launch.

We do not provide energy, legal, tax or data-centre engineering advice. These areas should be verified with the relevant specialists and authorities.

Contact Kodo

Sources

This article distinguishes planned, reserved and operational capacity. Project parameters and delivery dates may change as construction, contracting and the European AI Gigafactory process continue.