Chasing gigafactories: CEE’s AI infrastructure bet
Artificial intelligence is emerging as Central Europe's best answer to an exhausted growth model – but the race to build the infrastructure that could underpin that transformation reveals as much about the region's limitations as its ambitions. The EU's AI Gigafactory competition has quietly produced more cross-border cooperation between Czechia, Poland, and their neighbours than years of political declarations ever managed.
The growth model that carried Central and Eastern Europe through three decades of post-communist catch-up – cheap labour, rising employment, and integration into Western European supply chains – is running out of headroom. The workforce has largely been mobilised, dependence on Western economies has become a source of risk, rising energy prices and strong competition from China and the United States put the current energy-intensive industrial model in question.
As we have previously highlighted, new directions are being explored across the region as policymakers, business leaders, and researchers search for the next engine of growth. These range from strengthening supply-chain resilience – reducing exposure to the German economy – to changing the energy mix toward a larger proportion of nuclear and renewables, and investing heavily in emerging technologies, above all artificial intelligence.
A McKinsey Global Institute analysis published in June 2026, focused on Hungary but relevant across the region, put a figure to that bet: AI could generate at least €15 billion in automation-driven value creation for Hungary alone by 2030, equivalent to around 6 to 7 percent of GDP. The total impact could be larger still if companies move beyond automating existing workflows toward using AI to design entirely new products, services, and business models.
Building the stack: from AI factories to gigafactories
This transformation has received a new boost from the European Union's AI Continent Action Plan, legally anchored in Council Regulation (EU) 2026/150 and administered by the European High Performance Computing Joint Undertaking (EuroHPC JU). The plan channels an EU direct contribution of up to €4.12 billion from Horizon Europe, Digital Europe, and the Connecting Europe Facility, with a total mobilisation target of €20 billion including private and member-state co-financing. It operates across two distinct tiers.
The smaller tier – the AI Factories – are high-performance computing hubs built around existing or new EuroHPC supercomputers, optimised for AI workloads, with capacity of around 25,000 AI-optimised chips per site. They serve as hubs linking supercomputing centres, universities, SMEs, industry, and financial actors, providing access to computing resources for research, startups, and industry. The upper tier – the AI Gigafactories – is a different proposition altogether: large-scale facilities equipped with up to 100,000 AI-optimised processors, capable of handling the complete lifecycle of very large AI models from training through to industrial-scale deployment.
By October 2025, EuroHPC had selected 19 AI Factories across the EU, including facilities in Czechia and Poland. The Czech AI Factory, known as CZAI, is anchored at IT4Innovations National Supercomputing Center in Ostrava and built around the KarolAIna system. Poland has gone further, with two separate AI Factory selections: the PIAST-AI Factory in Poznań, backed by the Poznań Supercomputing and Networking Center, and the Gaia AI Factory in Kraków, operated by Cyfronet AGH. Together, PIAST and Gaia are designed to form an integrated AI ecosystem in Central Europe in collaboration with Finland's LUMI AI Factory. Hungary and Slovakia, by contrast, secured only the lighter-touch AI Factory Antennas – a participation model that allows countries to access existing AI Factory capacity remotely without building their own supercomputing infrastructure.
The competition for Gigafactory status is intense. EuroHPC plans to select seven AI Gigafactories in two categories: medium-scale facilities with around 75,000 GPUs and larger ones with approximately 100,000 GPUs. The funding model fixes the EU contribution at 17 percent of capital expenditure, matched by an equal national contribution, with the remaining two-thirds from private investors. With 76 expressions of interest submitted from 16 member states ahead of the June 2025 deadline, representing over €230 billion in proposed total investment, the field far exceeds available capacity. The formal call for proposals, originally expected in late 2025, has been delayed at least twice; it is now expected in July 2026, with selection to be completed by year-end and the first facilities becoming operational in 2028.
On paper the selection process is open and non-discriminatory, with the EuroHPC regulation requiring evaluation by independent experts and no role for member states in the process. In practice, bids from Germany, France, Spain, and the Netherlands are backed by deeper private capital markets, stronger industrial consortia, and greater state co-financing capacity. For CEE applicants, this means competing on formally equal but materially unequal terms, which is precisely why Poland and Czechia have concluded that a joint submission gives them better odds than going it alone.
Within Central Europe, two distinct strategies have emerged. Romania has pursued a solo bid anchored in energy economics; Czechia and Poland have moved toward a joint submission anchored in political and institutional cooperation.
A joint CEE bid takes shape: Czechia and Poland pool their chances
Poland initially pursued the most diplomatically ambitious approach. In August 2025, Warsaw led a four-country consortium – Estonia, Latvia, and Lithuania alongside Poland – to submit the Baltic AI GigaFactory proposal, valued at approximately €3 billion. The logic was geopolitical: pool the market size and political weight of the Baltic-Nordic digital corridor to strengthen a bid that none of the countries could plausibly win on their own. Estonia and Latvia subsequently withdrew as the European Commission revised its funding mechanics, but Poland and Lithuania continued, and the Czech Republic joined the consortium in November 2025.
Following a bilateral meeting later that month, the Czech and Polish governments agreed to establish joint working groups to prepare a combined Gigafactory application. Czech Industry Minister Lukáš Vlček framed the negotiations as opening the way toward a major CEE supercomputer centre for artificial intelligence. Former AI commissioner Jan Kavalírek made the strategic logic explicit: if the two countries competed separately, they would undermine each other's chances given the limited number of slots, while a joint submission would significantly strengthen the region's position against better-capitalised western consortia.
The Czech government's recent approval of CZK 2.5 billion in state backing in June 2026 is best understood in this context. Under EuroHPC rules, member states must provide a firm national financial commitment before any formal Gigafactory application can be filed, regardless of whether that application is solo or multinational. It is the legal and financial precondition for participation, not the submission itself. The project – formally known as Prague Gateway DC, located on land in Jílovište near Prague developed by České Radiokomunikace (ČRa) – is a commercial data-centre build elevated into a gigafactory bid through state co-financing. ČRa bears the financial risk and serves as owner and operator, committing approximately CZK 70 billion of the total CZK 100 billion project cost. The physical site already has a building permit and a construction timeline – a rare advantage in a competition where most rival bids remain at the planning stage. Whether the final submission will be a Czech-Polish bilateral application or a broader multinational consortium including Lithuania has not been formally confirmed as of mid-2026, but the strategic logic strongly favours consolidation.
Romania's Black Sea AI Gigafactory represents a fundamentally different kind of bid. A dedicated AI infrastructure project valued at up to €5 billion, spanning two sites at Cernavodă and Doicești chosen specifically for their proximity to nuclear power generation capacity, it is built around an explicit energy-economics argument: that stable, low-carbon nuclear baseload power – with a planned energy mix of up to 1,500 megawatts – is the gigafactory's true comparative advantage. The Romanian government mandated implementation in November 2025, with the Ministry of Energy and Ministry of Finance coordinating a proposal supported by World Bank technical expertise. The contrast with the Czech approach is instructive: where Czechia is leveraging a commercial data-centre buildout already underway – the Prague Gateway DC, whose total development cost reaches CZK 100 billion but which would proceed regardless of the Gigafactory competition's outcome – Romania has designed its bid as a purpose-built AI infrastructure project from the ground up.
Beyond borders: AI as a new axis of Central European cooperation
Perhaps the most significant development in Central Europe's AI story has received the least attention: the gigafactory competition is quietly generating more cross-border cooperation than the region's own AI strategies ever managed.
Poland's Baltic AI GigaFactory consortium drew over 270 companies from multiple EU countries into shared working processes. When Estonia and Latvia withdrew, Warsaw and Prague chose to deepen their bilateral cooperation rather than retreat to solo bids – establishing joint working groups, coordination councils, and a growing private-sector network that the region's political declarations had never managed to produce.
Slovakia's incoming V4 presidency, beginning July 2026, may extend this dynamic further, with digital and AI cooperation among the stated priorities. The V4's record on coordinated action is fitful at best, but AI infrastructure may prove more tractable than the ideological fault lines – rule of law, migration – that have historically paralysed the group. Building compute capacity and training technical talent is something that governments of markedly different political orientations can agree on without being required to agree on much else.
