AI gigafactory in Poland: where should public data stay?
Author: Andrzej Albera, Responsible for strategy and the company's direction
Poland is bidding for one of the European AI gigafactories. It is an investment that may exceed PLN 10 billion and good news for the Polish AI stack. Yet amid the talk of billions and processors, one question is rarely asked, the one every municipal secretary asks: where will residents' data physically be stored, and who will have access to it? We calculated what the gigafactory will give public offices and what offices have to take care of themselves.
How we calculated it
| Step | Value |
|---|---|
| Investment in the Polish gigafactory | may exceed PLN 10 billion |
| Initial capacity | at least 25,000 AI accelerators |
| Target capacity | at least 75,000 accelerators |
| Share bought by the Polish state | 17% |
| Share bought by the EU | 17% |
| Result | approx. 4,000 accelerators for Poland at the start |
That is a lot of power for training Polish models such as Bielik and PLLuM, and for research. But a gigafactory is a power plant for computation, not an archive of case files. Applications, decisions and data with PESEL numbers do not need to go there for an office to use AI.
Poland and the world
| Who | What | Numbers |
|---|---|---|
| European Union | first tender for AI gigafactories | 7 projects: 4 medium-scale and 3 large-scale |
| Poland | bidding for a medium-scale gigafactory | may exceed PLN 10 billion, bids due 12 November 2026, contracts early 2027, operational in 2028 |
| Countries in the region | are to use the power of the Polish gigafactory | Lithuania, Hungary, Croatia and Slovakia; with Czechia, Poland cooperates under a memorandum of September 2026 |
| Poland today | two AI factories are being built | PIAST-AI in Poznań and Gaia in Kraków |
The consortium that wins the tender will decide on the location. The government earlier identified 10 possible sites across the country.
How to read these numbers
- The gigafactory computes, the office decides. Models can be trained in a gigafactory and run where the data is: in the office's own server room or a trusted data centre in Poland.
- Data sovereignty is the decision about where data is stored and processed. That is how Poland's Ministry of Digital Affairs defines it. Every office can make that decision today, without waiting for 2028.
- There is no need to wait for the gigafactory. Polish models already run on ordinary servers today.
What this means for institutions
- Keep the data in-house. TWIN:DESK, our Digital AI Desk, runs in the office's own server room, on Polish models. Residents' data does not leave the building.
- Decide what goes outside. If an office wants to use the gigafactory's power or an external model, PROXY:AI checks every query and lets through only what the office's rules allow.
- Plan wisely. The CDF methodology starts by assessing which data can be processed where. We wrote more about it in Digital sovereignty: one day without your AI vendor.
The gigafactory is an opportunity for Polish AI. But the Polish AI stack does not begin with a hall full of processors. It begins with the decision that an office's data stays in the office. We will talk about it at the 2nd Forum on AI in Public Administration in Wrocław, 4-6 November.
Sources
- Ministry of Digital Affairs, 30.07.2026: the tender for AI gigafactories has started (in Polish)
- Notes from Poland, 28.07.2026: Poland enters race to host EU-backed AI gigafactory
- CyberDefence24, 8.06.2026: the AI gigafactory is getting closer (in Polish)
- Rzeczpospolita, 22.12.2025: is the AI gigafactory in Poland at risk? (in Polish)
- Ministry of Digital Affairs, 28.03.2025: data sovereignty and cooperation with Polish cloud providers (in Polish)