The rapid spread of generative artificial intelligence is sharply increasing electricity demand from data centres, reshaping where global tech firms place major investments. The International Energy Agency (IEA), in its April 2025 report Energy and AI, estimates that global data centre electricity consumption could reach 945 terawatt‑hours (TWh) by 2030 — roughly double current levels and exceeding the annual electricity consumption of some countries. Goldman Sachs Research’s 2024 forecast is even more dramatic: AI‑related data centre electricity demand could rise by up to 160 percent by 2030 compared with 2023.
These are not abstract risks. Training a large language model (LLM) can consume as much energy as a small city uses in a year, and the systemic shock will come from inference — the continuous, real‑time operation of models to handle user queries. Inference generates a 24/7 base load rather than intermittent peaks, requiring different grid guarantees than those provided by solar or wind alone.
Hyperscalers are becoming energy players
Microsoft, Amazon and Google are increasingly large energy purchasers rather than just tech companies. Their strategic moves include Microsoft’s September 2024 twenty‑year power purchase agreement (PPA) with Constellation Energy tied to restarting the Three Mile Island Unit 1 reactor, aimed specifically at powering data centres. Amazon and Meta have pursued similar approaches, exploring small modular reactors (SMRs) and geothermal projects.
These examples show that for hyperscalers the green transition and grid stability are survival issues: they need jurisdictions that can provide guaranteed, reliable baseload power under clear regulatory frameworks.
Hungary’s paradox: a solar boom paired with baseload scarcity
Hungary has become a European solar success story: according to MAVIR Zrt. data, installed domestic photovoltaic capacity approached 9 gigawatts (GW) by 2027. That is a strong regional achievement, but it alone does not make the country an obvious choice for premium AI data centre investment.
High‑value AI workloads require reliable, dispatchable baseload capacity. Hungary faces critical constraints here: the operational lifetime of Paks Unit 1 is limited, construction of Paks 2 has been delayed, and conversion of the Mátrai Power Plant carries strategic and financing uncertainties. The transmission system is congested in several regions, and requests for grid connections in industrial parks often wait years. The rapid increase in solar generation has made the grid more sensitive to fluctuations, forcing industrial users to consider expensive battery systems or gas backup if the grid cannot provide firm capacity.
Neighbouring countries have advantages in the race for large investments
When a U.S. or Asian tech giant evaluates a 50–100 MW premium data centre in Central and Eastern Europe, decision criteria increasingly emphasize grid guarantees and green energy reliability rather than labor costs.
- Slovakia can offer predictable baseload backed by new blocks at the Mochovce nuclear plant and stable hydro resources.
- Poland is investing aggressively in offshore wind and gas infrastructure and benefits from links to the Baltic region and a large domestic market.
- The Baltic states and Finland already attract data centre activity due to cold climates that lower cooling costs, extensive district heating systems and reliable northern wind power.
Hungary risks seeing premium foreign direct investment flow elsewhere while retaining lower‑value, energy‑intensive industrial projects.
Battery factories vs. compute centres: an allocation dilemma for capital
Hungary’s industrial policy has successfully prioritized battery manufacturing and automotive investments, which bring sizable employment and fast payouts. But those projects are water‑ and logistics‑intensive and rely more on mid‑skilled labor, whereas a high‑end AI data centre uses less water and physical infrastructure but attracts the highest value from knowledge‑intensive roles (AI engineers, data centre operators, cybersecurity experts).
If constrained grid capacity and preferential industrial electricity tariffs are consumed primarily by heavy industry, Hungary could miss out on the Compute‑Energy industry that will drive future high‑value investments.
Practical steps for Hungary
Addressing the challenge requires more than tax incentives. Key measures proposed include:
- A National Data Centre and Energy Program: a dedicated strategy to align data centre development with grid planning and capacity expansion.
- Plug‑and‑play industrial zones: designated areas where MAVIR Zrt. and distribution companies pre‑guarantee 24/7 supply backed by green PPAs to lower investor risk.
- District heating synergies: linking Budapest’s and regional district heating systems and industrial waste heat producers to data centres so that waste heat from server cooling can be fed into district heating networks, improving environmental performance and operational efficiency.
- SMR strategy preparation: although SMRs are a 10–15 year prospect, legal and infrastructure planning should begin now to enable deployment near industrial parks and data centre clusters.
In the 21st century, attracting operating capital increasingly depends on the condition of electrical switchboards and guaranteed baseload rather than solely on labor markets. Without structured, grid‑level responses to AI’s energy appetite, Hungary risks falling behind in attracting some of the decade’s most valuable infrastructure investments.
The battle for kilowatts has already begun; the question is whether Hungary’s grid can withstand the strain.
