Octave Klaba, chief executive officer of OVHcloud, warned that European AI "gigafactories" would need annual revenues of €300–400 million just to reach break-even. He said existing public procurement promises would cover only a fraction of that requirement, making investments unlikely to be commercially viable if they rely solely on national markets.
The basis of Klaba’s concern
Klaba referred to specific program parameters and figures: the EuroHPC initiative is seeking industrial consortia to set up up to seven high-performance computing centers intended for training, fine-tuning and running AI models. The public-sector procurement commitments in the scheme amount to €200 million in total over five years — a level Klaba describes as insufficient to trigger the necessary private investment.
According to his calculations, public actors would purchase roughly €40 million of compute capacity per year, while a single gigafactory operator would spend at least €600 million on graphics chips alone. That gap means state orders would cover only about 10–15 percent of the expected annual revenue target.
Rapid asset depreciation and the limits of national strategies
Klaba also stressed that the required hardware — notably GPUs — lose value quickly, which further complicates the prospect of recouping large upfront investments. For this reason, he argued, country-by-country financing and market approaches cannot create a viable industry: individual national markets are too small to sustain the profitable operation of large AI-optimized data centers.
He concluded that a Europe-wide market and demand are necessary, and that purely national strategies are likely to fail. Klaba added that because funding and procurement frameworks remain nationally focused, OVHcloud was unable to join some French consortia that have already formed.
Hungary’s contribution and the regional picture
The article notes Hungary's involvement in the EuroHPC effort: the Hungarian government announced a €500 million concessional loan facility to support AI developments, and about €125 million of that could be committed to the EuroHPC gigafactory initiative, potentially granting Hungary access and stakes in the project later on.
While that contribution is not negligible from a Hungarian perspective, Klaba maintains that per-country approaches are insufficient; national demand alone will not guarantee the scale of annual revenue needed, reinforcing his call for integrated, Europe-level demand and strategies.
Why this matters
Building significant AI compute capacity requires very large capital expenditures, and without credible routes to revenue private investors are unlikely to provide the necessary funds. That could slow Europe’s ability to build independent AI infrastructure and increase reliance on external technology providers. Klaba’s analysis is therefore a key consideration for policymakers and industry when designing funding and procurement models for EuroHPC and similar initiatives.
