By Hamza Faisal

Europe's largest independent cloud is about to become an AI model maker. OVHcloud said on Tuesday it will launch its first proprietary AI model in early 2027, a mixture-of-experts design that has been training since June on JUPITER, Europe's fastest supercomputer.

The announcement came from CEO Octave Klaba after OVHcloud published its annual earnings report for the year to August 31. The French company posted revenue of €1.15 billion ($1.29 billion), up 7.2%, and — more importantly for a company that has been burning cash since its 2021 listing — its first positive levered free cash flow: €24.7 million, against a negative €27 million a year earlier.

Adjusted EBITDA rose 9.7% to €480.5 million, a 42% margin, while public cloud revenue jumped 18.9% to €255.1 million. For fiscal 2027, OVHcloud is targeting comparable growth above 10%, an EBITDA margin above 42.5% and positive free cash flow again.

Spare capacity as a weapon

Klaba also revealed something his competitors cannot claim: about half of OVHcloud's connected cloud capacity is sitting ready for new clients. The company has 250 megawatts of capacity connected, 140 MW in operation, and 110 MW ready to take on new business — meaning it can onboard AI customers without spending on construction while rivals must keep building.

Ahead of the model itself, OVHcloud will release a pilot of OVHai Workspace, a productivity platform tuned on the company's own internal data and running workplace apps like email. The mixture-of-experts model follows in early 2027.

Europe's challenger to Mistral

Klaba has been positioning OVHcloud as a European challenger to Mistral, France's AI darling, and training on JUPITER gives the claim some weight. Unlike the American hyperscalers' models, OVHcloud's pitch is sovereign European AI: models built in Europe, on European infrastructure, for customers who cannot or will not send their data to US clouds.

Whether a cloud operator can become a model lab is an open question, but OVHcloud's discipline-first year — Klaba called it a return to 'fundamentals' — gives it the balance sheet to try.

The gigafactory jab

Klaba used the moment to renew his criticism of Europe's state-backed computing plans. The financing for the EU's planned AI gigafactories — half EU, half member state — naturally pushes projects toward national initiatives, he said, producing 'a race of national data centre projects, none backed by a customer base to match.'

He has called national AI gigafactory strategies 'doomed to failure' and said a proposed €200 million public commitment to buy AI services over five years was too small to underpin the required private investment. His argument: national demand is too weak; Europe must operate at a continental scale.

By Hamza Faisal