OVHcloud AI Summit 2026: “A year ago, it was a vision. Today, it’s a reality”
More than 2,000 customers, partners and developers gathered at the Carrousel du Louvre on 19 November for OVHcloud’s 13th Summit: the OVHcloud AI Summit. Beyond the announcements, the keynote set a clear course: AI that acts, built in Europe and designed to remain under the control of those who use it. Here is a look back at Octave Klaba’s vision, the standout moments and the customers who took the stage.

After the era of experimentation, the era of production AI
Opening the keynote, Octave Klaba, founder and CEO of OVHcloud, set out his vision: it is in agentic AI that artificial intelligence reveals its true value. AI that reasons, acts and can work on a task for hours, even days. “The added value delivered by AI is now there, and the prices are justified.” The timing is right.
After years of experimenting with generative AI, businesses are starting to measure what it concretely changes. The cost of inference has fallen sharply, and more compact open models, such as DeepSeek-R1 in early 2025, have shown that a giant model is no longer essential to achieve top-tier results.
Yaniv Fdida, Chief Product & Technology Officer at OVHcloud, summed it up on stage: “The future of inference isn’t one giant model that does everything. It’s the right, specialised model for each business need.”

The challenge is turning that intuition into production. That is the promise OVHcloud came to the Louvre to make: bringing together, with a single European provider, everything needed to run these models and agents, without locking customers in or exposing their data.
“All the cards are on the table”
For Octave Klaba, AI is opening a period that is “both exciting and destabilising”: the world is accelerating, but visibility is lacking. “All the cards are on the table”, and the difference will come down to two things: vision, to read the weak signals, and speed, to prove you are moving in the right direction. For OVHcloud, this Summit marks the starting point of an acceleration: after web hosting, dedicated servers, private and then public cloud, it is now the turn of AI and quantum.
This acceleration begins with an admission. Over the past three years, many customers have asked for GPUs without volume constraints. “Yes, we were short of them,” Octave Klaba acknowledges, standing by the choice to reserve them for existing customers rather than use them as an acquisition lever. That chapter is now closed: with NVIDIA B300 and Rubin GPUs, available in Bare Metal and Public Cloud across three continents, and 110 megawatts available in its datacenters, OVHcloud can now supply up to 60,000 B300 or 35,000 Rubin GPUs, for economically viable projects.
The other pillar is freedom of choice. OVHcloud’s AI can be consumed in several deployment modes, cloud or on-premises, in different configurations tailored to each customer’s sovereignty requirements.
This independence also means being able to design its own AI building blocks, rather than relying on other people’s. That is the role of OVHai’s three research labs, dedicated to language models, voice and document search, which stem notably from the acquisitions of Dragon LLM, now OVHai LLM, and Gladia.
“AI isn’t a product. It’s a stack”
Taking over, Yaniv Fdida started from the questions every organisation investing in AI is asking: where is my data, who controls the chain, how much will it cost in six months, and what happens if the rules change overnight? His answer: think of AI as a stack, from the electricity that powers the GPU to the application you open every morning. A stack that OVHcloud masters end to end, since the company manufactures its own servers, builds its own datacenters, water-cools them and operates its own network. “A European, open, reversible and predictable AI isn’t a slogan. It’s an architecture.”
The keynote climbed this stack layer by layer, from infrastructure to applications, with 55 new offerings along the way, detailed in the article “OVHcloud AI Summit 2026: 55 new offerings to take AI into production”. Among the demonstrations, AI Fabric, designed with Blade.sh by Shadow, launched two hundred coding agents in parallel across two hundred isolated sandboxes. As Yohann Prigent, CTO of Blade.sh, put it: “An agent without an execution environment is just a chatbot.” Two further demonstrations showed what comes next: AI Agents, which hosts, secures and monitors agents once they move beyond the prototype stage, and AI Search, which gives them access to the right passages across millions of company documents.
Three highlights
Olala, already in production
The announcement first drew smiles from the audience: “Yes, yes, we really did call it Olala,” Octave Klaba said as he unveiled OVHcloud’s first LLM. Behind the name lie fifteen years of research led by Olivier Debeugny at Dragon LLM, now OVHai LLM. Designed for efficiency, Olala mini runs on a simple CPU server, and it is already in production at Qwant, where it generates the instant answer for 3 million daily queries, using three times fewer resources than its predecessor. The thesis set out at the opening, backed by the numbers. “Now we’ve entered the arena,” says Olivier Debeugny, for whom “AI is becoming core infrastructure, just like the cloud”.
Voice, tested on its own support
Formerly head of data and AI at OVHcloud and now founder of Gladia, Jean-Louis Quéguiner came “back home” with a conviction: in Europe, voice AI must understand accents, mixed languages and that reflex of starting a sentence in English and finishing it in your own language. OVHcloud put it to the test on its most demanding use case, telephone support: since 1 September 2026, Gladia voice agents have been handling part of the calls. Octave Klaba sees more in it than an efficiency gain: “Tomorrow, we hope, our AI-augmented support will be our best tool for acquiring new customers.”
OVHai Workspace: AI that works “for you, not on you”
The final layer brought the audience back to everyday use. Guillaume Gojard, Product Director, and Timothée Rebours, Team Leader, demonstrated an AI that summarises a twenty-email thread, finds a common time slot or catches you up on a meeting that started without you. That left the question on everyone’s mind: can you let an AI read your emails, documents and meetings? Guillaume Gojard answered in a single sentence: “We make AI work for you, not on you.” Data stays in Europe, is not used to train the models, and the entire Workspace will be end-to-end encrypted in 2027.
Nine customers, nine ambitions
After the technology came what it makes possible. Hosted by Pierre Byramjee and Bruno Ronsse, this segment brought nine customers on stage.
AIFE, the French State’s financial IT agency, operates the public e-invoicing platform, which began its rollout in France in September. It has entrusted the platform to OVHcloud’s Bare Metal SecNumCloud offering, to eventually handle 4 billion flows. Its director, Emmanuel Spinat, was clear: the time for proofs of concept is over; now is the time for AI services in production.
IGN, the French national mapping agency, hosts its Géoplateforme with OVHcloud, capable of serving up to 55,000 requests per second, and is preparing digital twins of French territories to help decision-makers anticipate the effects of climate change.
Skaleet, Ipsos Synthesio, Wiremind, Sekoia, Kbrw, La Poste Santé et Autonomie and Luganodes completed the picture, spanning banking, healthcare, cybersecurity and logistics. Very different businesses, all coming back to the same question, put by Asha Sumputh, master of ceremonies for the keynote: who do you trust with your data?

“We build the future together”
Returning to the stage to close, Octave Klaba summed up the road travelled: “A year ago, it was a vision. Today, it’s a reality.” One more step on a journey guided, for nearly 30 years, by the same conviction: “We must be able to remain free and independent in our technology choices.” Before ending with a promise: “We still have plenty to show you. We build the future together.”

After the keynote, the Summit changed pace. From 1:30 pm to 6:30 pm, more than 60 breakout sessions took over, led by more than 100 speakers from OVHcloud teams and their partners. Ranging from introductory to advanced level, they explored the issues that shape AI adoption: sovereignty, cost control, cybersecurity, data, multicloud, environmental impact and quantum. Technical teams and decision-makers alike came looking for concrete answers to the question raised that morning: how do you move from experimentation to production?
Find out more