F/AI Deal Day: What We Learned in 3 Months Inside Europe's Top AI Program

On April 9, Rippletide took the stage at Station F for the Deal Day of the first F/AI cohort. Three months into the program, here is what we took away, and why it matters for our customers, partners, and the agentic ecosystem at large.

Why F/AI is different

F/AI was launched by Station F to support around twenty early-stage AI startups with a clear objective: reach one million euros in revenue in under six months. This is not a networking program. It is an execution program.

Selection is recommendation-only. 80% of founders in the first batch are repeat entrepreneurs, and a third have already sold a company for at least 10 million euros. That context changes the dynamic. No beginner pitches, no theory. People who have built before, who are starting again, and who know exactly why.

F/AI is designed to have in Europe the same impact that Y Combinator has had in the United States. That is ambitious. But when you look at the partners (Anthropic, OpenAI, Google, Microsoft, AWS, Mistral, along with Sequoia Capital, General Catalyst, and Lightspeed on the VC side), you understand this is not an empty promise.

What the program meant concretely for Rippletide

The program gives access to over one million dollars in compute credits, along with direct relationships with the technical teams at Anthropic, OpenAI, AWS, Microsoft, Google, and Mistral.

For us, that translated into three concrete levers:

Accelerated integrations. Having a direct contact at each platform, not just API access, changes the depth of what you can build together. Our integrations with Claude Code and AWS Bedrock AgentCore benefited from technical feedback that documentation alone would never have provided.

Problem validation. F/AI selects for the hardest unsolved problems in AI. Being selected on this positioning (decision runtime, pre-execution enforcement) confirms that agentic infrastructure is no longer a niche topic.

A high-quality peer network. Twenty startups selected through a recommendation-only process, with 34 million euros raised collectively. These founders become natural partners, not competitors.

What we saw on stage on April 9

The Deal Day is not a typical demo day. The program concludes with a Deal Day where startups pitch to corporates to sign partnerships. The objective is commercial, not about media coverage.

The most interesting conversions we observed in the cohort came from startups that already had real traction: customers in production, defensible metrics. That is exactly the dynamic we have been building at Rippletide since our first deployments with Renault, Pernod Ricard, and CircleCI.

Rippletide operates between San Francisco and Paris. That is how we combine speed and determination on one side with rigor, research, and depth of talent on the other. F/AI anchored that duality even more firmly.

What 2026 confirms for agentic infrastructure

2026 is the year agents move from proof-of-concept to real enterprise production. We see it in every conversation with our customers: the question is no longer "does it work?" but "how do we deploy it without taking on risk?"

That is exactly the problem Rippletide solves. Not monitoring. Not guardrails in the prompt. A decision runtime that sits before execution: deterministic, auditable, blockable.

The LLM proposes. Rippletide decides.

What comes next

F/AI is ending. Our work is accelerating.

If you want to understand how we deploy this infrastructure in an enterprise context, start with our dedicated F/AI page.

If you are building with Claude Code, see how Rippletide enforces your agent rules in production: rippletide.com/dev

Patrick Joubert, CEO & co-founder, Rippletide