Remanence raises $30M seed to train specialized enterprise models
Three H Company alumni raised $30M from EQT Ventures and FirstMark to turn enterprise agent traces into specialized models that companies own, weights included.
Also see more funding coverage: /funding/
Paris, France
Paris-based Remanence has raised a $30M seed round co-led by EQT Ventures and FirstMark to train specialized open-weight models from the traces of the AI agents that companies already run.
Thirty million dollars, two co-leads, one announcement day
Remanence announced the round on 17 November 2026, with EQT Ventures and FirstMark leading together. Zero Prime, Emblem, Drysdale, Global Founders Capital and Yellow also invested, alongside angels who include Datadog CEO Olivier Pomel, Mirakl’s Philippe Corrot, Pigment’s Romain Niccoli and Synthesia co-founders Victor Riparbelli and Steffen Tjerrild (Tech.eu and ActuIA, 17 Nov 2026).
Get the Weekly brief
The week’s rounds, last calls and events, every Monday.
The company says the money goes to research and compute, including a dedicated GPU cluster, plus reinforcement learning environments and evaluation tools (ActuIA, 17 Nov 2026). Hiring is mostly in Paris, with first offices planned in London and San Francisco. In our fundraising data the round is logged at €27.2M, converted at the ECB reference rate of 1.1014 dollars per euro (16 Nov 2026).
Three H Company alumni and a pipeline built on agent traces
Remanence is led by CEO Pierre-Louis Cedoz, with Hamza Benchekroun and Axel Moyal as co-founders. All three worked at H Company, the Paris AI startup founded by DeepMind veterans, where they led post-training for its open Holo models. Their earlier employers include Google DeepMind, InstaDeep and Meta (ActuIA, 17 Nov 2026).
The product plugs into the models a company already uses, turns their execution traces into training environments, and post-trains open-weight models on them with reinforcement learning. It retrains as new traces arrive. The customer keeps the model weights and can keep its data on its own infrastructure. Design partners sit in legal, customer support and enterprise software, in Europe and the US. Cedoz sums up the pitch: “We started Remanence to close that gap and help businesses turn their own agent data into specialised models they control” (Tech.eu, 17 Nov 2026).
The company’s site calls Remanence a continual-learning platform for companies running agents in production, under the line “Turning agent execution into compounding intelligence.” Its pipeline covers trace analysis, environment synthesis, long-horizon reinforcement learning on open-weight models, and inference. Remanence claims the resulting models handle multistep business tasks more reliably, and possibly at lower cost, than general-purpose ones (Tech.eu, 17 Nov 2026). No customer results are published yet, so that claim stays a claim for now.
An $8B agent market growing 43% a year
Remanence sells to companies that already run agents, so agents are the market to size. Research and Markets puts it at $8 billion in 2025 and $48.3 billion by 2030, a 43.3% CAGR (report published 24 Nov 2025). That is one analyst house’s forecast. No credible sizing exists yet for the narrower segment of post-training services sold to enterprises.
The round also has neighbors in our own data. Sumary raised a $25M seed for AI accounting on 10 November, and Nebuly raised a $21M Series A on 12 November. A $30M seed for a company with no disclosed founding year or headcount sits above both.
Owning the weights is becoming a fundable position
A TCS study published in early November found that 94% of French executives consider independence from AI models essential (as reported by ActuIA). The reality is that businesses keep paying for proprietary models from US labs, which is the dependence Tech.eu says Remanence targets. The silver lining is that open-weight models give them something to retrain, and the first checks for that idea came from EQT Ventures and New York’s FirstMark, with a founder-angel list that reads like a roll call of European software.
What to watch is whether the legal, support and enterprise software design partners turn into paying customers, and whether London and San Francisco open on schedule. Imagine if every French mid-cap owned the weights of the model that runs its claims desk. The opportunity is clear. Let’s build it!