Remanence raises $30M seed co-led by EQT Ventures, FirstMark

· VC Investment

TLDR

  • Paris start-up Remanence raised a $30M seed round co-led by EQT Ventures and FirstMark.12
  • Remanence says it turns companies' AI-agent traces into post-trained open-weight models they own; no customer results are published.23
  • Sifted reported on 21 October the founders sought $30M; no source we could read gives a valuation.4

Remanence, a Paris-based start-up founded by three alumni of the French computer-use AI lab H Company, announced a $30M seed round on 17 November 2026.12 Axios Pro, which carried the story as an exclusive, wrote that the round was led by EQT Ventures and FirstMark, while both investors' own posts say they co-led it.125 Zero Prime, Emblem, Drysdale, Global Founders Capital and Yellow also took part, alongside angels including Datadog's Olivier Pomel, Mirakl's Philippe Corrot, Pigment's Romain Niccoli, Synthesia's Victor Riparbelli and Steffen Tjerrild, and Mehdi Ghissassi of AI71.5 No valuation is given in the pages opened.

Remanence says its platform plugs into a company's existing AI models, converts their execution traces into tailored training environments, and post-trains open-weight models on them through reinforcement learning.5 The customer keeps the model weights, and the platform can run fully managed or in the customer's own cloud; FrenchWeb notes that exporting the model and continuing its training depends partly on the base model's licence.53 "AI companies spend more on intelligence every quarter and own none of it," CEO Pierre-Louis Cedoz said in FirstMark's post.5 The money will fund research, a dedicated GPU cluster and hiring, mainly in Paris with an early presence in London and San Francisco.6 The company says it works with design partners in legal, customer support and enterprise software; FrenchWeb notes no identities, production results or pricing have been published.63

Cedoz, Hamza Benchekroun and Axel Moyal worked together at H Company, where, according to FirstMark and EQT, they led post-training of the open-weight Holo models.52 EQT names Cedoz as CEO, Benchekroun as CTO and Moyal as CPO, while FrenchWeb says the Paris company registration names Benchekroun as président and Moyal as directeur général; EQT also says Holo set the state of the art on the OSWorld-Verified benchmark in March 2026, a claim Wikipedia reports only as "reportedly".28 Sifted reported on 21 October that Cedoz had left H at the start of September to launch the venture with two former colleagues and that the trio was seeking $30M; it carried the completed round on 17 November.47 Wikipedia records that H raised a $220M seed in May 2024; FrenchWeb says FirstMark was in it, a detail Wikipedia's investor list does not mention.83

On FirstMark's side, general partner Matt Turck argues that companies are moving to a hybrid stack of rented closed frontier models and owned, post-trained open-weight ones, and that the hard parts are training environments and scarce talent.9 Turck lists Synthesia, Dataiku, Pigment and Gradium among FirstMark's earlier European backings, and founders of Synthesia and Pigment are angels in this round.9 EQT Ventures partner Tom Mendoza said, in a quote carried by Tech Funding News, that "the next phase of enterprise AI will be about who owns the intelligence."6 EQT Ventures led an $11.9M round in dottxt, a tool for structured LLM output, in October 2024, and backed Pruna AI, which optimises AI models, in November 2024, saying it is "actively investing in AI tooling and infrastructure."1011

The segment is crowded. Bespoke Labs announced seed and Series A financing in July 2026, reported as $40M, to build reinforcement-learning environments for agents.12 Plurai said on 11 November it had raised a $13M seed led by Team8 to simulate company systems and post-train small open-weight models.13 FrenchWeb names Adaptive ML as the most direct competitor among the offers it reviewed, and cautions that the economics depend on volume, training cost and how many cases still need a generalist model.3

Why it matters: a $30M seed for a team with no disclosed customers or valuation shows investors backing the idea that companies will own post-trained open-weight models, though the performance and cost claims are the company's and unproven in production.532

Sources

  1. Remanence raises $30M to help companies train their own models (Axios Pro)
  2. Own, Don't Rent AI: Why We Co-Led Remanence's $30M Seed Round (EQT Ventures)
  3. Trois anciens de H COMPANY lèvent 25,5 millions d'euros pour REMANENCE (FrenchWeb)
  4. Exclusive: Former H Company CTO seeks $30m for new startup (Sifted)
  5. Intelligence That Compounds: Why We Co-Led Remanence's $30M Seed (FirstMark)
  6. Ex-H Company researchers' Remanence raises $30M to turn AI agent data into models companies own (Tech Funding News)
  7. EQT and FirstMark back trio of H Company alumni in $30m seed round (Sifted)
  8. H (company) (Wikipedia)
  9. New Investment: Remanence, Building Intelligence That Compounds (Matt Turck)
  10. Announcing Our Investment in dottxt: From Back-and-Forth Prompts to Truly Dependable Computers (EQT Ventures)
  11. On EQT Ventures Investment in Pruna AI - Making AI Faster, Cheaper, and Greener (EQT Ventures)
  12. Bespoke Labs Raises $40M to Build Environments that Enable Reliable Agents (Bespoke Labs)
  13. Plurai wants to help AI agents work better using the same approach that helped get self-driving cars to roads (TechCrunch)