What CoreWeave’s Trajectory Signals for Antimatter
.png)
Neo-cloud: What CoreWeave’s Trajectory Signals for Antimatter
The market has just placed a $63 billion valuation on the "neo-cloud", the category we have been investing in since 2022
Since the beginning of the year, CoreWeave has been racking up major announcements: a $35 billion commitment from Meta, a multi-year partnership with Anthropic, and $6 billion from Jane Street in cloud capacity and equity investment. In the first quarter of 2026, the company reported $2.08 billion in revenue and a backlog of $99.4 billion.
To us, these announcements have a very specific meaning. Wall Street has just validated the AI infrastructure thesis we have supported since our first investments in Hivenet in 2022, followed by Policloud in 2025.
A Six-Year Trajectory: From Atlantic Crypto to Wall Street
CoreWeave was born in 2017 under the name Atlantic Crypto, focusing on Ethereum mining using NVIDIA GPUs. In 2019, its founders sensed a more promising opportunity: VFX studios and early machine learning researchers were struggling to get quick access to GPUs from traditional hyperscalers (AWS, Azure, Google). The company was renamed CoreWeave and fully pivoted toward this new thesis. With the explosion in GPU demand starting in 2022, it became one of the few players capable of delivering H100s at scale.
The milestones followed in rapid succession:
- April 2023: $221M Series B, including $100M from NVIDIA (valuation ~$2B).
- August 2023: $2.3B debt facility led by Blackstone and Magnetar, using GPUs as collateral, a first for the sector.
- May 2024: $1.1B raise led by Coatue, $19B valuation.
- March 2025: IPO on the Nasdaq (CRWV), $1.5B raised. The largest VC-backed tech IPO in the US since 2021.
- September 2025: Initial $14.2B contract with Meta and a $6.5B extension of the OpenAI contract.
Between 2023 and 2025, revenue soared from $229M to $5.13B. CoreWeave is now worth approximately $63 billion on the public market, nearly thirty times its Series B valuation from just three years ago. Players like Meta, Anthropic, OpenAI, and Jane Street, who would have chosen AWS, GCP, or Azure three years ago, now prefer specialized infrastructure. The neo-cloud is establishing itself as a strategic asset class, and capital is following.
Three Companies, One Platform
It is in this context that Antimatter was launched in Cannes on April 21st, following months of work led by David Gurlé to unite three complementary companies:
- Hivenet: A distributed cloud that transforms unused computing capacity worldwide into productive infrastructure. We have supported this thesis since its seed round in July 2022.
- Policloud: A "France 2030" award winner that designs and operates modular, containerized micro-datacenters. Each can house up to 400 GPUs and be deployed in just five months. We signed their pre-seed in April 2025.
- Data Factory: Provides the American anchor of the equation. Co-founded by Richard Esteve and Richard Détente, the company was born in the wake of the 2021 Texas winter storm, which collapsed the state's power grid and revealed a massive opportunity for flexible power farms. These farms can absorb surplus renewable energy and release it back when needed. Today, it owns electrical infrastructure on 50 MW sites across the US, representing a hosting capacity for over 115,000 GPUs.
These three companies now cover the entire value chain: Energy (Data Factory), Modular Hardware (Policloud), and Distributed Software (Hivenet).
Antimatter: Riding the Inference Wave
CoreWeave’s extraordinary trajectory has demonstrated the scale of demand for specialized, independent AI infrastructure. It also signals the next stage of this industry. As CEO Mike Intrator noted, “as the market moves from training to inference, that distinction matters more than ever.”
Inference is the day-to-day use of AI models once they are trained: every query to a chatbot, every action by an autonomous agent, every API call from an application. With the arrival of open-source models like Kimi K2.6, more and more companies want to deploy their own models on dedicated infrastructure, close to their users and their data.
This is exactly the opportunity Antimatter anticipated. The company meets this demand with a simple offer: a server rented at a fixed hourly rate, data that stays with the client, and deployment possible wherever energy is available. This model targets a massive segment (SMEs, labs, administrations, regulated sectors) that will represent the bulk of AI demand in the coming years.
The Logical Next Step
CoreWeave wrote the first public chapter of the neo-cloud story. We are convinced Antimatter will write the next: a vertically integrated, distributed, sovereign platform built for inference and already operational. The demand is massive and durable, and the next wave of this history will be distributed.
We want to salute David Gurlé, Richard Esteve, and Richard Détente for carrying this vision, as well as the entire Antimatter team: Bastien Vidal, Queenie Chan, Guillaume Goualard, Nishka Hemrajh, Sébastien Perret, Romain Barbier, and Delphine Seguin, for building this platform under conditions that would have stopped most. Now, it’s up to Antimatter to ride the wave.
