Antimatter Revolutionizes AI Inference with Its Micro Data Centers

Antimatter: A New Era for AI Inference

Antimatter, a company recently formed from the merger of three specialized firms, is launching with the ambition to transform the landscape of AI inference. This Tuesday, it unveiled its neocloud project, a global network of micro data centers specifically dedicated to AI inference.

The three entities behind Antimatter are Datafactory, Policloud, and Hivenet, each bringing unique expertise. Datafactory focuses on energy infrastructure in the United States, Policloud designs modular micro data centers, and Hivenet provides the necessary cloud software. Together, they form the first fully integrated neocloud for AI inference. Based in Cannes, France, Antimatter plans to deploy its computing units where renewable energy is already available, thus avoiding long electrical connection delays. The company aims to secure 300 million euros to achieve its goal of 1,000 micro data centers by 2030. Leading the charge is David Gurlé, a renowned French entrepreneur in the tech sector.

An Infrastructure Designed for AI Inference

Antimatter did not emerge from nowhere. It is the result of the merger of three companies with well-defined competencies:

• Datafactory: specialist in energy infrastructure in the United States.
• Policloud: creator of modular, compact, and rapidly deployable micro data centers.
• Hivenet: provider of the software layer that orchestrates everything.

The entire chain, from electricity production to AI query processing, is designed and controlled in-house. David Gurlé, CEO of Antimatter, summarizes the company's philosophy: “In the age of AI, intelligence is not the bottleneck; energy is.” In other words, the deployment of AI is hindered not by the lack of high-performing models but by the capacity to power them. Policloud units are therefore installed near existing renewable energy sources, converting often wasted energy into AI computing power in just a few months.

Antimatter's Global Ambitions

David Gurlé, at the helm of Antimatter, is far from a stranger in the tech world. This French entrepreneur, a knight of the Legion of Honor, notably founded the technology at Microsoft that would become Teams, used by hundreds of millions of people. He also led the enterprise division of Skype during its acquisition by Microsoft and launched Symphony Communication Services, valued at $1.4 billion. His ambitions for Antimatter are thus commensurate with his background.

To achieve its goals, Antimatter is seeking to secure 300 million euros to deploy its first 100 Policloud units by 2027. These micro data centers, designed as compact containers filled with electronics, can house up to 400 GPUs each, essential for running AI models. In addition to being operational in five months compared to twenty-four for a traditional data center, these units will represent 40,000 GPUs and 3.6 exaFLOPS, an immense computing power capable of processing billions of AI queries daily.

Antimatter aims to deploy 1,000 Policloud units in dozens of countries by 2030, with 400,000 active GPUs and computing power equivalent to that of five traditional hyperscale data centers, all for half the capital investment.

A Reduced Environmental Footprint

Antimatter highlights a significant argument in the current debate on the environmental impact of AI: its infrastructures would emit 70% less CO₂ than the industry average, and they do not consume water to cool their equipment, unlike traditional data centers that can consume millions of liters each year.

Antimatter has already put 3,344 GPUs into operation, with demand exceeding 10,000 units, indicating that the market is receptive. Its first clients come from various sectors: energy (35%), public (30%), agriculture (15%), and enterprises (20%), illustrating the versatility of its infrastructure.

Financially, the goals are ambitious: to exceed $250 million in revenue in the next 18 months, and reach $3 billion by the end of 2030.