Antimatter has launched as a vertically integrated “neocloud” focused on AI inference, formed through the combination of three companies: Datafactory (energy and power infrastructure in the US), Policloud (a modular micro data center network), and Hivenet (a distributed cloud provider). The company says it has secured more than 1 GW of power capacity via grid connection agreements and reserved sites across distributed “micro-power” locations in the US, Europe, and the GCC, and plans to deploy 1,000 distributed micro data centers to serve inference workloads.
Antimatter’s buildout centers on Policloud modular, containerized micro data centers. Each unit is described as housing up to 400 GPUs and being deployable in as little as five months, compared with “24+ months” for traditional hyperscale builds. Antimatter says it currently operates 10 units across eight sites, and has a commercial pipeline of more than 500 additional units.
On power siting, the company describes an “energy-first” approach that places Policloud units “at or near existing power assets,” including wind, solar, hydro, or biogas sites. Antimatter says more than 160 MW is already operational across Texas and Oregon, and that deploying at existing generation can convert “stranded generation” into AI infrastructure “in a matter of months.”
For the software layer, Antimatter describes a proprietary distributed computing and storage platform intended to orchestrate distributed hardware into a single “sovereign cloud fabric,” with “global default Tier 3 capability.” The company says the platform is designed to support billions of inference requests per day, with sub-10 ms latency for edge workloads and “full data sovereignty” for regulated industries.
Antimatter says it is “securing €300 million” to fund deployment of its first 100 Policloud units by 2027, which it equates to 40,000 GPUs and 3.6 exaFLOPS of active compute capacity (using an RTX 5090 assumption of ~90 TFLOPS FP32). By the end of 2030, Antimatter’s stated plan is a 1,000-Policloud network totaling more than 400,000 GPUs and over 36 exaFLOPS, which the company calls “the equivalent of five traditional hyperscale data centers,” deployed across “dozens of countries.”
Some of the company’s comparisons are aggressive. Antimatter lists capex per fully loaded MW at “~$7M,” versus “~$35M” for “traditional hyperscale,” along with “~50% below hyperscalers” customer pricing, “~70% lower” carbon reduction, and “zero water cooling.” Those are big swings for any operator to bank on without third-party validation, but the direction of travel—siting compute at available power and shipping capacity in modular blocks—fits the real constraint many teams are living with: interconnect and grid timelines can dominate the critical path.
Antimatter is led by David Gurlé, Cofounder, Executive Chairman, and CEO. “In the age of AI, intelligence is not the bottleneck—energy is,” Gurlé said. “The inference era requires a different model: more distributed, faster to deploy, and sovereign by design.”
On early commercialization metrics, Antimatter reports $20 million in forward-looking revenue, 3,344 GPUs deployed with demand for “10,000+,” and a customer mix it breaks out as Energy (35%), Public sector (30%), Agriculture (15%), and Corporates (20%).
Source: Antimatter
