Edgevana Signs More Than $700 Million in Agreements to Deploy Blackwell Systems
Key Facts
Edgevana announced on November 9, 2026, that it had signed multiyear agreements worth more than $700 million to deploy NVIDIA B300 units and GB300 systems in North America and Europe. The company described the agreements as fully committed and structured on a “pay the agreed minimum whether the service is used or not” basis. The deals include infrastructure deployment coordination, as customer demand for AI computing accelerates, according to its statement published via PR Newswire. Edgevana said contracted customer demand had reached $7 billion at the time of the announcement, while its active pipeline had exceeded $20 billion. These three metrics represent different commercial figures: the value of the new agreements, demand associated with customer contracts, and opportunities that remain in the pipeline.
The new agreements are based on Blackwell systems, including B300 units and GB300 systems, to be deployed at sites in two key regions: North America and Europe. The company says its model combines power-ready land, data center facilities and high-speed networks in a single virtual campus. The arrangement is intended to make tens of megawatts of high-density AI computing available within 90 to 120 days, according to Edgevana’s estimate. The company is betting on accelerating access to computing capacity through ready-to-use sites, rather than waiting for a traditional data center to be built from the ground up. The statement did not identify the locations for the new agreements or the start-up dates for each site.
Edgevana attributed demand growth to power constraints delaying traditional data center construction. This obstacle intersects with a market where available space is shrinking despite expanding supply, according to CBRE’s first-quarter 2026 report. Supply across the 16 largest data center markets reached 16 gigawatts, up 25% from the same period a year earlier. But the average vacancy rate fell to 6.7% from 8.3% a year earlier, the report said. CBRE said demand is outpacing even large increases in supply, prompting companies to secure space early and favor markets with reliable power supplies.
Edgevana aims to capitalize on these conditions through a network it said includes 17 operational data center sites, with 10 more under development. It also said platforms running on its infrastructure serve more than 10,000 later-stage enterprise customers. These figures describe the reach of the network and the customers served by the platforms, while the announced agreements remain tied to deployments of specific systems in North America and Europe. Co-CEO Ryan Fay linked the company’s strategy to changes in the nature of AI workloads, saying: “Now that AI is starting to shift to inference, the market came to us.” He added that Edgevana’s expertise was built on providing ultra-low latency for high-frequency trading, which he said aligns with the market’s shift toward inference.
The focus on inference takes on significance in light of the specifications NVIDIA lists for its GB300 systems. The company describes DGX GB300 as a rack-scale, liquid-cooled system designed for training, post-training and inference workloads. NVIDIA says the system’s dense FP4 performance is 1.5 times that of DGX GB200, while its attention performance is twice as high. These specifications position the system as a tool for running AI workloads, including inference, which Edgevana said the market is moving toward. Deployment of these systems remains tied to the availability of the sites, power and networks required, elements Edgevana said it brings together in its operating model.
The power dimension shows the scale of the challenge facing plans to expand data centers in the United States. In September 2026, a Reuters report said grid interconnection requests from major power consumers, mostly for data centers in parts of the Midwest, Mid-Atlantic and South, had exceeded 700 gigawatts. Regulators warned that some requests could be duplicates or infeasible, according to the report. Earlier, in June, the U.S. Federal Energy Regulatory Commission ordered grid operators to consider measures to speed up interconnections for major power consumers while limiting the impact on grid stability and electricity bills. The commission gave operators 60 days to respond, according to a Reuters report.