Equinix bets on inference era with Nvidia-backed AI cloud service

Equinix, which describes itself as a colocation provider, rents data center space, power, cooling and security to more than 10,500 customers. Its 281 facilities span 77 metropolitan areas across six continents. The company says its urban locations give it an edge as AI shifts from training to inference, which requires high-speed communication between servers and end users. According to McKinsey, inference will make up half of all AI compute and 30% to 40% of total data center demand by 2030.

The new offering, called Equinix Inference Exchange, is scheduled to be available in the first quarter of 2027. It will allow customers to connect to different clouds and providers, run inference on open-source models, and manage costs. Together AI, which offers access to 200 open-source models, will be the seller of record and bill end customers, according to Maryam Zand, Equinix's vice president for AI ecosystem strategy. Equinix data centers are optimized for Nvidia's B300 Blackwell Ultra GPUs, and some liquid-cooled facilities can support newer Vera Rubin chips, she said.

Equinix has largely maintained its strategy of building smaller data centers near city centers and positioning them as network interconnection hubs. Its xScale business serving hyperscalers mostly operates facilities under 100 megawatts, while some AI data centers are measured in gigawatts. Vlad Galabov, a data center analyst and host of the AIDC Debate podcast, said Equinix was "too slow" to react to gigawatt-scale demand, and that colocation companies now see the need to plan more strategically.

Equinix's stock is up 33% this year, lifting its market cap to $100 billion, by far the largest among data center REITs and ahead of Digital Realty's $68 billion. The company also announced Equinix Fabric One, a connectivity service meant to simplify networks running on multiple clouds and AI models.