Equinix Launches 2027 AI Inference Platform With Nvidia as $5 Trillion Data Center Race Shifts
Updated
Updated · CNBC · Sep 2
Equinix Launches 2027 AI Inference Platform With Nvidia as $5 Trillion Data Center Race Shifts
3 articles · Updated · CNBC · Sep 2
Summary
Equinix said its new Equinix Inference Exchange, built with Nvidia and delivered through Together AI, will launch in the first quarter of 2027 to let customers run open-source AI models across multiple clouds.
Together AI will bill end customers for access to roughly 200 open-source models, while Equinix positions the service as a way to cut inference costs and link users to different cloud and compute providers.
281 Equinix facilities across 77 metro areas give the company an urban footprint it says is better suited to latency-sensitive inference, even as analysts say it reacted too slowly to gigawatt-scale AI training demand.
McKinsey estimates inference will account for half of AI compute and 30% to 40% of total data center demand by 2030, a shift Equinix is targeting as hyperscalers are projected to spend more than $5 trillion on infrastructure.
As AI inference spending surges in 2026, can Equinix's upcoming decentralized network actually defeat hyperscalers in the race for zero-latency compute?
With the 2027 launch approaching, how will Equinix solve the massive urban power demands required for its new liquid-cooled AI inference hubs?