Equinix has put a quarter on its next enterprise AI infrastructure offering, but not yet a price or a launch map. The company announced Sept. 2 that Equinix Inference Exchange, developed with NVIDIA and Together AI, is expected to become available starting in the first quarter of 2027.
The planned offering could give enterprises another way to run AI inference closer to users, applications, and regulated data. But Equinix has not disclosed initial deployment locations, pricing, detailed hardware configurations, capacity information, or specific performance commitments, leaving buyers without enough detail for serious procurement planning.
What Equinix has confirmed
The Sept. 2 Equinix announcement says Inference Exchange will combine NVIDIA Enterprise Reference Architectures with Together AI’s inference platform and Equinix’s data center infrastructure. Together AI will support more than 200 open-source models through multitenant and dedicated single-tenant deployments.
Equinix will supply power, cooling, infrastructure operations, and connectivity through Equinix Fabric. The design reflects a broader move toward integrated AI infrastructure, where compute, networking, cooling, and security increasingly have to be planned as one deployment rather than separate purchases.
Equinix is positioning Inference Exchange for metro-edge inference, open-model migration, and sovereign AI workloads. Those use cases depend heavily on geography, particularly for organizations that need to keep data or processing within specific jurisdictions.
An Equinix product release note says deployment locations and customer-access details will be disclosed later. The company has also not published a pricing model, making direct cost comparisons with hyperscaler APIs, GPU clouds, or internally managed infrastructure premature.
Those economics are becoming harder to ignore as AI usage scales. Recent data showing AI agents consuming about five times more LLM tokens than human-driven activity illustrates why inference capacity and consumption costs can quickly become infrastructure-planning issues.
What eWeek found: Equinix’s global footprint is not its Inference Exchange rollout
Equinix’s infrastructure footprint is much broader than the rollout information disclosed for Inference Exchange. Its current metro-edge AI page lists 281 AI data centers across 77 metros in 36 countries, while the new offering’s initial deployment locations remain unspecified.
The gap is clearer against Equinix’s earlier AI strategy. In its March 11 Distributed AI Hub announcement, Equinix said enterprises could connect privately to model companies, GPU clouds, data platforms, and other AI providers through its 280 high-performance data centers. The September Inference Exchange materials do not make an equivalent site-level deployment commitment.
Equinix is also adding capacity in markets such as Singapore, where it recently received a provisional 50MW allocation as part of a broader 200MW data center expansion. That additional capacity should not be assumed to support Inference Exchange until Equinix identifies the offering’s launch metros.
Enterprises can assess the proposed architecture and technology partners, but not yet its geographic fit or economics. Procurement teams still need deployment locations, pricing, capacity, and service commitments before they can compare Inference Exchange with existing production-inference options.
Read more: As inference hardware choices continue to evolve, NVIDIA’s reported talks with inference-chip specialist Rebellions show why infrastructure buyers may want to avoid locking in long-term assumptions before more deployment and capacity details emerge.
Want to learn more AI tips, tricks, and prompting techniques? eWeek readers get free 7-day access to The Neuron Academy, our practical learning platform designed to help professionals use AI more confidently at work. Browse all lessons →