Equinix held its first Horizon customer event on September 2 in San Francisco, launching two products - Fabric One and Inference Exchange - alongside a keynote pairing CEO Adaire Fox-Martin with NVIDIA CEO Jensen Huang. The announcements mark a strategic push to position Equinix's interconnection fabric as the neutral meeting point for enterprise AI infrastructure, and the market has already responded: shares have climbed roughly 40% over the past year to exceed $1,000.
Fox-Martin framed the strategy through a historical lens, tracing trade exchanges from antiquity to position Equinix as the AI-era equivalent. Her argument: compute fragmentation, shifting network traffic, and rising AI token costs are reshaping enterprise infrastructure in ways that favor neutral interconnection platforms over hyperscaler lock-in.
Two products, one thesis
Fabric One is an intent-based connectivity service that configures itself based on desired outcomes rather than manual network specifications. It addresses a core pain point: enterprises spreading workloads across 20 or more clouds face geometric growth in point-to-point connection costs. Equinix's connect-once model aims to simplify that.
Inference Exchange, built with NVIDIA and Together AI, deploys distributed AI inference infrastructure within Equinix data centers. Both products are scheduled for general availability in Q1 2027, with a Fabric One pilot expected in Q4 2026.
The NVIDIA partnership is notable for what it excludes. Unlike CoreWeave and Nebius, which hold direct capital relationships with NVIDIA, Equinix's arrangement is architectural rather than financial - no exclusive GPU access, no privileged silicon. Together AI remains free to partner elsewhere. That reinforces Equinix's neutrality but also means the exchange's value will depend on performance and pricing, not exclusivity.
The physical infrastructure advantage
Equinix also announced expansions in liquid cooling, a power-first site strategy, and a broader Managed Solutions portfolio with NVIDIA and Dell. On sustainability, the company cited 1,400 MW of contracted wind and solar, on-site fuel cells, and over 1 GW of next-generation nuclear capacity in the pipeline.
The PG&E relationship offers a concrete data point: each gigawatt of steady data center demand reduces neighboring ratepayers' tariffs by 1%, according to the utility. That counters the narrative that data centers burden local communities - a real constraint in markets like Northern Virginia and Dublin where development has faced opposition.
The Equinix Foundation commitment, now doubled to $100 million, signals the company recognizes social license to operate as a growth constraint, not just a PR concern.
What to watch
The strategic logic is sound, but execution questions remain. Fabric One's five-month development cycle was aggressive; whether intent-based provisioning can handle complex hybrid environments at scale is unproven. The real test will be Q4 pilot performance under production network loads.
Also worth monitoring: whether NVIDIA eventually extends Equinix a direct capital relationship, and whether the neocloud ecosystem consolidates around neutral interconnection points or retreats into cloud-provider relationships. The more workloads run on independent GPU clouds, the more valuable Equinix's position becomes.
Why this matters for executives and strategy leaders
The Inference Exchange and Fabric One announcements are bets that enterprises want optionality in AI infrastructure - the ability to plug into multiple clouds, model providers, and GPU sources without committing to a single vendor stack. For strategy leaders evaluating AI deployment, the question is whether your organization actually needs that optionality today or whether it's insurance against vendor lock-in you haven't yet experienced. Equinix's 40% share gain suggests investors believe the demand is real. The next two quarters of customer deployments will show whether enterprises agree.
Your membership also unlocks:
