Equinix's soft Q3 guide signals interconnection demand is still growing—but value may be leaking back to hyperscalers

Equinix issued its updated outlook with a clear top-line framework for Q3 2026, but it did not provide a Q3-specific interconnection revenue range alongside total company guidance. In practice, that means investors get a read on platform momentum, yet they don’t get a hard confirmation that the incremental AI buildout is flowing to interconnection at the same rate (or with the same pricing power) as the overall ecosystem.

Equinix’s model depends on third parties needing low-latency connectivity across clouds, SaaS, and enterprise systems. But AI buildouts don’t only raise compute demand—they also increase the share of infrastructure spend that can be optimized internally by hyperscalers (especially when they expand their own GPU and AI-ready campus footprints and then connect workloads with private networking). The investment-relevant question is not whether interconnection demand exists—it’s whether the marginal AI dollar is “routed through” the interconnection REIT or is absorbed upstream by captive builds.

In the most recent quarter disclosures we opened, Equinix highlighted record interconnection net additions and provided interconnection revenue levels. Those facts matter for the baseline thesis (platform demand is intact). However, they do not answer the counterfactual investors care about: whether incremental AI-driven connectivity spend is increasingly being satisfied through hyperscaler-owned or hyperscaler-directed paths that reduce third-party interconnection revenue per incremental AI capacity.

A practical supply-chain chain looks like this: (1) AI GPU and accelerator capacity is built in data centers, (2) the workloads need private, predictable connectivity to clouds, data stores, and enterprise applications, and (3) operators monetize either through wholesale network services (interconnection ecosystems) or through integrated captive networking within hyperscaler campuses and partner programs.

If hyperscalers increase captive networking share, interconnection operators can still benefit from the existence of more AI-ready capacity they colocate on—but they may capture less of the incremental networking “integration” layer per added AI workload.

This isn’t a claim that Equinix is losing interconnection demand. It’s a narrower claim: the market may be paying for an AI buildout “tax” that interconnection REITs might not fully collect if hyperscalers keep integrating more of the fabric and orchestration inside their own spend.