Vapor IO

Vapor IO spent a decade building what everyone now says AI needs: physical compute at the true network edge. Its Kinetic Grid of automated micro data centers at fiber intersections and wireless aggregation hubs — funded by a ~$90M Berkshire Partners-led Series C with Crown Castle participation, over $100M total — is infrastructure that would take any challenger years of permitting, fiber negotiation, and construction to replicate. The strategic moment has finally arrived at the asset: Zero Gap AI delivers network-adjacent GPU inference across 36 U.S. markets with Supermicro and NVIDIA, the company deployed the world's first private 5G AI-RAN with NVIDIA Aerial for the City of Las Vegas (supporting thousands of municipal cameras), and the February 2025 Veea partnership packages turnkey AI-as-a-Service for enterprises and municipalities that will never build their own edge. As inference shifts from chat to physical-world AI — video analytics, robotics, autonomous systems, agentic workloads with real latency budgets — round-trips to hyperscale regions become the binding constraint, and the operator already standing at the tower base captures the workload. Founder Cole Crawford's Open Compute pedigree and the Crown Castle relationship position Vapor IO as either the category winner or the obvious strategic acquisition for any carrier, tower REIT, or CDN that needs an edge-AI story overnight.

Vapor IO's last disclosed equity round was 2018 — eight years without a fresh mark is itself the market's verdict, and an 11-50 headcount against "nationwide edge infrastructure" ambitions implies a company sustaining itself rather than scaling. The deeper problem is that the killer low-latency application has been perpetually eighteen months away since 2014: 5G monetization disappointed, AR never arrived, autonomous vehicles do compute onboard, and even AI inference — the latest hope — is mostly batch-tolerant and cache-friendly, adequately served by regional zones from Cloudflare, hyperscaler Local Zones, and metro colocation, all of which undercut the micro-data-center thesis without new construction. The graveyard is instructive: EdgeMicro died, American Tower quietly retreated from edge compute, and Crown Castle — Vapor IO's own strategic backer — has been shedding fiber ambitions. The business model compounds the demand problem with capital intensity: micro facilities carry fixed costs against uncertain utilization, and GPU-equipped edge sites add depreciation risk on top. The recent posture — partnerships with Veea, Supermicro press releases, municipal showcases — reads like a company marketing capacity it cannot yet fill. Without a major capital event or an anchor tenant at scale, the plausible path is an asset sale to a carrier or REIT at infrastructure, not platform, multiples.