Meridian is a stealth sub-150MW neocloud building an asset-light wedge into AI infrastructure: originate 3–5 year GPU-as-a-service contracts, collect ~20% of TCV upfront, finance the balance with debt against contracted cash flows, and later layer sovereign-cloud software on top of an installed base. The founding team is unusually credentialed for the category — Vic Rose was a primary lead on AWS Local Zones/Wavelength and most recently Director of Inferencing at Microsoft (compressing GPU region deployment timelines for OpenAI-related workloads); co-founder Anthony scaled Overwatch Mission Critical from ~3 to ~450 people and raised a ~$20M Series A. Jim Trout (founder of CoreSite and Vantage) is described as a prospective board member/lead.
The thesis is that operator credibility, partnership-native execution, and contract-first financing can compress the path to a mid-scale GPU business faster than vertically integrated neoclouds — but venture-scale returns require Meridian to escape a project-finance profile by owning a durable software/control layer, which today is aspirational. The internal read (Miles, 2026-04-06) explicitly flags that value in an OpCo/PropCo split accrues to the assets, not the equity we would own, and that likely exits (infra funds, PE, strategics) price on yield rather than growth multiples. Continued conversation is warranted on founder quality alone; conviction depends on signed customer contracts, a financeable first deployment, and a sharper platform thesis.
The neocloud tailwind is genuinely enormous and durable: analyst forecasts cluster in the 40–60% CAGR range through the end of the decade, driven by inference demand, sovereign compute mandates, and continuing hyperscaler supply constraints. Meridian is targeting the sub-150MW, sub-100MW-per-site slice that hyperscalers under-serve. The company’s own $920M i3D opportunity, if it converts, would validate demand density in the mid-market segment they are chasing; even a fraction of the 4 in-flight LOIs converting into signed contracts would seed a real financing base. The strategic question is whether Meridian captures durable multiples on that TAM or whether the value accrues to whichever entity owns the depreciating hardware.
| Company | Positioning | Scale / Signal Web |
|---|---|---|
| Meridian | Asset-light sub-150MW neocloud; contract-first GPUaaS with sovereign-cloud software vision | Pre-revenue; 4 LOIs; live 1.05 PUE lab in West Sacramento |
| CoreWeave | Full-stack GPU cloud; hyperscaler-alternative | $5.1B FY25 revenue, $12–13B FY26 guide, $99.4B backlog, public |
| Nebius | Inference-as-a-service via Token Factory; managed endpoints | Public; pivoted to inference SLAs Nov 2025 |
| Crusoe | Renewable-powered H100/H200 clusters | Valued >$10B as of Oct 2025 |
| Lambda Labs | Vertically integrated; owns 100MW KC AI Factory (10K+ Blackwell Ultra) | Raised >$1.5B late 2025; IPO track |
| Nscale | European/sovereign neocloud | $14.6B valuation; $2B Series C March 2026 |
| IREN | Full-stack via Mirantis + Nostrum acquisitions | $625M Mirantis buy in 2026 to add orchestration |
Moat sentence: The defensible edge is not proprietary hardware or a novel software layer — it is speed and partnership choreography: Vic’s ability to assemble GPUs (Positron, AMD), colo/build capacity (Overwatch), cooling (Anthonium), orchestration (Ray.io), and financing (Tower Digital) against a signed customer contract faster than a vertically integrated competitor could plan a build. Whether that choreography compounds into a moat — rather than degrading into a commodity broker margin as capacity floods in — is the central uncertainty.
Named comparable acquirers cluster in three buckets: strategics buying GPU capacity + orchestration (Meta’s April 2026 $21B CoreWeave commitment and February 2026 SpaceX/xAI absorption of terrestrial compute set the tone Web); full-stack neoclouds rolling up software layers (IREN’s $625M acquisition of Mirantis and simultaneous Nostrum Group deal Web); and infrastructure funds/PE valuing on contracted yield rather than growth multiples. Meridian’s OpCo, absent a durable software layer, defaults to the third bucket — a good business, but not the 30–50× outcome Fund II underwrites. Public comps (CoreWeave >$50B market cap, Nebius, Crusoe >$10B, Nscale $14.6B) confirm the strategic bid exists but at scale, not for a sub-150MW operator without proprietary IP.