Customer Segments
Different neoclouds win different customer segments. Mapping segment to vendor clarifies who competes for what.
Frontier labs
OpenAI, Anthropic, Meta AI, Google DeepMind, etc.
- Primary providers: CoreWeave (Microsoft / Meta capacity), Crusoe (Stargate), Nebius (Microsoft).
- Why: Multi-thousand-GPU clusters; multi-billion-dollar commitments; deepest operational integration.
- Off-limits: Marketplaces, smaller dedicated.
Mid-market AI
Series A through Series D AI startups and AI-focused mid-enterprises.
- Primary providers: Lambda, RunPod Secure, sometimes Nebius for European customers.
- Why: Smaller commitments; need reliability without negotiating CoreWeave-scale deals.
- Marketplace use: Often for research / experimentation alongside production on dedicated.
Indie developers
Hobbyists, students, indie ML builders, ML consultancies.
- Primary providers: Vast.AI, RunPod Community, TensorDock, Hyperbolic.
- Why: Price-sensitive; high tolerance for variability; need broad hardware selection.
- Off-limits: Enterprise neoclouds (don't sell at indie scale).
Enterprise IT
Non-AI-native enterprises adopting AI as one capability among many.
- Primary providers: Hyperscalers, with CoreWeave / Nebius / Lambda as secondary options.
- Why: Bundled with broader cloud relationships; enterprise procurement comfort.
- Smaller neoclouds: Limited penetration here.
Inference-as-a-service
Companies wanting OpenAI-style per-token APIs.
- Primary providers: Together.AI, Fireworks (not in this guide), Hyperbolic (inference side), RunPod Serverless.
- Adjacent: Closed-source (OpenAI, Anthropic) for highest-quality use cases.
Takeaway
The segmentation is clean. Each major segment maps to a small set of providers; no single neocloud wins all segments. The market is heterogeneous by design. The next chapter examines the strategic positioning of each company.