Vast.AI
The original GPU marketplace. Founded 2016 on the thesis that the world is full of underutilized GPU hardware and the people who need it most can't afford hyperscaler rates. Twelve chapters covering founding through outlook.
Section A · Orient
00Start HereHow to read this guide, why Vast.AI matters in the landscape.
01The CompanyFounding, mission, scale, ownership, organizational shape.
Section B · The marketplace
02The Marketplace ModelTwo-sided dynamics; how Vast differs from a traditional GPU cloud.
03The Supply SideWho provides GPUs — gaming rigs, mining farms, small DCs, labs. Onboarding, ranking, payouts.
04The Demand SideWho rents GPUs — researchers, students, indie ML builders, occasional enterprises. Use cases.
05Pricing & BiddingBid mechanic, on-demand vs interruptible, why Vast can be 60–80% cheaper than hyperscalers.
Section C · Economics & operations
06Hardware & TiersThe range of available GPUs (RTX consumer, A100, H100, B200). Tier distribution, reliability.
07Trust, Verification & NetworkingDLPerf benchmarks, host scoring, ssh access, bandwidth limits, why trust is the hard problem.
08Unit EconomicsVast's take rate, provider economics, why margins are thin but scale is interesting.
09Customer SegmentsWho actually uses Vast — and who doesn't. Where the boundary against enterprise neoclouds sits.
Section D · Position & outlook
10Competitive PositioningVast vs RunPod, vs hyperscalers, vs CoreWeave/Crusoe. Strengths, weaknesses, defensibility.
11Recent DevelopmentsWhat's changed at Vast in 2025–2026. Product moves, partnerships, capacity expansion.
12Outlook & Investment ThesisWhere Vast goes from here — bull case, bear case, and the wildcards (compute futures, GPU price normalization).