Layer 2.5 · head to head
RunPod vs Vast.ai
Curated marketplace against open bidding: the price gap is the reliability gap.
| RunPod | Vast.ai | |
|---|---|---|
| Can you leave | Movable with effort. Expect to redo images, storage wiring and networking. | Movable with effort. Expect to redo images, storage wiring and networking. |
| Counterparty | Low exposure for you: you are renting by the second, so the switching cost of a supplier failing is hours, not quarters. That is the honest advantage of the marketplace model. | You are renting from individuals and small operators. Assume any single machine disappears without notice and design for it. |
| What it is | Brokers capacity somebody else owns. Cheapest headline rates, and the machine you get is not the machine you chose. | Brokers capacity somebody else owns. Cheapest headline rates, and the machine you get is not the machine you chose. |
| How you reach it | You push a container image and it runs. No cluster to operate. | You push a container image and it runs. No cluster to operate. |
| Accelerators | H100, H200, A100, L40S, RTX 4090, RTX 5090, and a long consumer tail | Whatever the market is offering — RTX 5090, 4090, PRO 6000, V100, H100 and a long tail |
| Regions | Global, community and secure clouds | Global, wherever a host has hardware |
| Pricing model | Per-second billing, on-demand and spot. Two tiers: 'Secure Cloud' in real datacentres, 'Community Cloud' on other people's hardware. | Open bidding marketplace, interruptible and on-demand. The clearing price is public without an account. |
| Getting started | A few dollars. | Cents. |
| Capacity | Generally good on consumer parts, variable on datacentre parts. | Deep on consumer silicon, thin and volatile on datacentre parts. |
| Ownership | Private. | Private. |
RunPod
For: Fine-tuning, batch jobs, inference experiments, anyone whose workload can checkpoint and move.
The catch: Community Cloud is somebody's machine somewhere. For anything with a compliance story attached, that distinction is the whole decision, and it is easy to miss in the pricing table.
Economics: Reported around $2/hr for an H100 on-demand — roughly a third of CoreWeave list. Per-second billing genuinely matters for bursty work.
Vast.ai
For: Price-sensitive batch work, research, and anyone benchmarking what compute *should* cost before signing a contract.
The catch: Reliability is the price of the price. Hosts vary enormously, verification tiers matter, and a cheap unverified box that dies at hour nine is not cheap.
Economics: Consistently the market floor — reported around $1.49/hr for an H100 where hyperscalers ask $7. Its public API is the closest thing this market has to a spot index, which is why this site measures it.
Neither table row is a price
Deliberately. Published on-demand rates at this layer move weekly, and essentially nobody signing a real contract pays them — every serious buyer pays less than every list figure either of these companies publishes. Quoting one here would date this page within a month.
The GPU rental price index carries dated, sourced figures instead, and the durable finding there is the spread: the identical H100 rents from roughly $1.38 to $12.29 an hour depending only on who you rent it from.
The layers underneath both
Whichever you pick is renting you chips in a building that needs power. In 2026 that is the constraint that binds: Microsoft has disclosed an Azure backlog it cannot fill for want of megawatts rather than accelerators, and the US interconnection queue exceeds 2,600 GW with roughly 80% of projects withdrawing before they energise.
Layer 2 — Silicon · Layer 1 — Energy · The interconnection queue