# CoreWeave

Official pricing: https://www.coreweave.com/pricing  
Category: gpu-cloud · Isolation: container

## Pricing regimes (raw)

- **B200 HGX, on-demand (×8 node)** (resource): $0/vCPU-h, $0/GiB-h
- **H100 HGX, on-demand (×8 node)** (resource): $0/vCPU-h, $0/GiB-h
- **H200 HGX, on-demand (×8 node)** (resource): $0/vCPU-h, $0/GiB-h
- **L40S, on-demand (×8 node)** (resource): $0/vCPU-h, $0/GiB-h
- **L40, on-demand (×8 node)** (resource): $0/vCPU-h, $0/GiB-h
- **A100-80GB, on-demand (×8 node)** (resource): $0/vCPU-h, $0/GiB-h
- **RTX-PRO-6000 High Memory, on-demand (×8 node)** (resource): $0/vCPU-h, $0/GiB-h
- **GH200, on-demand (×1)** (resource): $0/vCPU-h, $0/GiB-h

## Features

Yes: snapshots, fork/clone, persistent disk, volumes, ≥24 h sessions, custom image (Docker or snapshot), start from your own snapshot, your own Docker/OCI image, Docker inside, preinstalled agents, SSH, public IPv4, HTTPS preview URLs, egress allowlist, static egress IP, private networking, SOC 2, SSO, GPU, arm64, Python, Node.js, live fork (no pause), EU data residency, MCP server, snapshots on demand, agent harness API, inbound access rules, extra volumes, shared volumes

No: memory snapshots, pause/resume, idle auto-stop, full VM (own kernel), nested virtualization, browser, desktop GUI, computer-use API, browser + desktop control, anti-bot stealth, CAPTCHA solving, residential IPs, self-hosting, BYOC, open source, HIPAA, Windows, macOS, HTTP method/path egress rules, wake on request, live resize, memory fork, automatic snapshots, hosted agent API (their own agent), gVisor or VM (no shared kernel), firewall inside (nftables), secret proxy, secret proxy for any API

Unknown: everything else. Evidence (source + quote) per feature: https://battleships.dev/data/providers/coreweave.json → feature_evidence

## Caveats

- Current node tables bundle CPU/RAM. Do not reuse obsolete per-resource A100 prices. Spot rates are regional point-in-time quotes. Multi-GPU node minimum makes divided rate unsuitable for a single-GPU sandbox.
- Any null count, CPU/RAM, billing increment or minimum is unverified, not unlimited/free.
- This is GPU-product research, not a claim that every feature of the entire vendor documentation was audited. Unverified features are null.
- GPU multi-count bundles, variant/region selection and fractional slices require a SKU-aware estimator; unsupported rows remain unpriced in strict cards.
- Modes with gpu=null are intentionally excluded, not free GPU modes. Never substitute 0 for a null rate. Flags alt indicate DIY/ML infrastructure, not the same thing as managed untrusted-code sandboxes.
- Unknown fee, storage or bandwidth means full monthly total may be unavailable. Rates exclude taxes.
- 2026-09-28: 8-GPU node regimes now priced at list (node price / 8 per GPU, gpu_min_count 8, plan_limits_exempt); alt removed. H100 is only sold as an 8x HGX node ($49.24/h). Access is via sales (plan flagged sales).
- Single-GPU prices on coreweave.com/pricing (B200 $8.60, H200 $6.31, H100 $6.16, A100 $2.70, RTX PRO 6000 $2.50, L40S $2.25, L40 $1.25 per GPU-h) apply exclusively to CoreWeave inference-platform customers via an account executive; general compute is sold as 8-GPU nodes.

## How this provider charges

As of 2026-09-28. GPU nodes on Kubernetes / virtual servers.

Current node tables bundle CPU/RAM. Do not reuse obsolete per-resource A100 prices. Spot rates are regional point-in-time quotes. Multi-GPU node minimum makes divided rate unsuitable for a single-GPU sandbox.

| Regime | When applicable | Billing | Numbers ($/GPU-hour unless slice) | Source |
|---|---|---|---|---|
| on-demand  | B200 HGX; count [8]; US table; region-dependent availability and spot prices. | unknown; min unknowns | 8.6; list; per_gpu_hour | [source](https://www.coreweave.com/pricing) |
| on-demand  | H100 HGX; count [8]; US table; region-dependent availability and spot prices. | unknown; min unknowns | 6.155; list; per_gpu_hour | [source](https://www.coreweave.com/pricing) |
| on-demand  | H200 HGX; count [8]; US table; region-dependent availability and spot prices. | unknown; min unknowns | 6.305; list; per_gpu_hour | [source](https://www.coreweave.com/pricing) |
| on-demand  | L40S ; count [8]; US table; region-dependent availability and spot prices. | unknown; min unknowns | 2.25; list; per_gpu_hour | [source](https://www.coreweave.com/pricing) |
| on-demand  | L40 ; count [8]; US table; region-dependent availability and spot prices. | unknown; min unknowns | 1.25; list; per_gpu_hour | [source](https://www.coreweave.com/pricing) |
| on-demand  | A100-80GB ; count [8]; US table; region-dependent availability and spot prices. | unknown; min unknowns | 2.7; list; per_gpu_hour | [source](https://www.coreweave.com/pricing) |
| on-demand  | RTX-PRO-6000 High Memory; count [8]; US table; region-dependent availability and spot prices. | unknown; min unknowns | 2.5; list; per_gpu_hour | [source](https://www.coreweave.com/pricing) |
| on-demand  | GH200 ; count [1]; US table; region-dependent availability and spot prices. | unknown; min unknowns | 6.5; list; per_gpu_hour | [source](https://www.coreweave.com/pricing) |
| spot  | B300 ; count [8]; US table snapshot; another combined/default table shows H100 $19.71 and H200 $20.93 per node. Not a guaranteed quote. | unknown; min unknowns | 4.5875; dynamic-snapshot; per_gpu_hour | [source](https://www.coreweave.com/pricing) |
| spot  | B200 ; count [8]; US table snapshot; another combined/default table shows H100 $19.71 and H200 $20.93 per node. Not a guaranteed quote. | unknown; min unknowns | 4.35875; dynamic-snapshot; per_gpu_hour | [source](https://www.coreweave.com/pricing) |
| spot  | H100 ; count [8]; US table snapshot; another combined/default table shows H100 $19.71 and H200 $20.93 per node. Not a guaranteed quote. | unknown; min unknowns | 2.43875; dynamic-snapshot; per_gpu_hour | [source](https://www.coreweave.com/pricing) |
| spot  | H200 ; count [8]; US table snapshot; another combined/default table shows H100 $19.71 and H200 $20.93 per node. Not a guaranteed quote. | unknown; min unknowns | 2.58; dynamic-snapshot; per_gpu_hour | [source](https://www.coreweave.com/pricing) |
| spot  | RTX-PRO-6000 ; count [8]; US table snapshot; another combined/default table shows H100 $19.71 and H200 $20.93 per node. Not a guaranteed quote. | unknown; min unknowns | 1.37625; dynamic-snapshot; per_gpu_hour | [source](https://www.coreweave.com/pricing) |

## Gotchas
- Current node tables bundle CPU/RAM. Do not reuse obsolete per-resource A100 prices. Spot rates are regional point-in-time quotes. Multi-GPU node minimum makes divided rate unsuitable for a single-GPU sandbox.
- Any null count, CPU/RAM, billing increment or minimum is unverified, not unlimited/free.
- This is GPU-product research, not a claim that every feature of the entire vendor documentation was audited. Unverified features are null.
- GPU multi-count bundles, variant/region selection and fractional slices require a SKU-aware estimator; unsupported rows remain unpriced in strict cards.

## Worked example (required protocol workload)
4 vCPU / 8 GiB, 50 concurrent × 8 hours/day × 22 days = **8,800 instance-hours/month**. 30% CPU utilization does not reduce allocated GPU uptime. 50 GiB snapshots and 100 GiB egress are additional. The requested CPU-only workload has no GPU type/count, so its full GPU-provider total is **null**, not a fictitious CPU equivalent.

Explicit GPU sensitivity example: add **1 × GH200** to each of those 50 workers at the GH200 quoted shape. Compute component = 8,800 × $6.5 = **$57,200.0000**. CPU/RAM are included in that SKU (not independently resizable).
This is compute-only, before platform fees, storage, egress, setup, minimum rounding, cold-start/idle tails and capacity quotas. 50 concurrent GPUs are not promised by a unit rate.

## Evidence scope
Sources read are listed in the provider/features JSON. Raw pricing pages, relevant docs and API payloads are retained under `../raw/`. All unknown feature toggles remain null.