# Nebius

Official pricing: https://docs.nebius.com/compute/resources/pricing  
Category: gpu-cloud · Isolation: vm

## Pricing regimes (raw)

- **B300, on-demand** (resource): $0/vCPU-h, $0/GiB-h [alt]
- **B300, spot** (resource): $0/vCPU-h, $0/GiB-h [alt, spot]
- **B200, on-demand** (resource): $0/vCPU-h, $0/GiB-h
- **B200, spot** (resource): $0/vCPU-h, $0/GiB-h [alt, spot]
- **H200, on-demand** (resource): $0/vCPU-h, $0/GiB-h
- **H200, spot** (resource): $0/vCPU-h, $0/GiB-h [alt, spot]
- **H100, on-demand** (resource): $0/vCPU-h, $0/GiB-h
- **H100, spot** (resource): $0/vCPU-h, $0/GiB-h [alt, spot]

## Features

Yes: snapshots, fork/clone, pause/resume, persistent disk, volumes, ≥24 h sessions, custom image (Docker or snapshot), start from your own snapshot, your own Docker/OCI image, full VM (own kernel), Docker inside, root, SSH, public IPv4, HTTPS preview URLs, egress allowlist, open internet, static egress IP, private networking, SOC 2, HIPAA, SSO, GPU, Python, live fork (no pause), EU data residency, snapshots on demand, gVisor or VM (no shared kernel), inbound access rules, firewall inside (nftables), extra volumes, shared volumes

No: memory snapshots, idle auto-stop, browser, desktop GUI, computer-use API, browser + desktop control, anti-bot stealth, CAPTCHA solving, residential IPs, preinstalled agents, self-hosting, BYOC, open source, arm64, Windows, macOS, Node.js, HTTP method/path egress rules, wake on request, live resize, memory fork, MCP server, automatic snapshots, agent harness API, hosted agent API (their own agent), secret proxy, secret proxy for any API

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

## Caveats

- Use rates BEFORE October 1, 2026. Pricing landing spot floors differ materially from documented preemptible rates: both retained but not conflated. L40S separately bills CPU/RAM; other GPU platform reference bundles include them. Dynamic spot starts October 8, not September 22 announcement date. As of Sep 28 docs fixed preemptible prices still apply.
- 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.
- B300/B200/H200/H100 on-demand prices rise 2026-10-01 to $9.50/$8.50/$5.40/$4.50 per GPU-h (from $7.85/$7.15/$4.50/$3.85); RTX PRO 6000 and L40S unchanged. Spot prices are dynamic (from $0.79) and left unpriced.
- CPU-only VMs (cpu-d3 AMD Genoa all regions, cpu-e2 Intel Ice Lake eu-north1) added 2026-09-28 at pre-October rates $0.012/vCPU-h + $0.0032/GiB-h; from 2026-10-01 AMD becomes $0.015 + $0.0045 and Intel $0.012 + $0.0045 (nebius.com/prices landing: "from $0.10" -> "from $0.13" AMD, "$0.05" -> "$0.06" Intel). Plain VMs: no sandbox API, disk ($0.071/GiB-month network SSD) billed separately.
- Serverless AI (Devlabs, jobs, endpoints: container workloads on Compute VMs) has no own pricing; it bills Compute VM + disk rates and counts toward Compute quotas. A stopped Devlab still pays its whole container disk.
- Preemptible: fixed prices ($2.15 H100, $2.45 H200, $3.95 B200, $4.30 B300, $0.95 RTX PRO 6000) apply until 2026-10-07; from 2026-10-08 a dynamic spot price updated every 15 min, floor 'from $0.79' (H100/H200/RTX) / 'from $0.99' (B200/B300), ceiling one cent below on-demand. L40S preemptible stays flat ($0.65 GPU + CPU/RAM). Spot modes remain unpriced.
- Commitment discounts: up to 35% below on-demand for multi-month reservations of large clusters (prepaid, sales, company accounts); rates unpublished.

## How this provider charges

As of 2026-09-28. GPU VMs and preemptible VMs.

Use rates BEFORE October 1, 2026. Pricing landing spot floors differ materially from documented preemptible rates: both retained but not conflated. L40S separately bills CPU/RAM; other GPU platform reference bundles include them. Dynamic spot starts October 8, not September 22 announcement date. As of Sep 28 docs fixed preemptible prices still apply.

| Regime | When applicable | Billing | Numbers ($/GPU-hour unless slice) | Source |
|---|---|---|---|---|
| on-demand  | B300 ; count None; uk-south1/eu-west2/us-north1 | per-second; min unknowns | 7.85; list; per_gpu_hour | [source](https://docs.nebius.com/compute/resources/pricing) |
| spot  | B300 ; count None; uk-south1/eu-west2/us-north1 | per-second; min unknowns | 4.3; list; per_gpu_hour | [source](https://docs.nebius.com/compute/resources/pricing) |
| on-demand  | B200 ; count None; us-central1/me-west1 | per-second; min unknowns | 7.15; list; per_gpu_hour | [source](https://docs.nebius.com/compute/resources/pricing) |
| spot  | B200 ; count None; us-central1/me-west1 | per-second; min unknowns | 3.95; list; per_gpu_hour | [source](https://docs.nebius.com/compute/resources/pricing) |
| on-demand  | H200 ; count None; eu-north1/eu-north2/eu-west1/us-central1 | per-second; min unknowns | 4.5; list; per_gpu_hour | [source](https://docs.nebius.com/compute/resources/pricing) |
| spot  | H200 ; count None; eu-north1/eu-north2/eu-west1/us-central1 | per-second; min unknowns | 2.45; list; per_gpu_hour | [source](https://docs.nebius.com/compute/resources/pricing) |
| on-demand  | H100 ; count None; eu-north1 | per-second; min unknowns | 3.85; list; per_gpu_hour | [source](https://docs.nebius.com/compute/resources/pricing) |
| spot  | H100 ; count None; eu-north1 | per-second; min unknowns | 2.15; list; per_gpu_hour | [source](https://docs.nebius.com/compute/resources/pricing) |
| on-demand  | RTX-PRO-6000 ; count None; us-central1/uk-south2/eu-south1 | per-second; min unknowns | 1.8; list; per_gpu_hour | [source](https://docs.nebius.com/compute/resources/pricing) |
| spot  | RTX-PRO-6000 ; count None; us-central1/uk-south2/eu-south1 | per-second; min unknowns | 0.95; list; per_gpu_hour | [source](https://docs.nebius.com/compute/resources/pricing) |
| on-demand  | L40S Intel; count None; eu-north1 | per-second; min unknowns | 1.35; list; per_gpu_hour | [source](https://docs.nebius.com/compute/resources/pricing) |
| spot  | L40S Intel; count None; eu-north1 | per-second; min unknowns | 0.65; list; per_gpu_hour | [source](https://docs.nebius.com/compute/resources/pricing) |
| on-demand  | L40S AMD; count None; eu-north1 | per-second; min unknowns | 1.35; list; per_gpu_hour | [source](https://docs.nebius.com/compute/resources/pricing) |
| spot  | L40S AMD; count None; eu-north1 | per-second; min unknowns | 0.65; list; per_gpu_hour | [source](https://docs.nebius.com/compute/resources/pricing) |

## Gotchas
- Use rates BEFORE October 1, 2026. Pricing landing spot floors differ materially from documented preemptible rates: both retained but not conflated. L40S separately bills CPU/RAM; other GPU platform reference bundles include them. Dynamic spot starts October 8, not September 22 announcement date. As of Sep 28 docs fixed preemptible prices still apply.
- 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.

No fixed single-full-GPU USD quote suitable for a numeric example was verified. Compute = 8,800 × selected offer $/hour, or required node count × node price; keep the estimate null until the offer/FX/commitment is resolved.

## 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.