Provider record · updated 2026-09-10
Scaleway — GPU pricing & deliverability
5 tracked offerings · grade mix A:0 · B:0 · C:2 · D:2 · U:1. Official pages publish 'up to 20 Gbps' network bandwidth for H100/H100-SXM instances and 2.5-20 Gbps for L4; NVLink is published for H100-SXM. No InfiniBand or cluster-fabric details published on fetched pages. GPU instances are single-node VMs (max 8 GPUs); no contiguous multi-node cluster size published. Official prices are EUR before tax and zone-dependent (PAR-1/PAR-2/WAW-2); the pricing page only rendered PAR-1 (L4) prices in static fetch, so PAR-2 GPU prices were taken from third-party tracker gpuperhour.com and marked secondary. No A100, H200, B200 or MI300X listed.
Tracked offerings
| GPU | SKU | $/GPU-hr | Terms | Grade | Conf | Fabric (as published) | Src |
|---|---|---|---|---|---|---|---|
| L4 | L4-1-24G EU (Paris PAR-1) | $0.87 | on-demand | D | publ | 2.5-20 Gbps Ethernet | ↗vendor |
| L40S | L40S (L40S-*-48G) EU (Paris PAR-2) | $1.68 | on-demand | U | seco | not published | ↗3rd-party |
| H100-PCIe | H100-1-80G EU (Paris PAR-2, Warsaw WAW-2) | $3.34 | on-demand | D | seco | Up to 20 Gbps Ethernet | ↗3rd-party |
| H100-SXM | H100-SXM (H100-SXM-*-80G) EU (Paris PAR-2, DC5) | $3.61 | on-demand | C | seco | Up to 20 Gbps Ethernet | ↗3rd-party |
| B300 | B300-SXM-8-288G EU (Paris PAR-2) | $8.74 | on-demand | C | seco | not published | ↗3rd-party |
Grades measure what sellers publish — deliverability of the written claim, not measured performance. No fleet holds a measured (Verified) badge yet; that tier arrives with the verification harness. Source classes: API = provider's live pricing API · vendor = provider's published page · mkt API = marketplace API · 3rd-party = aggregator (provider publishes no price on its own pages) · quote = quote-gated or "starting at", not a walk-up price. Quote-class prices are shown in tables but excluded from headline ranges, spreads, and fair-price bands.
Terms & trust (as published, 2026-09-14)
| Uptime SLA | 99.5% SLO for GPU Instances (tiers: Shared Instances 99%, Dedicated 99.5%, Specialized 99.5%, GPU 99.5%; some legacy families have no SLO) — measured on: External connectivity of the instance — 'Unavailability Period' = loss of external connectivity for at least 4 continuous minutes, measured as monthly availability rate (MAR); excludes components under client control |
| SLA remedy | credits (voucher on future invoices, 'cannot be reimbursed in cash') — GPU tier: 10% of monthly amount if 99.0%<MAR<99.5%, 25% if 95.0%<MAR<99.0%, 100% if MAR<95%; capped at 100% of monthly amount for the failing resources; claim within 30 days with log files. Notable exclusion: 'Unavailability of a hardware component due to a stock shortage.' |
| Capacity guarantee | Published reservation offers: 'GPU Cluster On Demand' (reserved H100 clusters, 16-504 GPUs, one week to a few months) and 'Custom-built Clusters' (DGX H100 SXM5) — both sales-led. No capacity guarantee for on-demand single instances (SLA even excludes stock shortages). |
| Egress fees | GPU instances page claims no data-movement fees: 'We ensure predictable billing by eliminating hidden costs for moving data in or out of your infrastructure'; per-instance bandwidth listed as a spec (e.g., 2.5 Gbps); no separate egress price on the GPU pricing page |
| Billing granularity | per-minute for GPU Instances ('billed per minute of uptime, including startup and standby time', except RENDER-S which is per-hour); CPU instances per-hour; flexible IPs and storage billed for the entire allocation period even when powered off |
| Self-serve signup | yes — on-demand GPU instances (incl. H100-SXM) via console/CLI/Terraform, subject to account quotas/IAM; multi-node reserved clusters are sales-led |
| Security certifications | ISO/IEC 27001:2022 (achieved), HDS health-data hosting (achieved), GDPR compliance claims; SecNumCloud in ANSSI qualification, NOT yet achieved; no SOC 2 claim on the security-and-compliance page |
Notes: Minor internal inconsistency: generic pricing doc says 'Instances are billed per hour' while the Instances FAQ and GPU pages specify per-minute for GPU Instances — the more specific GPU statement recorded. SLA compensation is voucher-only.
Sources: ↗ ↗ ↗ ↗ ↗ ↗ ↗ · Fields marked "not published" are exactly that — the seller does not state them anywhere we could find. Absence is a finding.
Reliability record (public status feed, last 90 days)
50+ incidents on Scaleway's own status page in the trailing 90 days, collected 2026-09-15. The feed caps at 50 items — the true count is higher. Self-reported by the provider — a floor on true incident count, not a ceiling; it still beats the war-story thread you were going to read instead. High counts often reflect granular per-rack reporting, not worse infrastructure: read the incidents, not just the number.
| Date | Incident | Impact | Duration |
|---|---|---|---|
| 2026-09-14 | [PGWY] - [fr-par-1] - Public Gateay stuck in a transient state ↗ | none | 0 h |
| 2026-09-14 | [Support] Phone assistance is unavailable ↗ | minor | — |
| 2026-09-14 | [Support] Delays in ticket processing ↗ | minor | — |
| 2026-09-14 | [WBHE] - [fr-par] - Customer websites are slow on pf-014.whm.fr-par.scw.cloud ↗ | minor | 0.9 h |
| 2026-09-11 | [NETI] - DC2/S103/A7 - Public switch reboot ↗ | minor | 1.2 h |
| 2026-09-10 | [generative-apis] - [fr-par ] - gemma4 unhealthy ↗ | minor | — |
| 2026-09-09 | [NETI] - DC3/S45/C5 - Public switch has rebooted ↗ | minor | 0.1 h |
| 2026-09-08 | [NETI] - AMS1/H6/6B52 - Public switch has rebooted ↗ | minor | 0 h |
| 2026-09-08 | [NETI] - DC2/301/D4 - Public switch unavailable ↗ | minor | 0.7 h |
| 2026-09-08 | [VPCL] - it-mil-1 - Network Incident on going at IT-MIL-1 ↗ | major | 0.1 h |
| 2026-09-08 | [Jobs][fr-par] some jobs cannot be started due to internal error ↗ | minor | 5.3 h |
| 2026-09-08 | [MGDB] - [fr-par] - Restore Snapshot is broken ↗ | minor | 25.4 h |
What Scaleway does not publish
- official EUR rates for H100/H100-SXM/L40S/B300 in PAR-2 (client-rendered)
- H100-SXM gpus-per-instance options
- inter_node_fabric beyond 20 Gbps Ethernet
- max_contiguous_gpus
- reserved pricing
- min_commit
- notice_period
Unpublished fields are recorded as unknown and grade conservatively. Sellers improve their grades by disclosing — that is the point.
Weekly context on every provider
The Delivered Compute Report tracks how these prices and grades move.
Subscribe free