# CoreWeave

Canonical: https://inetgeek.com/gpu/coreweave/

GPU cloud selling whole HGX nodes by the hour, with free egress and object storage priced separately.

## Who it suits

Training runs that want a whole HGX node and move large datasets — free egress and free I/O operations remove the charge that usually dominates a data-heavy job.

Consider something else if You want one H100 on demand: it is sold as an eight-GPU instance at $49.24 an hour, and the $6.16 per-GPU figure CoreWeave prints applies only to its inference-platform customers.

## iScore

55 / 100 (80% interval 49–61), rank 8 of 13 in gpu — interval overlaps the next rank, so this is an ordering, not a finding.

inetGeek's own reading, computed from sourced facts against category peers by disclosed rules. Not a provider claim. Intervals are 80%; a provider with few documented facts is pulled toward the category middle rather than scored on what little is known.
Read this score with care:
- Price is thin: 4 of 10 criteria are scored, so it sits near the category middle with a wide interval (32–53). Read the interval, not the point.
- Capacity and limits is thin: 1 of 3 criteria are scored, so it sits near the category middle with a wide interval (37–64). Read the interval, not the point.
- Capabilities is thin: 0 of 1 criteria are scored, so it sits near the category middle with a wide interval (94–100). Read the interval, not the point.
- Documented but not scored, because the page states it as text with no comparable number or a yes/limited/no (a one-time, usage-based or other-vendor price, or a per-site choice): regions. Read them on the provider page; the score neither rewards nor penalises what it could not compare.

Method: https://inetgeek.com/iscore/. Full output with every criterion's basis: https://inetgeek.com/scores.json. Every gpu peer ranked: https://inetgeek.com/gpu/coreweave/alternatives/

- Price: 43 (32–53), weight 50%, 4 of 10 criteria scored; scored for peers, not here — undocumented or stated without a number or a yes/limited/no: budget_input_price, flagship_input_price, flagship_output_price, free_card_required, free_offer_type, free_tier
- Capacity and limits: 51 (37–64), weight 24%, 1 of 3 criteria scored; scored for peers, not here — undocumented or stated without a number or a yes/limited/no: context_window, regions
- Capabilities: 100 (94–100), weight 3%, 0 of 1 criteria scored; scored for peers, not here — undocumented or stated without a number or a yes/limited/no: prompt_caching
- Trust and compliance: 74 (68–80), weight 13%, 9 of 12 criteria scored; scored for peers, not here — undocumented or stated without a number or a yes/limited/no: pci_dss, support_included, uptime_sla
- Evidence: 86 (86–86), weight 11%, 4 of 4 criteria scored

## What the documentation says

### Pricing

- Pricing model: Four CKS capacity plans - Flex Reservations, Reserved Instances, Spot Instances and On-Demand; On-Demand instances are billed on actual use with no long-term commitment ([source](https://docs.coreweave.com/platform/capacity-plans), read 2026-10-17)
- Egress over allowance: Free. CoreWeave charges nothing for egress or for I/O operations, and runs a zero-egress migration programme for moving data in. ([source](https://www.coreweave.com/pricing), read 2026-10-17)

### Infrastructure

- Regions: Availability Zones listed across seven Super Regions (CA-EAST, EU-NORTH, EU-SOUTH, EU-WEST, US-CENTRAL, US-EAST, US-WEST), sited in the US, Canada, UK, Spain, Sweden, Norway and Denmark ([source](https://docs.coreweave.com/platform/regions/all-availability-zones), read 2026-10-17)

### Storage

- Storage price: $0.06 per GB per month for AI Object Storage Hot; Warm is $0.03, Cold $0.015 and Archive $0.0125 (limited availability). Distributed File Storage is $0.070 per GB per month. CoreWeave bills storage in binary GB (GiB). ([source](https://www.coreweave.com/pricing), read 2026-10-17)

### GPU

- H100 SXM, per GPU-hour: $6.16 per GPU-hour for NVIDIA HGX H100 (80GB) is CoreWeave's inference-platform price, printed in its "Inference Single GPU Price" column, which it says applies only to inference platform customers. The on-demand instance price is $49.24/hour for the 8-GPU HGX H100 instance, spot $19.71/hour (North America). ([source](https://www.coreweave.com/pricing), read 2026-10-17)
- Cheapest GPU, per hour: $1.25 per GPU-hour for NVIDIA L40 (48GB) is CoreWeave's inference-platform price, from its "Inference Single GPU Price" column, which applies only to inference platform customers. The on-demand instance price is $10.00/hour for the 8-GPU L40 instance, spot $6.27/hour (North America). ([source](https://www.coreweave.com/pricing), read 2026-10-17)
- Largest GPU offered: 279GB per GPU on GB300 NVL72, which is Contact sales only. HGX B300 is next at 270GB, with no on-demand price and a $35.84/hour spot price for the 8-GPU instance. The most VRAM with a published on-demand price is GB200 NVL72 at 186GB per GPU: $42.00/hour instance price for 4 GPUs, or $10.50 per GPU-hour on the inference platform. ([source](https://www.coreweave.com/pricing), read 2026-10-17)
- Billing granularity: Invoice line items are per SKU with quantity in hours; usage attribution across Reserved, Flex and On-Demand capacity runs every 30 seconds ([source](https://docs.coreweave.com/platform/capacity-plans), read 2026-10-17)
- Reserved pricing: Up to 60% below on-demand for reserved compute capacity. ([source](https://www.coreweave.com/pricing), read 2026-10-17)

### Compliance

- SOC 2: SOC 2 Type II is listed as a badge in the CoreWeave Trust Center; detailed documentation is available by requesting access. ([source](https://trust.coreweave.com/), read 2026-10-17)
- ISO 27001: ISO 27001 (with ISO 27017 and ISO 27018) is among the frameworks CoreWeave's compliance evidence maps to, available via its Compliance Documentation Portal. ([source](https://docs.coreweave.com/security/trust-compliance/compliance-programs), read 2026-10-17)
- HIPAA: Security docs say multi-tenant isolation and encryption controls help customers address regulatory requirements such as GDPR, HIPAA and CCPA. Stated as helping customers meet requirements, not as a HIPAA attestation or an offer of a BAA. ([source](https://docs.coreweave.com/security/architecture), read 2026-10-17)
- GDPR / data residency: Regions are placed to meet data residency requirements, and Geos are used to ensure compliance with regulatory requirements. ([source](https://docs.coreweave.com/platform/regions/about-regions-and-azs), read 2026-10-17)
- Encryption at rest: AI Object Storage encrypts object data at rest by default with AES-256 at no additional cost; object metadata is not encrypted by default. Read for CoreWeave AI Object Storage; other storage products document encryption separately. ([source](https://docs.coreweave.com/products/storage/object-storage/about), read 2026-10-17)
- Customer-managed keys: SSE-C lets you encrypt AI Object Storage objects with AES-256 keys you provide and manage; CoreWeave retains only a hash of the key, never the key itself. ([source](https://docs.coreweave.com/products/storage/object-storage/buckets/server-side-encryption/about-sse-c), read 2026-10-17)
- Role-based access control: IAM Access Policies assign roles to users or groups across CoreWeave services; CKS clusters use Kubernetes RBAC and Object Storage its own bucket policies. ([source](https://docs.coreweave.com/security/iam/access-policies), read 2026-10-17)
- Private networking / VPC: VPCs define network isolation for attached products such as CKS clusters; VPC prefixes reach other VPCs, remote clouds or on-premises networks over VPN or Direct Connect. ([source](https://docs.coreweave.com/products/networking/vpc/about-vpcs), read 2026-10-17)
- Dedicated infrastructure: Dedicated Access AZs are single-tenant: no other customer's workloads run in the AZ and no other tenant shares the underlying network fabric; reserved for select customers. ([source](https://docs.coreweave.com/platform/regions/about-regions-and-azs), read 2026-10-17)

## Compared with

- [CoreWeave vs RunPod](https://inetgeek.com/compare/coreweave-vs-runpod/)
- [CoreWeave vs Hyperstack](https://inetgeek.com/compare/coreweave-vs-hyperstack/)
- [CoreWeave vs Lambda](https://inetgeek.com/compare/coreweave-vs-lambda/)
- [CoreWeave vs Vast.ai](https://inetgeek.com/compare/coreweave-vs-vast-ai/)