CoreWeave — research brief

Slug: coreweave · Date: 2026-09-21 · Theme: ai-infra (GPU Cloud)

Labels

Company snapshot

Teaching arc (revenue ladder)

Spot (interruptible, discounted) → On-demand PAYG GPU/CPU hours → Reserved / committed capacity (up to 60% off on-demand) → Multi-year take-or-pay enterprise & hyperscaler contracts (+ storage/networking/managed software).

Unlike freemium SaaS, CoreWeave has no $0 consumer free tier on the published rate card; land is on-demand/spot self-serve; expand is reserved + long-term contracts that dominate revenue.

Pricing (FACT — coreweave.com/pricing, North America unless noted)

SKU On-demand / hr Spot / hr Inference single-GPU / hr (where listed)
HGX H100 (8×) $49.24 $19.71 $6.16
HGX H200 (8×) $50.44 $20.93 $6.31
HGX B200 (8×) $68.80 $34.11 $8.60
GB200 NVL72 (4) $42.00 N/A $10.50
GH200 (1) $6.50 N/A $6.50
A100 (8×) $21.60 $9.65 $2.70
L40S (8×) $18.00 $7.88 $2.25
L40 (8×) $10.00 $6.27 $1.25
RTX PRO 6000 Blackwell HM (8×) $20.00 $11.09 $2.50

Financials (FACT — FY2025 IR release 2026-02-26)

Customers / partners (FACT — IR highlights + company news)

Products

Sources

  1. https://www.coreweave.com/pricing
  2. https://www.coreweave.com/
  3. https://investors.coreweave.com/news/news-details/2026/CoreWeave-Reports-Strong-Fourth-Quarter-and-Fiscal-Year-2025-Results/
  4. https://www.coreweave.com/news/coreweave-announces-agreement-with-openai-to-deliver-ai-infrastructure
  5. https://www.coreweave.com/news/nvidia-and-coreweave-strengthen-collaboration-to-accelerate-buildout-of-ai-factories
# CoreWeave — research brief

**Slug:** coreweave · **Date:** 2026-09-21 · **Theme:** ai-infra (GPU Cloud)

## Labels
- **FACT** — primary company page, pricing, or IR release
- **SECONDARY** — press / analyst paraphrase of filings
- **INFERENCE** — model teaching, explicitly labeled

## Company snapshot
- **Legal:** CoreWeave, Inc. (Nasdaq: **CRWV**); IPO March 2025 (FACT — company about / IR)
- **HQ:** Livingston, NJ (FACT — site footer)
- **Founded:** 2017 (FACT — schema.org / about)
- **Positioning:** “The Essential Cloud for AI™” — specialized AI hyperscaler / GPU cloud (FACT — homepage, IR)

## Teaching arc (revenue ladder)
**Spot (interruptible, discounted) → On-demand PAYG GPU/CPU hours → Reserved / committed capacity (up to 60% off on-demand) → Multi-year take-or-pay enterprise & hyperscaler contracts (+ storage/networking/managed software).**

Unlike freemium SaaS, CoreWeave has no $0 consumer free tier on the published rate card; land is on-demand/spot self-serve; expand is reserved + long-term contracts that dominate revenue.

## Pricing (FACT — coreweave.com/pricing, North America unless noted)
| SKU | On-demand / hr | Spot / hr | Inference single-GPU / hr (where listed) |
|---|---:|---:|---:|
| HGX H100 (8×) | $49.24 | $19.71 | $6.16 |
| HGX H200 (8×) | $50.44 | $20.93 | $6.31 |
| HGX B200 (8×) | $68.80 | $34.11 | $8.60 |
| GB200 NVL72 (4) | $42.00 | N/A | $10.50 |
| GH200 (1) | $6.50 | N/A | $6.50 |
| A100 (8×) | $21.60 | $9.65 | $2.70 |
| L40S (8×) | $18.00 | $7.88 | $2.25 |
| L40 (8×) | $10.00 | $6.27 | $1.25 |
| RTX PRO 6000 Blackwell HM (8×) | $20.00 | $11.09 | $2.50 |

- **Reserved compute:** up to **60%** discount vs on-demand (FACT — pricing page); contact sales
- **Storage:** AI Object Storage Hot $0.06 / Warm $0.03 / Cold $0.015 / Archive $0.0125 per GB-mo; Distributed File $0.070/GB-mo; **egress free** (FACT)
- **Networking:** NAT, internet transfer, VPC free; Public IP $4/mo; Direct Connect e.g. 10G $1,250 dedicated / $1,500 virtual (FACT)
- **Kubernetes control plane / SUNK:** Free (excl. CPU for SUNK) (FACT)

## Financials (FACT — FY2025 IR release 2026-02-26)
- FY2025 revenue **$5.131B** (vs $1.915B FY2024)
- Q4’25 revenue **$1.572B** (vs $747M Q4’24)
- Revenue backlog **$66.8B** as of 2025-12-31 (company definition: RPO + estimated future under committed contracts)
- Active power >**850 MW**; contracted power ~**3.1 GW** YE2025 (IR)
- Adj. EBITDA FY2025 **$3.093B** (non-GAAP; see IR reconciliation)

## Customers / partners (FACT — IR highlights + company news)
- Named: Cognition, CrowdStrike, Cursor, Mercado Libre, Midjourney, Runway; expanded hyperscaler relationships
- Historical / S-1 era named ecosystem: Microsoft, OpenAI, Meta, IBM, Cohere, Mistral, NVIDIA (S-1 / company news — treat as disclosed relationships; concentration evolves)
- OpenAI infrastructure agreement (company news Mar 2025; up to $11.9B contract value claim)
- NVIDIA Exemplar Cloud for GB200 NVL72 training; NVIDIA collaboration / investment news (Jan 2026 company/NVIDIA release)
- Acquisitions noted in FY25 IR: OpenPipe, Marimo, Monolith (platform expansion)

## Products
- GPU / CPU compute, storage, networking, managed software, Mission Control, Kubernetes-native platform, serverless RL, CoreWeave Federal (IR + homepage)

## Sources
1. https://www.coreweave.com/pricing
2. https://www.coreweave.com/
3. https://investors.coreweave.com/news/news-details/2026/CoreWeave-Reports-Strong-Fourth-Quarter-and-Fiscal-Year-2025-Results/
4. https://www.coreweave.com/news/coreweave-announces-agreement-with-openai-to-deliver-ai-infrastructure
5. https://www.coreweave.com/news/nvidia-and-coreweave-strengthen-collaboration-to-accelerate-buildout-of-ai-factories