[CRWV] CoreWeave Thesis 2026: GPU Cloud Backlog Tests Hyperscaler AI Capex Cycle
Key Takeaways
- CoreWeave Inc. (NASDAQ: CRWV) FY2025 revenue ~$5.0-5.5B (+700-770% YoY) with diluted EPS
-$0.50 to +$0.30 reflecting continued explosive AI cloud GPU compute demand growth ($26B aggregate post-IPO backlog March 2025) plus selected post-2024 Microsoft + OpenAI + selected major hyperscaler + AI lab GPU compute customer concentration plus continued ~26 data center facility expansion (~360+ MW operational + ~1.6+ GW contracted) plus selected post-March 2025 ~$1.5B IPO closing under continued co-CEO Michael Intrator (founder + CEO since 2017; ex-Hudson Ridge Asset Management; ~8-year company career). - ~$26B aggregate post-IPO GPU compute backlog: selected primary AI cloud GPU compute provider with ~$26B aggregate backlog March 2025 (selected ~5x trailing FY2024 revenue); ~80%+ Microsoft + OpenAI + selected major hyperscaler + AI lab GPU compute customer concentration; selected post-September 2024 OpenAI ~$11.9B 5-year compute partnership + post-March 2025 OpenAI ~$11.9B incremental partnership.
- 26 data center facility footprint: ~360+ MW operational + ~1.6+ GW contracted FY2025 (vs ~14 facilities + ~145 MW operational FY2024 post-IPO); selected continued post-2024 NVIDIA H100 + H200 + Blackwell + GB200 + GB300 GPU procurement; selected post-2024 ~$1.4B Core Scientific (CORZ) acquisition closing potential.
- Capital structure: selected ~$15-18B aggregate cumulative debt FY2025 (combined senior secured term loans + delayed-draw + other facilities) supporting GPU procurement + data center capex + selected various; no dividend; no buybacks (capital priority on growth + AI infrastructure investment); selected continued NVIDIA strategic partnership including ~$1.3B aggregate FY2024-2025 NVIDIA-led equity investment; investment-grade Caa1/CCC+ credit rating.
Company Background
CoreWeave Inc. (NASDAQ: CRWV) is the leading AI cloud GPU compute provider with FY2025 revenue ~$5.0-5.5B (+700-770% YoY) reflecting continued explosive AI cloud GPU compute demand growth, ~$26B aggregate post-IPO backlog March 2025, and ~80%+ Microsoft + OpenAI + selected major hyperscaler + AI lab GPU compute customer concentration. The company employs ~1,000+ globally with operations across 26 data center facilities + ~360+ MW operational + ~1.6+ GW contracted in US + selected international.
Founded 2017 in Roseland New Jersey by Michael Intrator + Brian Venturo + Brannin McBee as Atlantic Crypto cryptocurrency mining operator (~8-year heritage); selected post-2017 initial focus on Ethereum cryptocurrency mining; selected post-2019 strategic shift from cryptocurrency mining toward general-purpose GPU compute serving visual effects + machine learning + AI customers; selected post-2021 rebranding from Atlantic Crypto to CoreWeave reflecting AI cloud GPU compute focus; selected post-2022 ~$200M Series B + post-April 2023 ~$221M Series B-1 + post-May 2024 ~$1.1B Series C + post-March 2025 ~$1.5B NASDAQ IPO; selected post-2023 NVIDIA strategic partnership + ~$1.3B aggregate FY2024-2025 NVIDIA-led equity + debt investment.
Headquartered in Roseland New Jersey; ~1,000+ employees globally with ~$5.0-5.5B revenue. Single primary product offering: AI cloud GPU compute (~80%+ revenue from Microsoft + OpenAI + selected major hyperscaler + AI lab customers; selected post-September 2024 + March 2025 OpenAI ~$11.9B + ~$11.9B partnership commitments). Geographic footprint: ~26 data center facilities + ~360+ MW operational + ~1.6+ GW contracted including selected post-2024 Texas + Pennsylvania + New Jersey + Virginia + selected various US data center expansion.
CEO + Co-founder Michael Intrator since 2017 founding (~8-year tenure); selected ex-Hudson Ridge Asset Management hedge fund; CFO Nitin Agrawal (since post-2024 Magnet Forensics); CTO + Co-founder Brian Venturo + Chief Strategy Officer + Co-founder Brannin McBee; selected continued NVIDIA strategic partnership.
$26B Aggregate Post-IPO Backlog
Post-March 2025 IPO ~$26B aggregate GPU compute backlog represents selected primary differentiation:
- Total backlog: ~$26B March 2025 (~5x trailing FY2024 revenue ~$1.92B; +1500%+ growth)
- OpenAI partnerships: post-September 2024 ~$11.9B 5-year compute partnership + post-March 2025 incremental ~$11.9B partnership commitment
- Microsoft: ~62% FY2024 revenue concentration; selected major Azure infrastructure partnership
- Selected various AI labs: Anthropic + selected various enterprise AI customers
- Backlog conversion timeline: ~2-5 year revenue recognition
FY2026 catalyst: continued backlog conversion + new customer wins + ~$1.0-2.0 incremental annual EPS contribution.
26 Data Center Facility Footprint
CoreWeave 26 data center facilities + ~360+ MW operational FY2025:
- Operational: ~360+ MW operational FY2025 (vs ~145 MW FY2024)
- Contracted: ~1.6+ GW contracted via long-term lease + selected various (vs ~360 MW FY2024)
- GPU portfolio: NVIDIA H100 + H200 + Blackwell + GB200 + GB300 + selected various
- Geographic mix: Texas + Pennsylvania + New Jersey + Virginia + selected various US states
- Selected post-2024 Core Scientific (CORZ) acquisition potential: post-2024 ~$1.4B aggregate offer (selected ongoing as of March 2025)
FY2026 catalyst: continued 26 → 30+ data center facility expansion + ~$0.30-0.50 incremental annual EPS contribution.
NVIDIA Strategic Partnership
CoreWeave NVIDIA strategic partnership represents selected primary GPU procurement + technical partnership:
- NVIDIA equity investment: ~$1.3B aggregate FY2024-2025 NVIDIA-led equity + selected debt investment
- GPU procurement priority: selected priority H100 + H200 + Blackwell + GB200 + GB300 GPU allocation
- NVIDIA Inception Premier Partner: selected joint go-to-market + technical partnership
- NVIDIA DGX Cloud: selected co-developed AI cloud product offering
- Selected post-2025 NVIDIA Blackwell ramp: continued Blackwell + GB200 + GB300 GPU procurement priority
FY2026 catalyst: continued NVIDIA strategic partnership + Blackwell + GB200 + GB300 GPU procurement.
Risks
- Hyperscaler AI capex cycle: hyperscaler AI capex sustainability vs cyclical adjustment
- Customer concentration: ~80%+ Microsoft + OpenAI + selected major hyperscaler + AI lab GPU compute concentration; major customer churn or capacity reduction risk
- GPU technology cycle: NVIDIA Blackwell + GB200 + GB300 + post-2025 Rubin generation GPU obsolescence + depreciation
- Debt burden: ~$15-18B aggregate cumulative debt + selected continued debt-funded GPU procurement + data center capex
- Hyperscaler in-house capacity: Microsoft Azure + AWS + Google Cloud expanding internal AI cloud capacity could moderate CoreWeave growth
Key Core Metrics
| Metric | FY2025 | FY2024 | FY2023 | FY2022 | FY2026 outlook |
|---|---|---|---|---|---|
| Revenue | $5.0-5.5B | $1.92B | $0.23B | $0.02B | $7-9B |
| Adj. EBITDA | $2.5-3B | $1.05B | $0.07B | -$0.05B | $4-5B |
| Diluted EPS | -$0.50 to +$0.30 | -$1.49 | -$0.62 | -$0.10 | $0.30-0.85 |
| Backlog | $26B+ | $4-5B | $0.5-1B | $0.02B | $30-35B |
| MW operational | 360+ | 145 | 50 | 10 | 600-800+ |
| Capital return | FY2025 | FY2024 | FY2026 outlook |
|---|---|---|---|
| Dividend | None | None | None |
| Buybacks | None | None | None |
| Cumulative debt | $15-18B | $11B | $20-25B |
| Capital priority | growth + AI infrastructure | growth + AI infrastructure | growth + AI infrastructure |
Market Evaluation
CoreWeave trades at selected ~25-35x FY2026 P/E premium vs Microsoft Azure (~30-35x parent) + Amazon AWS (~25-30x parent) + Google Cloud (~25-30x parent) reflecting selected ~$26B aggregate post-IPO backlog + ~80%+ Microsoft + OpenAI customer concentration + selected continued AI cloud GPU compute demand acceleration + NVIDIA strategic partnership. Selected re-rating catalysts include: (1) continued backlog conversion + new customer wins; (2) 26 → 30+ data center facility expansion; (3) NVIDIA Blackwell + GB200 + GB300 GPU procurement priority; (4) selected continued AI inference + training compute demand; (5) FCF positive trajectory toward FY2027.
$26B Aggregate Post-IPO Backlog Deep Dive
CoreWeave $26B aggregate post-IPO GPU compute backlog represents selected unique AI cloud GPU compute differentiation vs hyperscaler AI cloud (Microsoft Azure + AWS + Google Cloud) via (a) selected pure-play AI cloud GPU compute focus eliminating hyperscaler AI cloud profitability vs internal AI workload optimization tradeoffs; (b) selected NVIDIA strategic partnership + ~$1.3B aggregate FY2024-2025 NVIDIA-led equity investment supporting selected priority GPU allocation; (c) selected purpose-built GPU compute infrastructure including liquid cooling + 100Gbps + 400Gbps InfiniBand + selected various AI training-optimized infrastructure; (d) selected ~80%+ Microsoft + OpenAI + selected major hyperscaler + AI lab customer concentration creating selected stable recurring revenue base. Selected backlog progression FY2022 ~$0.02B → FY2023 ~$0.5-1B → FY2024 ~$4-5B → March 2025 IPO ~$26B reflects selected explosive AI cloud GPU compute demand growth driven by hyperscaler AI capex cycle ($300B+ aggregate FY2024 hyperscaler AI capex + $400B+ aggregate FY2025) + selected OpenAI compute partnership commitment. FY2026 catalyst: continued backlog conversion + new customer wins + ~$1.0-2.0 incremental annual EPS contribution.
FY2026 thesis: $26B aggregate backlog conversion + 26 → 30+ data center facility expansion + NVIDIA Blackwell + GB200 + GB300 GPU procurement + AI cloud GPU compute demand sustainability.