CRWV

1. Executive Summary

• CoreWeave is accelerating capacity-led growth with a newly announced £1.5B expansion in the UK (total UK commitment now £2.5B) and an intended up to $6B US build in Pennsylvania, while deepening vertical integration via an all-stock acquisition of Core Scientific (closing targeted Q4’25) and software stack expansion (Weights & Biases closed; OpenPipe acquisition announced).
• Commercial momentum is anchored by a multi‑year, up to $11.9B OpenAI infrastructure contract (Mar’25) plus enterprise partnerships (IBM Granite, Aston Martin Aramco) and “first-to-deploy” leadership across NVIDIA Blackwell platforms, supporting premium pricing and utilization.
• Cost structure and cash efficiency poised to improve via elimination of >$10B cumulative future lease overhead and $500M run‑rate cost savings by 2027 (pending Core Scientific close), offsetting capex intensity from rapid global buildouts.

Key Takeaways

• Rapid capacity scale-up: £1.5B new UK phase; US $6B Lancaster plan; continued first‑to‑market GPU deployments (GB300 NVL72, RTX PRO 6000 Blackwell).
• Margin and cash discipline: planned $500M annual cost savings by 2027 and >$10B lease overhead eliminated via verticalization (post‑close).
• Contract visibility: OpenAI deal up to $11.9B plus IBM/Aston Martin partnerships underpin utilization and revenue visibility.
• Stack differentiation: W&B integration and OpenPipe RL acquisition extend from infrastructure to tooling, improving customer lock‑in and monetization.

2. Financial Performance

Capital Raises & Proceeds

• IPO (Mar 28, 2025; Nasdaq: CRWV): 37.5M shares priced at $40; CoreWeave sold ~36.59M shares (gross proceeds ≈ $1.46B before fees). Underwriters’ option up to 5.625M (~$225M additional potential gross). Latest and primary equity raise on record.
• Strategic investment: $350M OpenAI investment (stock issuance) tied to the up to $11.9B infrastructure contract (Mar’25), signaling strong customer alignment.
• No convertibles/notes disclosed in the provided materials.

Early Revenue Initiatives

• OpenAI multi‑year contract up to $11.9B (Mar’25) for training and delivery infrastructure—anchor customer with high utilization potential.
• IBM: GB200 Superchip-enabled AI supercomputer for Granite model training (Feb’25) and broader collaboration on hybrid AI cloud.
• Aston Martin Aramco (May’25): Official AI Cloud Computing Partner; migration of on‑prem to CoreWeave cloud for AI‑accelerated engineering.
• Continued traction with leading AI labs and enterprises; first-access to next‑gen NVIDIA platforms supports premium SKUs and throughput.

Expense Management & Cash Flow

• Pending Core Scientific acquisition (Jul’25): expected to eliminate >$10B cumulative future lease obligations; $500M run-rate cost savings by end‑2027; leverage‑neutral with access to broader financing sources.
• Efficiency tailwinds from technology mix: GB200 NVL72 instances cited up to 25x lower TCO and energy consumption for real-time inference vs. prior gen (Feb’25)—benefits customers and supports pricing power/volume.
• Near-term headwind: outsized capex from capacity expansion (UK £2.5B cumulative; US Lancaster up to $6B). Risk managed by verticalization and contracted demand signals.

3. Guidance and Future Outlook

Production Ramp–Up

• UK: Two AI data centers (Crawley, Docklands) operational as of Jan’25; new £1.5B phase adds sovereign AI deployments with NVIDIA GB300 and RTX PRO Blackwell GPU fleets; powered by renewables and closed‑loop cooling (Sep’25).
• US (Lancaster, PA): Intent to invest up to $6B; initial 100 MW facility with expansion potential to 300 MW; job creation and regional hub positioning (Jul’25).
• Ongoing global scale: 33 AI data centers cited (28 US) as of mid‑2025; continued rapid deployments expected.

Expansion Plans

• Accelerated UK and US builds; partnerships with NVIDIA, DataVita, Digital Realty, Global Switch enhance speed-to-capacity and sustainability profile.
• Platform expansion via W&B (closed May’25) and OpenPipe (definitive agreement Sep’25) adds training, evaluation, RL/fine‑tuning features to the AI cloud.
• CoreWeave Ventures (Sep’25) to seed ecosystem growth (capital + compute-for-equity), driving pipeline and platform stickiness.

Operational Targets

• Post‑Core Scientific close: $500M annual run‑rate cost savings by 2027; >$10B lease overhead eliminated.
• Sustainability/efficiency: renewable‑powered sites; closed‑loop cooling; next‑gen GPUs with materially higher perf/Watt and lower TCO for customers.

4. Strategic Positioning and Initiatives

Cost Management

• Verticalization of data centers (Core Scientific deal) to compress unit costs, improve margin per GPU hour, and de‑risk lease exposure.
• Mix shift to Blackwell and GB200 NVL72 improves utilization and revenue density per rack/power footprint.

Product Development

• W&B integration: Mission Control Integration, W&B Inference, Online Evaluations (Jun’25) streamline model lifecycle on CoreWeave.
• OpenPipe RL: Adds reinforcement learning and fine‑tuning capabilities, improving model performance/reliability at lower cost.
• Continuous “first-to-offer” cadence on NVIDIA platforms (GB300 NVL72; RTX PRO 6000 Blackwell Server Edition).

Market Expansion

• Ventures program provides capital + compute + go‑to‑market, catalyzing customer creation and deep technical integration.
• Strategic partnerships (OpenAI, IBM, Aston Martin Aramco) showcase cross‑industry use cases, enhancing enterprise credibility.
• UK sovereign AI positioning aligns with the UK Compute Roadmap; US builds support national competitiveness.

5. Competitive Positioning and Market Trends

Market Positioning

• Positioned as a specialist AI cloud with fastest access to cutting‑edge GPUs and software tooling, competing on performance, time‑to‑capacity, and developer experience rather than generic compute.

Competitive Strengths

• First-mover on GB200/GB300 and RTX PRO 6000 Blackwell; Platinum rating in SemiAnalysis GPU Cloud ClusterMAX.
• Contracted demand evidence (OpenAI; IBM) enhances utilization visibility and GPU allocation priority with NVIDIA.
• Full-stack approach (infra + W&B + RL via OpenPipe) deepens stickiness and drives higher‑margin software‑adjacent revenue.

Emerging Industry Trends

• Shift to agentic AI and larger context/reasoning models increases demand for tightly coupled training + inference clusters.
• Rising sovereign AI demand favors local, sustainable, compliant deployments (e.g., UK roadmap).
• Vertical integration among AI clouds to control power, real estate, cooling, and fleet costs—CoreWeave is aligning ahead of trend.

6. Technology and Innovation Strategy

Technological Advancements

• Operationalization at scale of NVIDIA GB200 NVL72 and GB300 NVL72, enabling high‑throughput training and lower‑latency inference.
• Infrastructure optimized for AI workloads (Kubernetes and observability stack), enabling better scheduling/utilization.

New Product Developments

• W&B Inference and Online Evaluations accelerate deploy/iterate cycles; OpenPipe ART toolkit extends RL training for agents.
• RTX PRO 6000 Blackwell instances offer cost‑efficient alternatives for LLMs and text‑to‑video, broadening SKU coverage.

Alignment with Market Needs

• Performance per watt, sustainability, and rapid capacity turn‑up align with enterprise and lab priorities amid GPU scarcity.
• Tooling integrations reduce time‑to‑results, a key driver for AI labs seeking faster iteration.

7. Risk and Reward Analysis

Growth Catalysts

• Conversion of pipeline into long‑term contracts similar to OpenAI deal; continued first‑access to NVIDIA Blackwell supply.
• Closure and integration of Core Scientific unlocking material cost savings and margin expansion.
• UK/US capacity go‑lives translating into high utilization and premium pricing on next‑gen SKUs.

Downside Risks

• Execution risk on mega‑capex buildouts (power, permitting, supply chain, commissioning).
• Customer concentration (OpenAI) and potential pricing pressure as hyperscalers and peers scale Blackwell fleets.
• Regulatory/approval risk for Core Scientific acquisition; integration complexity across operations and systems.
• GPU allocation risk if supply/demand dynamics shift or if competitors secure priority allocations.

Valuation Metrics

• Public financials not provided in the documents; quantitative valuation requires upcoming filings. A pragmatic framework:
• Revenue base case anchored by contracted/committed capacity (e.g., OpenAI up to $11.9B, multi‑year), layered with enterprise/sovereign wins.
• Apply AI infra peer ranges on EV/Sales given negative/low GAAP earnings during buildout: high‑growth, capacity‑constrained GPU clouds and AI infra names typically command mid‑to‑high single‑digit EV/Sales on forward 12–24 months in strong demand cycles.
• Sensitivity drivers: utilization of new UK/US capacity, pace of Blackwell deliveries, and realization of $500M run‑rate savings by 2027.
• Conclusion: Until audited figures are available, anchor relative valuation to forward revenue visibility and capacity ramp milestones rather than P/E or EBITDA multiples.

8. Investment Thesis

Investment Rationale

• Exposure to the scarce‑asset bottleneck in AI (cutting‑edge GPU clusters), validated by first‑mover deployments and blue‑chip customers.
• Vertical integration and scale economics (Core Scientific) set the stage for structural margin improvement while maintaining growth.
• Software‑enhanced moat via W&B and OpenPipe positions CoreWeave as more than a commodity GPU lessor.

Price Target Justification

• Use a scenario grid tied to forward installed capacity, utilization, and contract mix:
• Base case: staged UK/US capacity ramps, steady NVIDIA allocations, and partial realization of cost savings by 2027; apply mid‑single to high‑single digit forward EV/Sales typical for AI infra leaders.
• Upside: faster Blackwell deliveries, higher‑than‑expected sovereign wins, accelerated Ventures flywheel; premium EV/Sales warranted.
• Downside: GPU/power delays or slower absorption; multiple compression toward low single digits.
• Update numeric targets upon first post‑IPO financial disclosures (revenue, backlog recognition cadence, share count for per‑share math).

Influencing Market Dynamics

• NVIDIA supply cadence, power availability, and hyperscaler pricing posture are the key exogenous drivers.
• Regulatory posture toward sovereign AI and national compute roadmaps support regional premium pricing and localization.

9. Macroeconomic and Industry Trends

Regulatory Changes

• UK Compute Roadmap alignment and government backing for sovereign AI infrastructure support localized, compliant deployments.
• US emphasis on national competitiveness in AI and streamlined permitting (per PA statements) could accelerate time‑to‑build.

Supply Chain Dynamics

• Continued GPU scarcity favors providers with priority allocations and fast install capabilities; partnerships with NVIDIA, Dell, Switch, Vertiv mitigate deployment risk.
• Power and cooling constraints drive value for renewable‑backed, closed‑loop systems; vertical control (post‑Core Scientific) reduces exposure to third‑party leases.

Technology Adoption Trends

• Expansion of agentic AI and long‑context models increases demand for tightly coupled Blackwell clusters.
• Enterprises are migrating from on‑prem to specialized AI clouds for performance and time‑to‑market, as evidenced by recent partnerships.