AI Infrastructure / Supporting Systems

CRWV

CoreWeave

Live Quote

······

In Your Sleeves

Speculative Individual Stocks
5.0%
···

Why I Own It

CoreWeave went public with a differentiated pitch: purpose-built GPU cloud for AI workloads, not a general-purpose cloud with GPUs bolted on. The architecture — bare-metal NVIDIA GPU access, InfiniBand fabric, low-latency storage tuned for AI training — delivers performance per dollar that AWS, Azure, and Google Cloud can't match with their generalist infrastructure. The Microsoft anchor relationship and $15B+ contracted backlog provide meaningful near-term revenue visibility, which matters for a capital-intensive business still in its early innings. I sized this as a higher-risk, higher-conviction bet on the structural shift of AI training and inference workloads toward purpose-built infrastructure.

Why This Sleeve

CRWV is in the taxable retail portfolio because it's a recently IPO'd company that will have significant lock-up expiration volatility and needs ongoing capital to fund GPU fleet expansion. Active position management as the story evolves is important here — the IRA's long-duration orientation doesn't suit a business requiring frequent reassessment of capital structure and competitive dynamics.

Investment Thesis

CoreWeave is a specialized cloud provider purpose-built for GPU-intensive AI and machine learning workloads. Unlike the hyperscalers, which serve general-purpose enterprise computing, CoreWeave runs an infrastructure stack optimized exclusively for AI: NVIDIA H100 and H200 clusters connected via InfiniBand, with storage, networking, and orchestration specifically tuned for distributed training jobs. This architectural focus allows CoreWeave to deliver meaningfully better GPU utilization and performance per dollar than general-purpose cloud providers. The Microsoft relationship — which includes both a cloud partnership and a large contracted workload commitment — validates the differentiated positioning.

The core thesis is that enterprises running serious AI training workloads will increasingly choose purpose-built GPU clouds over general-purpose alternatives, and CoreWeave is the clearest expression of that trend as a public equity. The contracted backlog provides revenue visibility that most growth companies at this stage don't have. The key risks — capital intensity and customer concentration — are real, but the market is pricing them at a discount to what I believe the long-term revenue trajectory supports. If CoreWeave executes on customer diversification beyond Microsoft and converts its pipeline to contracts, the multiple should expand meaningfully.

Scenario Analysis

Bull Case

Purpose-Built GPU Cloud Wins Enterprise

Enterprise AI workloads standardize on CoreWeave's optimized GPU cloud, capturing a structural share of the market from general-purpose clouds.

  • Enterprise AI training workloads migrate from AWS/GCP/Azure to CoreWeave's optimized infrastructure

  • NVIDIA partnership deepens, providing priority access to next-generation GPU generations

  • Microsoft contract expands; additional hyperscaler partnerships announced

Base Case

Contracted Backlog Executes on Schedule

CoreWeave's contracted pipeline converts to predictable revenue, with customer diversification proceeding.

  • $15B+ contracted backlog delivers revenue through 2026 with high visibility

  • Enterprise customer additions diversify revenue beyond hyperscaler anchor contracts

  • GPU utilization rates sustain above 80%, supporting unit economics

Bear Case

Hyperscalers Commoditize GPU Cloud

AWS, Google, and Azure aggressively expand their own GPU capacity, undermining CoreWeave's differentiation and pricing.

  • AWS Trainium and Google TPU deployments reduce demand for NVIDIA-based clusters

  • General-purpose clouds leverage existing enterprise relationships to win AI workloads

  • CoreWeave's capital intensity creates refinancing risk if revenue growth disappoints

Key Risks

  1. 01

    Business model requires continuous large-scale capital raising against long-duration GPU leases, creating refinancing risk.

  2. 02

    Revenue highly concentrated in a small number of customers — Microsoft alone represents a large portion.

  3. 03

    Technology risk if NVIDIA GPU pricing or availability shifts materially.

  4. 04

    Competition from well-capitalized hyperscalers with existing enterprise relationships and bundled services.

What I'm Watching

  • Customer diversification — new enterprise customers and the Microsoft revenue concentration ratio.

  • GPU utilization rates as a measure of asset efficiency against the capital deployed.

  • New contracted backlog announcements — pipeline conversion is the key forward indicator.

  • Debt structure and refinancing timeline — capital market conditions matter significantly here.

  • Competitive GPU cloud pricing from AWS, Google, and Azure, which sets the ceiling on CoreWeave's rate card.