Philippe Laffont's Strategic Shift: Why CoreWeave Represents a Superior AI Infrastructure Play Over Nvidia

The AI infrastructure race is heating up, and one of Wall Street's most astute investors, Philippe Laffont of Coatue Management, has placed his bets. Over the past two years, Laffont has sold 83% of his firm's Nvidia (NVDA) stake—a position once valued at over $4 billion—while simultaneously loading up on CoreWeave (CRWV), an AI cloud infrastructure upstart. This strategic shift reflects a seismic shift in the sector: a move away from hardware monopolies toward the companies building the scalable, GPU-powered ecosystems that will power the next era of AI. Let's dissect why CoreWeave is now the better bet for investors seeking exposure to AI's growth, while Nvidia faces mounting headwinds.

CoreWeave: The Undervalued Leader in AI Cloud Infrastructure

CoreWeave's stock has surged over 300% in 2025, climbing from its $40 IPO price to an all-time high of $187, fueled by its role as a specialized AI data center operator. The company's 33 facilities, housing 250,000 Nvidia GPUs, are purpose-built for AI workloads, offering 95%+ utilization rates—critical in an era where compute efficiency defines profitability.

Valuation Case for CoreWeave:
While CoreWeave trades at a 39.7x sales multiple, this premium is justified by its $25.9 billion revenue backlog, a 63% year-over-year jump driven by a $11.2 billion OpenAI contract and a $4 billion deal with another hyperscaler. Analysts project 2025 revenue of $5.0 billion, up from $1.9 billion in 2024. Crucially, CoreWeave's $88 billion market cap (as of June 2025) still trades at a discount to its growth runway: its backlog alone implies 3.5x current revenue, suggesting upside if it converts contracts into cash flow.

Competitive Advantages:
- Strategic Partnerships: Beyond OpenAI, CoreWeave's May 2025 deal with Google Cloud diversifies its client base, reducing reliance on Microsoft (which once accounted for 62% of revenue).
- Scalability: Plans to expand data center capacity to 1.6 gigawatts by 2026, backed by a $23 billion capital expenditure war chest.
- Software Integration: Its acquisition of Weights & Biases (a platform used by 1,400 AI teams) adds end-to-end MLOps capabilities, making it a full-stack infrastructure provider.

Why Philippe Laffont Is Right to Exit Nvidia

Laffont's $4 billion pivot from NVDA to CRWV isn't just about chasing growth—it's a valuation and risk calculus. Nvidia's stock has underperformed in 2025, dropping 15% since January despite its dominance in AI chips. Here's why CoreWeave is a safer, higher-growth play:

1. Margin Pressures on Nvidia's Hardware Model

Nvidia's gross margins have fallen from 78% in 2020 to ~60% today, as competition intensifies. Rivals like AMD (with its MI300X GPUs) and Intel (Gaudi processors) are eroding its edge. Worse, clients like Meta and Google are designing their own AI chips, reducing reliance on NVDA. This threatens its pricing power—a $1,500 H100 GPU now sells for closer to $1,000 as supply expands.

2. Regulatory Risks: The DOJ's Crosshairs

The U.S. Department of Justice has launched an antitrust probe into Nvidia's alleged anti-competitive practices, including tying software to hardware sales and leveraging its data center dominance to block rivals. In China, regulators are investigating its 2020 Mellanox acquisition—a move that contributed to a 2.7% stock drop in December . These risks could force divestitures or fines, complicating its ability to invest in R&D.

3. CoreWeave's Debt vs. Nvidia's Monopoly Decline

While CoreWeave carries $8.7 billion in debt, its path to profitability is clearer: it projects narrowing losses to $0.56 per share by 2026. Nvidia, by contrast, faces a maturity wall—its H100/ H200 GPUs are nearing obsolescence as next-gen models (like the Grace Blackwell) emerge. CoreWeave's modular data centers can upgrade hardware faster, ensuring it stays ahead.

The Investment Thesis: CoreWeave's Upside vs. Nvidia's Downside

Buy CoreWeave on dips. Its $185 price target (set by Bank of America despite a "Neutral" rating) reflects its $400 billion total addressable market by 2028. Even skeptics acknowledge its backlog-to-revenue ratio implies 5x upside over the next three years.

Avoid Nvidia unless you're a long-term hardware bettor. While its CUDA ecosystem remains vital, the stock's 30x sales multiple and regulatory overhang make it overpriced for a maturing player. A correction to $150 (2024 lows) is plausible if margin pressures or antitrust settlements materialize.

Conclusion: Infrastructure Wins, Monopolies Fade

The AI revolution isn't about owning the chip—it's about owning the cloud infrastructure that delivers compute at scale. CoreWeave's data center network, strategic partnerships, and software stack make it the sector's most compelling growth story. Meanwhile, Nvidia's reliance on hardware margins and regulatory risks position it as a speculative play at best.

Final Take: CoreWeave is undervalued relative to its backlog and scalability. Investors should buy shares below $150—where Coatue's cost basis lies—and hold for the next wave of AI adoption. Nvidia? Save it for when the dust settles on its regulatory battles.

Disclosure: This analysis is for informational purposes only and not personalized investment advice.