Marvell Technology, Inc.

Marvell Technology, Inc. (MRVL) — Investment Memo

🐂 The Bull Case (Warren's voice)

• The Power of the "Design-In": Once Marvell’s electro-optics or custom ASICs are designed into a hyperscaler's architecture, they are locked in. Ripping out this high-speed plumbing is akin to trying to replace the plumbing in a skyscraper while the tenants are still taking showers.
• Exceptional Core Unit Economics: When the semiconductor cycle runs hot, the economics are beautiful. Operating margins of 32.6% in 2026 show that the world is willing to pay a premium for specialized silicon that eliminates data center bottlenecks.
• Picks and Shovels of the AI Gold Rush: Marvell does not need to guess which AI software or model will win. They simply sell the high-speed connectivity chips that every data center requires to prevent expensive GPUs from sitting idle.
• The Buyout Price: Berkshire would only become interested if the market priced Marvell as a boring, cyclical hardware manufacturer rather than an AI darling. We would want a price that values the business on its normalized, historical cash flows rather than its optimistic, cycle-peak projections.

🐻 The Bear Case (Charlie inverts)

• Scenario 1: The In-Sourcing Guillotine (Most Likely, 5–10 Year Horizon): Marvell's greatest strength is also its structural flaw. Relying on "four major customers" for the bulk of revenue is a hostage situation; if those hyperscalers continue their current trajectory of designing custom silicon in-house, Marvell’s moat is instantly bypassed.
• Scenario 2: The Taiwan Strait Choke Point: With sales heavily concentrated in China and Taiwan, and manufacturing outsourced to leading-edge foundries in Taiwan, a single geopolitical spark in the region would stop Marvell's business cold. They cannot sell what they cannot fabricate, and they cannot fabricate without Taiwan.
• Scenario 3: The Debt and Dilution Treadmill: This is not a self-funding compounder. To grow, management loaded the balance sheet with $4.5B in permanent debt and routinely dilutes shareholders with stock-based compensation, resulting in a five-year GAAP net loss of $2.7B from 2021 to 2025.

💰 Valuation & Margin of Safety

Our DCF model assumes an optimistic 15.0% FCF growth rate, a 10% discount rate, and a 3% terminal growth rate:

• Intrinsic Value Estimate: $59.04 per share (Total valuation of $50.0B).
• 25% Margin of Safety Entry: $44.28 per share.
• 50% Margin of Safety Entry: $29.52 per share (Buffett's preferred entry).
• The Verdict on the Price: The DCF model is deceptively precise. Relying on a 15.0% FCF growth rate for a highly cyclical business with a volatile margin history (from -16.9% in 2024 to 32.6% in 2026) is a dangerous game. The current valuation is expensive because it prices in permanent AI prosperity while ignoring the debt, dilution, and customer concentration.

Verdict: PASS

We choose to PASS on Marvell Technology because the business lacks the predictable, long-term moat required to protect our capital over a twenty-year horizon. The narrow switching costs are highly vulnerable to vertical integration by a tiny handful of dominant customers who hold all the leverage. Even at our 50% margin of safety price of $29.52, we will not partner with a management team that uses aggressive leverage and heavy dilution to run on a highly cyclical technology treadmill.