NVIDIA Corporation — Through the Warren Buffett Lens
Thesis
NVIDIA designs the accelerated computing platforms that power the world's AI infrastructure. Its GPUs, software ecosystem, and networking solutions form a deeply entrenched standard for training and inference. The economic engine is simple: customers pay for performance that reduces their own cost and time-to-insight, creating powerful switching costs. Revenue has surged to over $130 billion annually, with gross margins above 70%, generating enormous free cash flow. The durable edge lies in a widening moat built on the CUDA software ecosystem, a relentless innovation cadence, and scale advantages that competitors struggle to match. The key number is the return on invested capital, which has averaged well over 50% in recent years, signaling a business that compounds intrinsic value at exceptional rates. The question is whether the current price already discounts a decade of AI dominance, leaving little room for error. Supporting evidence: • Q4 FY2026 revenue reached $68.13 billion, a 73% year-over-year increase, driven by insatiable demand for Blackwell architecture in data centers (Source 5). • Gross margins expanded to 75.0% in FY2025, up from 72.7% in FY2024, reflecting pricing power and a favorable mix shift toward Data Center (Source 3). • The company repurchased 108.3 million shares for $20.0 billion in the three months ended April 26, 2026, demonstrating disciplined capital return at what management likely views as attractive prices (Source 2). Key uncertainties: • Can NVIDIA sustain its current gross margin levels if hyperscaler customers develop in-house AI chips or if competition intensifies? • What is the normalized level of owner earnings after adjusting for the current AI capex super-cycle, and how much of today's revenue is pull-forward from future demand? Management & culture: Jensen Huang (CEO, verify) has led NVIDIA since its founding, instilling a culture of relentless innovation and long-term vision. The company pioneered the GPU and later the CUDA platform, betting on parallel computing years before AI became mainstream. Management communicates with unusual candor about technological challenges and market opportunities. Capital allocation has been exemplary: massive R&D investment to maintain the performance lead, strategic acquisitions (Mellanox) that strengthened the data center position, and aggressive share repurchases when the stock was arguably undervalued. The leadership team thinks in decades, not quarters, and has consistently reinvested in the ecosystem to widen the moat. This is a founder-led, owner-oriented culture that aligns with long-term shareholders.
Key Value Drivers
- Data center AI adoption: hyperscaler and enterprise capex on accelerated computing.
- CUDA ecosystem lock-in: software moat that raises switching costs.
- Product cycle cadence: annual architecture upgrades (Hopper → Blackwell → Rubin) sustaining performance leadership.
- Capital allocation: aggressive buybacks and strategic investments.
- Gross margin durability: pricing power from performance-per-watt advantage.
Key Risks
- Hyperscaler customers develop in-house AI chips, reducing NVIDIA's data center revenue growth. (Trigger: Two or more top-5 cloud customers announce production deployment of internal AI accelerators that match NVIDIA's performance per watt.)
- Gross margins compress due to competitive pricing or mix shift to lower-margin products. (Trigger: Non-GAAP gross margin falls below 70% for two consecutive quarters.)
- AI capex cycle peaks, leading to a sharp deceleration in data center revenue. (Trigger: Data center revenue growth drops below 20% year-over-year for two consecutive quarters.)
- Export restrictions on advanced semiconductors to key markets materially reduce revenue. (Trigger: New U.S. government regulations block sales of Blackwell-class GPUs to China, and NVIDIA's China revenue falls by more than 50% year-over-year.)
Key Metrics to Monitor
- Owner earnings (NI + D&A − maint capex) — not disclosed in fetched sources
- ROE (5yr average) — not disclosed in fetched sources
- ROIC — not disclosed in fetched sources
- Gross margin trend (5yr) — not disclosed in fetched sources
- Retained earnings / market cap — not disclosed in fetched sources
- Debt / equity — not disclosed in fetched sources
- Moat description (type + durability) — not disclosed in fetched sources
- Capital allocation grade (buybacks / debt / M&A) — not disclosed in fetched sources
Want to see how different lenses come to different conclusions?
Run your own thesis →