Company profile
SEC EDGAR ↗- Industry
- Semiconductors & Related Devices
- Listed
- Nasdaq: NVDA
- Incorporated
- DE
- Headquarters
- Santa Clara, CA
- Fiscal year end
- Jan 31
Financials
FY2026- Revenue
- $215.9B
- Net income
- $120.1B
- Total assets
- $206.8B
| Fiscal year | Revenue | Net income | Total assets |
|---|---|---|---|
| FY2026 | $215.9B | $120.1B | $206.8B |
| FY2025 | $130.5B | $72.9B | $111.6B |
| FY2024 | $60.9B | $29.8B | $65.7B |
| FY2023 | $27.0B | $4.37B | $41.2B |
| FY2022 | $26.9B | $9.75B | $44.2B |
| FY2021 | $16.7B | $4.33B | $28.8B |
Reported fiscal years from SEC filings (10-K) — SEC EDGAR ↗. Stratrix is a strategy publication, not a real-time financial data service.
In depth
Decisions on record
5 verified Decision Records on Nvidia, and the casebook each one belongs to.
| Fork date | Decision type | The story | Casebook |
|---|---|---|---|
| 2006 | Bet-the-Company | Read the story → | Bet-the-Company Casebook → |
| 2006 | Adjacency Expansion | Read the story → | Adjacency Expansion Casebook → |
| 2006 | Moat Anatomy | Read the story → | Moat Anatomy Casebook → |
| 2022 | The Counterfactual | Read the story → | The Counterfactual Casebook → |
| 2022 | Vertical-Integration Bet | Read the story → | Vertical-Integration Bet Casebook → |
Run one of these yourself
The decisions Nvidia faced, as kits you can run.
Each pack takes one of these decisions and gives you the framework, a deck for the room, a model that resolves to one number, and the sourced cases behind it. $499, one-time.
Decision ForksThe All-In BetHow much does this bet commit before anybody outside the company acts on it?See the pack →Growth & PortfolioThe Adjacency ExpansionWhen does the business next door belong to you?See the pack →Decision ForksThe Road Not TakenWhat did the decision you did not take actually cost?See the pack →
The analyses
The Bet-the-Company · Decision Forks
The Decade of Looking Wrong: How Nvidia's CUDA Bet Survived Wall Street's Contempt
For nearly ten years, Nvidia spent billions making its chips do something almost no one wanted. Analysts called it a margin drag. Then the world needed exactly that — and a graphics-card maker became the most important company of the AI era.
7 min
The Adjacency Expansion · Growth & Portfolio
Nvidia Started CUDA in 2004 and Absorbed an 80% Crash Waiting for the Bet to Pay
The legend says Nvidia rode a sudden AI wave from gaming to the datacenter. The truth: it started building CUDA in 2004, saw almost no payoff for nearly a decade, and watched its stock fall ~80% in 2008. The pivot was a long, expensive infrastructure bet — not a vision.
7 min
The Adjacency Expansion · Growth & Portfolio
Nvidia's One Bet on Parallel Compute Was Re-Licensed From Gaming to Science to Data Centers
Nvidia is read as a graphics company that got lucky on AI. It wasn't luck. It made one bet — parallel compute — and re-licensed it from gaming to science to data centers. By FY2024, the data center it once didn't sell to was 78% of a $60.9B business.
8 min
The Counterfactual · Decision Forks
Five Jurisdictions Cornered Nvidia Out of the $40B Arm Deal, at a $1.353 Billion Charge
Everyone says the FTC blocked Nvidia's $40B Arm acquisition. It didn't - no regulator ever issued a final order. Nvidia walked away on February 7, 2022, cornered by five jurisdictions at once, and ate a $1.353 billion charge for the privilege.
8 min
The Moat Anatomy · Moat & Competition
Nobody Can Out-Build Nvidia. They're Trying to Out-Wait It Instead.
Nvidia did $115.2B in data-center revenue in fiscal 2025, then kept accelerating to a $51.2B single quarter at a 73% gross margin by late 2025. The threat isn't a better chip - it's the market quietly splitting in two, with a fast-growing slice of China already walled off by Washington.
8 min
The Moat Anatomy · Moat & Competition
5.9 Million Developers Keep Nvidia Winning Against Chips With 1.5x the Compute
Everyone says you can't port CUDA off Nvidia. You can - AMD's HIP translates most of it. The real wall is that 5.9 million developers and ~250 hand-tuned libraries make Nvidia the path of least resistance even when rival silicon has 1.5x the raw compute.
8 min
The Pricing Power · Business Model
The H100's Price Has Almost Nothing to Do With the Chip
An H100 costs an estimated $3,320 to build and sold for as much as $40,000. The reflex is to call that a hardware monopoly. It isn't. The premium is rent on seventeen years of software nobody wants to rewrite.
8 min
Vertical-Integration Bet · Boundaries of the Firm
Nvidia built the RTX 30 on Samsung's 8nm, then moved gaming chips back to TSMC's 4N for the RTX 40
In 2020 Nvidia put its flagship gaming GPUs on Samsung's cheaper 8nm process. Two years later it moved the RTX 40 series to TSMC's denser 4N node. It looks like a switch and a switch-back - but Nvidia never actually left Samsung, and the real driver wasn't the story everyone tells.
7 min