In April 1993, Jensen Huang founded Nvidia with $600 at a Denny’s diner in San Jose. In October 2025, Nvidia became the first company in history to exceed $5 trillion in market capitalisation. In FY2026, it reported $215.9 billion in revenue — up 65% year-on-year — and $120.1 billion in net income. These numbers describe three decades of bets, near-bankruptcies, and pivots by a founder who understood, years before the market, that the GPU would become the fundamental unit of modern computing.
Sources: Nvidia Annual Reports FY2023–FY2026 | Fortune Q4 FY2026 Earnings (Feb 2026) | Wikipedia: Jensen Huang | Bloomberg Billionaires Index 2025 | Quartr Insights | Fortune Leadership Profile (Nov 2024) | ABC News Nvidia Q3 FY2026 (Nov 2025)
The Man Before the Machine: From Taiwan to Trillion Dollars
Born in Tainan, Taiwan, on 17 February 1963, Jensen Huang was sent to the United States at age nine with his brother, initially living in Kentucky, later attending Oneida Baptist Institute before moving to Oregon. He earned an electrical engineering degree from Oregon State University in 1984 and a Master’s from Stanford in 1992. After stints at LSI Logic and AMD, he convinced two engineering colleagues to start a company at a Denny’s in San Jose. Initial capital: $600 — $200 each from Huang, Chris Malachowsky and Curtis Priem. A lawyer formalised the incorporation for the $200 in Huang’s pocket.
Sequoia Capital’s Don Valentine, introduced via LSI’s CEO, provided the first $20 million in venture capital. Priem reportedly turned to Huang and said: ‘Jensen, you’re the CEO, right?’ Huang has held that title unbroken for 33 years — one of the longest tenures in S&P 500 history.
“Building Nvidia turned out to have been a million times harder than we expected. We probably would not have done it if we had realized the pain and suffering, the challenges we were going to endure.”
— Jensen Huang, Nvidia co-founder and CEO
\🏦 Nvidia initial capital: $600 (3 × $200), Denny’s San Jose, April 1993 (Wikipedia: Jensen Huang)
📊 First VC funding: $20 million, Sequoia Capital and Sutter Hill Ventures (1993) (Wikipedia: Jensen Huang)
💡 CEO tenure: Unbroken since April 1993 — 33 years as of 2026; founding stake ~3.5% (Bloomberg Billionaires Index 2025)
💰 Personal net worth: ~$176 billion (Oct 2025) — 9th richest person globally (Bloomberg Billionaires Index 2025)
Three Pivotal Bets That Built a $5 Trillion Company
Bet 1: Survive 1997 — The RIVA 128 Pivot
Nvidia’s NV1 chip used a proprietary quadrilateral architecture. By 1996, the error was clear: Microsoft had standardised graphics around Direct3D triangle rendering. Thirty competitors had entered the space. With one month of payroll remaining, Huang ordered a complete redesign. Sega provided a $5 million lifeline. The RIVA 128, released August 1997, captured ~80% of the graphics card market within months. Huang later described this — not the trillion-dollar milestones — as Nvidia’s finest hour.
Bet 2: CUDA (2006) — $1 Billion Before a Market Existed
In 2006, Huang allocated approximately $1 billion (~$2.5 billion in 2025 dollars) to build CUDA — an open-source platform for running general-purpose workloads on GPUs. The gaming GPU business was thriving; AI was an academic backwater; no enterprise data centre had ever used a GPU for computation. The investment was, literally, a bet on a future that did not exist.
The logic crystallised in 2012, when AlexNet — trained on Nvidia GPUs — achieved a 41% reduction in image recognition error rates, running 100 times faster than on CPUs. Every major AI laboratory — Google Brain, OpenAI, DeepMind, Meta AI — subsequently built on Nvidia hardware and CUDA. The $1 billion created platform lock-in that competitors have spent billions trying to replicate and largely failed to overcome.
Bet 3: H100 and Blackwell (2022–2025)
When OpenAI released ChatGPT in November 2022, Nvidia was already shipping the H100 — designed specifically for generative AI training workloads before the market existed. The financial results were unlike anything in semiconductor history: Q1 FY2025 revenue was $26 billion, up 262% year-on-year. FY2025 data centre revenue: $115.2 billion. FY2026 data centre revenue: $197.3 billion. The Blackwell Ultra GB300 delivers 50 times the performance of the H100. By Q4 FY2026, quarterly revenue reached $68.1 billion, up 73% year-on-year.
Jensen Huang’s Pivotal Decisions: A Timeline
| Year | Pivotal Decision | Financial Outcome | Strategic Significance |
| 1993 | Founded Nvidia at Denny’s; $600 initial capital | Series A: $20m from Sequoia/Sutter Hill | GPU-first bet before market existed |
| 1997 | RIVA 128 pivot — abandoned NV1, rewrote for DirectX | Captured ~80% graphics market share | Near-bankruptcy survival; reset architecture |
| 2006 | CUDA launch — $1bn investment in general-purpose GPU | Opened non-gaming GPU market | Platform moat enabling AI a decade later |
| 2012 | AlexNet trained on Nvidia GPUs — 100x faster than CPU | Data center revenue begins rapid climb | Validated AI thesis; GPU supremacy confirmed |
| 2022 | H100 ‘Hopper’ launch for generative AI workloads | Data center revenue: $15.0bn (FY2023) | Designed for transformers before market existed |
| FY2026 | Blackwell ramp — $197.3bn data centre revenue | Total revenue: $215.9bn; net income: $120.1bn | First company to exceed $5 trillion market cap |
Sources: Nvidia Annual Reports FY2023–FY2026; Wikipedia: Jensen Huang; Fortune Nov 2024; Quartr Insights Nov 2025)
The Full-Stack Strategy: Why Competitors Cannot Simply Build a Better Chip
Nvidia does not sell a chip; it sells a vertically integrated computing platform: GPUs (Hopper, Blackwell), interconnect fabric (InfiniBand, NVLink), servers (DGX SuperPODs) and the CUDA software stack, including cuDNN libraries, NIM inference microservices and AI Blueprints. A hyperscaler purchasing Nvidia hardware is simultaneously purchasing a developer ecosystem with over a decade of optimised tooling and community knowledge.
AMD’s MI300 series is competitive on raw specifications — but CUDA developers face significant switching costs to move to ROCm. Google’s TPUs and Amazon’s Trainium serve internal workloads but cannot offer the universal deployability of the CUDA ecosystem. As Wedbush Securities’ Dan Ives stated in November 2025: ‘There is one company in the world that is the foundation for the AI Revolution.’ The AI spending boom contributed 0.5 percentage points to annualised US GDP growth in H1 2025 alone (Pantheon Macroeconomics) — and Nvidia captured the overwhelming majority of that capital expenditure.
“Computing demand is growing exponentially. The agentic AI inflection point has arrived.”
— Jensen Huang, Nvidia Q4 FY2026 Earnings Call, February 2026
Three Structural Risks — and Why They Matter for EU Markets
-
Hyperscaler Concentration
More than 50% of Nvidia’s revenue comes from five hyperscalers — Amazon, Microsoft, Google, Meta and Oracle — which collectively committed nearly $700 billion in capital expenditure in 2026 (Fortune, February 2026). Nvidia’s CFO stated the company has visibility to $500 billion in Blackwell and Rubin revenues through end-2026. If hyperscaler capex growth decelerates, the revenue concentration creates non-linear downside risk.
-
China Export Controls
Multiple rounds of US chip export restrictions have eliminated a structural revenue source: in Q3 FY2026, Nvidia reported zero H20 sales to China. Subsequent partial relaxation (contingent on 15% revenue sharing with the US government) has partially reopened the market, but regulatory uncertainty around China access remains an ongoing earnings risk no engineering decision can resolve.
-
EU Competitiveness Implications
European technology companies — SAP, ASML, Siemens and major EU financial institutions — are increasingly dependent on Nvidia infrastructure for AI deployment. The Draghi Report (2024) estimated Europe’s annual digital investment gap at €580 billion; Nvidia’s hardware is a direct input into whether that gap widens or narrows. ASML, which produces the EUV lithography machines used to fabricate Nvidia’s chips at TSMC, remains one of the rare European companies central to the AI supply chain: Q3 2025 revenue €7.5 billion; order backlog exceeding €36 billion.
Conclusion: A Founder Who Saw the Platform Before the Market
Jensen Huang’s decisive bets — CUDA in 2006, the H100 in 2020, the Blackwell ramp in 2024 — were each made before commercial demand was visible. CUDA preceded ChatGPT by 16 years. The H100 was architected for generative AI before generative AI was a commercial category. This is the pattern of genuine platform builders: create infrastructure for products that do not yet exist.
Nvidia’s FY2026 results — $215.9 billion in revenue, $120.1 billion in net income, $197.3 billion in data centre revenue — are the financial expression of those early, unpopular decisions. For EU investors, Nvidia is not simply a semiconductor company. It is the picks-and-shovels supplier of the most consequential technology buildout since the internet, led by a CEO who has held the same position for 33 years and shows no sign of running out of bets to make.
