Jensen Huang’s 33-year reign at Nvidia ended this week in scandal. But the story that matters to investors, entrepreneurs and the technology industry is how a $40,000 start-up that nearly went bankrupt became the engine of the AI revolution, and what his methods teach anyone who wants to build something that lasts.
There’s a number in Nvidia’s last annual report that explains, better than any tribute, what Jensen Huang built.
In the financial year to January 2026, Nvidia’s revenue rose by about $85 billion, to $215.9 billion. Its operating income rose by roughly $49 billion, to about $130 billion. For every additional dollar of sales, about 57 cents became operating profit (our calculation from its filings).
Few companies of any size manage that. Nvidia did it at a scale where its revenue grew by more than most companies earn in a decade. It also returned $41 billion to shareholders that year and generated $103 billion in operating cash.
How it got there is one of the great business stories of our time.
A diner, three engineers and a bad name
Nvidia was founded in April 1993 by Huang, then 30, and two engineers, Chris Malachowsky and Curtis Priem, who brought experience from Sun Microsystems and IBM. They worked out the plan over cheap coffee at a Denny’s in San Jose, a fitting venue for a founder who had washed dishes at a Denny’s as a teenager. Huang had arrived in the US as a child and has said that, at a Kentucky boarding school at age nine, his chores included cleaning bathrooms.
The founders started with $40,000. They wanted to call the company NVision, until they discovered the name belonged to a toilet paper maker.
Thirty days from the end
Nvidia’s first product almost killed it.
The NV1 chip, launched in 1995, cost about $10 million to develop, money raised from venture firms Sequoia Capital and Sutter Hill Ventures. It tried to do too many things, used a rendering approach that the rest of the industry didn’t adopt, and won few paying customers. Nvidia, barely two years old, was forced to lay off about half its staff.
Its biggest customer relationship was in trouble too. Nvidia had agreed to build a graphics chip for Sega’s next games console, but the design was built on the same flawed approach. Rather than deliver a chip he knew was wrong, Huang flew to Japan to tell Sega’s chief executive the truth. Sega responded with an eleventh-hour $5 million investment, which Huang has credited with keeping Nvidia alive.
The company bet everything that was left on one product, the RIVA 128, designed around the industry standards it had previously ignored. When it launched in August 1997, Nvidia had about a month of payroll left. The chip was a hit, selling about a million units within four months, and the company survived.
That near-death experience shaped Nvidia’s culture for the next three decades. It gave the company its unofficial motto: “Our company is thirty days from going out of business.” Huang used it not as fear, but as a weapon against complacency, even when Nvidia was worth trillions.
The company would face another brush with disaster. In the 2008 financial crisis, its stock fell about 85% from its late-2007 high, while Huang kept pouring money into technology Wall Street didn’t value. He has since said that, knowing how painful the journey would be, he would not start Nvidia again.
In 1999, Nvidia launched the GeForce 256, which it marketed as the world’s first GPU, or graphics processing unit, and went public on Nasdaq. For the next decade, it was a successful but unremarkable company selling graphics chips to video gamers.
The biggest gamble
The decision that defines Huang’s career came in 2006.
Nvidia launched CUDA, a software platform that let programmers use its graphics chips for general-purpose computing, not just for rendering video games. A GPU is designed to do thousands of simple calculations at once, and Huang bet that scientists, researchers and engineers would eventually want that power for problems far beyond gaming.
For years, the bet looked expensive. Nvidia spent heavily building CUDA, giving it away to developers and supporting a market that barely existed. Investors questioned the spending. But the platform kept growing, university by university, lab by lab.
In business terms, Huang was building a moat out of software. Rivals could design competitive chips, but they couldn’t easily replicate a decade of developer tools, libraries and habits. That is why Nvidia has been able to charge premium prices and keep gross margins above 70%.
The graphics chip started a revolution
The payoff arrived from an unexpected direction.
In 2012, a team of researchers at the University of Toronto trained a neural network on two Nvidia gaming GPUs and won a prestigious image-recognition competition by a stunning margin. The result showed that “deep learning,” training huge networks on huge datasets, worked, and that GPUs were the perfect machines to do it. Because researchers had already spent years learning CUDA, they used Nvidia’s chips. Today, virtually all major AI software frameworks are built on CUDA.
Huang saw what was coming and moved fast. Nvidia began designing chips and complete systems specifically for AI. In 2016, he hand-delivered Nvidia’s first DGX-1 AI server to a young research lab called OpenAI.
When OpenAI released ChatGPT in late 2022, demand for AI computing exploded, and Nvidia was the only company ready to supply it at scale. The world’s largest technology companies competed to buy its chips. Larry Ellison and Elon Musk once described pleading with Huang for more GPUs over a sushi dinner.
Huang reframed what Nvidia sold. These were no longer chips, he argued, but “AI factories”: data centres that turn electricity and data into intelligence. Governments bought the idea too. Nvidia’s sovereign AI business, selling to countries building national AI capacity, more than tripled last year to over $30 billion. Customers have deployed nearly 9 gigawatts of its latest Blackwell computing capacity.
The shift shows up starkly in the numbers. In fiscal 2026, Nvidia’s data centre business brought in $193.7 billion, almost 90% of total revenue (our calculation). Gaming, the business the company was founded on, contributed $16 billion.
Nvidia’s market value passed $1 trillion in 2023 and reached about $5.8 trillion at its peak in early October this year. In the latest reported quarter, to July 2026, revenue more than doubled to $96.2 billion.
How he led
Huang ran Nvidia in an unusual way. He reportedly avoided one-on-one meetings with his roughly 55 direct reports, preferring to share information openly in groups so that everyone heard the same thing. The structure was flat, the pace relentless and the strategy driven from the top. He was also its most effective salesman, crossing the globe in his trademark leather jacket and turning product launches into rock concerts.
He spoke openly about the cost. Last year, he said that “to be a CEO is a lifetime of sacrifice,” crediting his immigrant parents with preparing him for the pain of leadership.
His approach had a blind spot. Huang said he didn’t believe in succession planning, and it left Nvidia without an obvious successor when it suddenly needed one.
What others are saying
The tributes this week showed how central Huang had become to the technology industry.
President Donald Trump, a close ally, called the saga horrible and sad. He had phoned Huang several times this week to encourage him to fight calls to resign, CNN reported.
Shaun Maguire, a partner at venture capital firm Sequoia, called Huang “one of the greatest entrepreneurs of this generation.” Elon Musk replied to the post: “I agree.” Hours earlier, when a financial blog urged Musk to hire Huang, Musk wrote: “That would be an honor.”
Mark Zuckerberg had put it simply at a 2024 conference, telling Huang the AI boom was something “you kind of made this happen.”
Nvidia’s board, announcing his resignation, said: “Jensen’s contributions to NVIDIA are profound.” Interim chief executive Colette Kress pledged to build on “the extraordinary foundation Jensen created over the last 33 years.”
Not all reactions were warm. House Speaker Mike Johnson said that by stepping down, Huang “did the right thing for the company and for the country.”
The ending
Huang’s departure came after he admitted to an extramarital affair with a businesswoman whom US officials suspect of spying for China, and reports of a second relationship with a senior employee. China has denied the spying allegations, and the board’s investigation continues. The scandal ended his reign abruptly, and it will shape how his legacy is told.
But for investors and entrepreneurs, the lessons of how he built Nvidia stand on their own.
Survive first. Nvidia nearly died twice. Huang’s response, honesty with Sega, a single all-in bet and a culture of permanent urgency, kept it alive long enough to win.
Bet early on platforms, not just products. CUDA took years to pay off. Its value came from the ecosystem built around it.
Protect the moat with software. Hardware can be copied. Habits, tools and developer communities are much harder to displace.
Let operating leverage do the work. Once the platform was in place, each additional dollar of revenue was enormously profitable.
Plan for your own absence. The one thing Huang chose not to build, a succession plan, is the one Nvidia now needs most.
Nvidia faces its first test without him on 17 November, when it reports quarterly results. Whatever happens next, the company he built from a diner booth, and rescued when it was thirty days from the end, is one of the most consequential businesses in history.
About The Author
