This is a segment from The Breakdown newsletter. To read more editions, subscribe.
“I wish upon you ample doses of pain and suffering.”
— Jensen Huang, on building resiliency
The history of the world’s most valuable company is a study in how improbable its survival was, let alone its success.
In The Nvidia Way, Tae Kim tells the story of how Nvidia, founded in 1993, emerged from the small and volatile market for graphics chips — an industry in which everyone was always “thirty days from going out of business,” as Jensen Huang constantly reminded his employees.
That was more or less true for much of Nvidia’s existence, not only because money was tight for most of its existence, but because graphics chips were so interchangeable that one poorly designed product could sink the company.
After Nvidia’s first two chips were poorly designed, the company really was just weeks away from insolvency.
“We were just bad at our jobs,” Jensen told Kim, who refers to Huang by his adopted first name throughout.
(Born in Taiwan, Jen-Hsun Huang anglicized his first name because his parents sent him to a reform school for juvenile delinquents in Kentucky under the mistaken impression it was a boarding school for high achievers.)
Nvidia’s third chip only happened because Jensen took what little venture capital the company had left and gambled it all on an accelerated production process to get it done while he could still make payroll.
It was a hit and the company survived another cycle.
No one was more surprised than the industry leader at the time, 3dfx, which was forming plans to buy Nvidia out of bankruptcy.
But shortly thereafter, Nvidia was the industry leader and 3dfx was bankrupt.
(Nvidia bought 3dfx’s patents out of bankruptcy and hired 100 of their engineers.)
Nvidia would continue to “oscillate between near-death experiences and market-defining success,” Kim writes, which does much to explain the company’s culture — the world’s largest company still operates as if it’s 30 days from going out of business.
Much of that is a reflection of Jensen’s personality.
As Kim documents, the billionaire tech founder spends nearly all of his waking hours working, including weekends and holidays.
“Work is how I relax,” he explained to employees who asked why he was bombarding them with emails from a beach.
He expected his employees to do the same and they mostly did: Local competitors took notice that the employee parking lot at Nvidia’s Santa Clara headquarters was typically full on nights and weekends.
Even so, Jensen was tough on them.
His method of managing people, in his own words, was to “torture them into greatness.”
That wasn’t always appreciated.
Jensen used to save his most scathing performance reviews for all-company meetings, believing that a public dressing-down was an opportunity for everyone to learn from someone’s mistake.
(“Feedback is learning. For what reason should you be the only person to learn from this?”)
Many talented engineers left the company because of that kind of treatment: “It was the most embarrassing, humiliating thing I’ve ever seen,” an executive told Kim of one such incident.
But Jensen was equally hard on himself: “I look in the mirror every morning and say, ‘You suck.’”
Nvidia adopted that harsh personality — the “Nvidia way” of Kim’s title is essentially the mindmeld between Jensen and the company he co-founded with two friends.
(One co-founder quit the company in 2000 and sold his entire stake for about $30 million. His 3.5% of the company would be worth about $140 billion today.)
That may have made Nvidia the most unpleasant place to work in Silicon Valley, but it’s made it a survivor.
Even when Nvidia had survived long enough to IPO in 1999, people doubted its chances of holding on much longer.
“Why would we invest in a graphics company?” one investor asked. “You would be the 40th we’ve backed, and all the others have gone out of business. Why would we do this?”
The reason, Kim’s story makes clear, was Jensen.
His maniacal work habits, adopted by the whole company, kept Nvidia alive in the early years, and his uncanny ability to see around corners kept Nvidia at the cutting edge of technology.
Jensen was among the first to recognize that PC gaming would be a bigger market for graphics chips than business workstations.
He was among the first to recognize that graphics chips, like CPUs, could become programmable “general processing units” by running software, which Nvidia went on to invent.
(Nvidia contracted Mythbusters to create a fun explainer on the difference between GPUs and CPUs.)
Most importantly, he was among the first to recognize that GPUs would become the enabling technology of AI.
The same GPUs that Nvidia had optimized to do the math behind the arc of an arrow in Zelda turned out to be uncannily well-suited for the matrix math that powers today’s AI models.
From afar, it seems like Nvidia lucked into it: GPUs just happened to be good for AI and Nvidia just happened to make them.
But Kim’s account shows that Jensen spotted the breakthrough progress in deep-learning AI while it was still an academic curiosity.
It’s hard to imagine any other CEO being that early to such a deeply technical trend — and impossible to imagine any of them betting their company on it.
“Deep learning is going to be really big,” he told Nvidia executives in 2013, nine years before the release of ChatGPT. “We should go all in.”
Only with hindsight was that an obvious thing to do.
At the time, Nvidia sales were stagnant, its market capitalization was just $8 billion and its shares traded on 10x P/E, net of cash.
Nvidia’s own investors were so low on the company’s prospects that they badgered Jensen into spending $1 billion of the company’s cash buying back shares; they thought there was nothing else to do with it.
Fortunately, Jensen was simultaneously investing billions in getting Nvidia ready for the AI moment that hardly anyone else saw coming.
In many ways, Jensen created that moment.
Without the billions that Nvidia invested in optimizing its chips and software for AI (years before they could show investors any return on it), the ChatGPT phenomenon would not have happened how and when it did.
Nvidia’s GPUs made large-scale deep learning feasible and they wouldn’t have made them with any other CEO.
