Jensen Huang was BFFs with everyone in Silicon Valley this year, becoming the only person to walk across all three conference stages of Microsoft, Amazon, and Google. Why? Because everyone needs his GPUs to power their AI dreams.
Graphics Processing Units were once only coveted by PC gamers looking to max out their home setup, but now they’re the crown jewel of the most lucrative tech companies in the world. It’s a unique position to be in, but how did Jensen Huang get here? And what is he doing with this opportunity of a lifetime when he’s not on the conference stage?
The Envy of The Valley
Nvidia dominated the AI chip market this year: over 80% of chips to train and run AI models in 2023 were made by Nvidia according to Baird analysts. When you hear that Big Tech companies like Apple spend “millions of dollars a day” training AI, a chunk of that goes to Nvidia and even more money would flow if there were more units to go around. Jensen Huang named his company Nvidia after the Latin word for “envy,” and everyone in Silicon Valley certainly envies his company right now.
Microsoft and Meta bought 150,000 Nvidia H100s and H800s this year, which cost about $US30,000 each. Google, Amazon, and Oracle bought as many as they could get their hands on, roughly 50,000 each. Joe Biden deliberately set United States foreign policy to limit Nvidia chips getting into the hands of Chinese companies. 2023 was the year powerful people learned Jensen Huang’s name.
Big Tech executives lined up to shake hands with Jensen on stage and smile for photos, but there was tension in the air. Soon after Jensen Huang walked off stage in his iconic black leather jacket, his largest customers quickly announced their own competing chips to train AI models. Big Tech doesn’t want to be beholden to anyone. Being the biggest star in the Valley may mean you’re friends with everyone for now, but it also puts a giant target on your back.
History of GPUs
Jensen Huang’s position at the center of the AI revolution started long ago. The Nvidia CEO is worth roughly $US43 billion, but his preferred place to take meetings is a booth at Denny’s. He made a bet at one of those booths over 10 years ago that GPUs were the future of gaming, instead of CPUs.
A GPU can perform many tasks at once, whereas a CPU performs tasks one at a time. The Mythbusters demonstration from 2009 is still the best, comparing a GPU to 1,100 simultaneously firing paint guns to recreate the Mona Lisa in a fraction of a second. That’s how a GPU works, performing parallel tasks, whereas a CPU is more like a single paint gun, trying to paint a smiley face, creating something both less impressive and taking more time.
GPUs revolutionized the gaming industry, making video game graphics look way better than anything produced with CPUs. It turns out, through incredible luck or foresight from Jensen, that GPUs would prove to be essential for training artificial intelligence which, like graphics rendering, requires billions of simultaneous tasks to build large neural networks. GPUs have become an essential building block for any company working artificial intelligence, and every other chipmaker is years behind Nvidia.
Competitor or Customer?
Everyone made headlines for jumping into AI chip production this year. Microsoft has the Maia 100, Google has the TPU v5p, Amazon has its Trainium 2, and Meta announced its MTIA. They all sound great, but the truth is none of them hold a candle to Nvidia’s GPUs. Nvidia is years ahead of the competition, and all of these companies will be training most of their AI models on Jensen Huang’s chips for the foreseeable future.
“The initiatives that we see for vertical integration are very, very muted,” Tristan Gerra, Baird’s Senior Research Analyst covering semiconductors, told Gizmodo. “Meta is not going to have their own chip until 2025. AWS has talked about developing its own GPU, which we know is extremely complex and challenging, and that would take several years.”
Gerra says we are going to see in-house efforts to build AI chips for the next few years, but they’ll all be very limited. Nvidia will likely only lose a couple of percentage points of market share in the next year.
“For Nvidia, there’s really nothing compared to the high double-digit growth we see for them, and the whole space, for the next several years,” Gerra said.
So long term, Nvidia may have to worry about Big Tech as a chip competitor, but for now, they’re just customers. Big Tech can tout their AI chips at conferences all they want, and Jensen Huang won’t lose a lick of sleep. By the time they do catch up, Nvidia has even bigger plans for computing domination.
Jensen Huang’s Vision to Turn Leverage Into an Empire
Nvidia is trading its GPUs to cloud providers like Google, Microsoft, and AWS who are, in exchange, hosting Nvidia’s own AI cloud service, DGX Cloud. So just as these companies come for Nvidia’s AI chips, Jensen Huang is also building out cloud products to replace them. But who will win?
Big Tech’s best friend, and possibly biggest enemy, told The New York Times at its DealBook Summit that AI has completely changed computing. Huang is hoping he can use his position to reign in a new era of computing.
“We’re at the beginning of a brand new generation of computing. It hasn’t been reinvented in 60 years, this is why it’s such a big deal,” said Huang. He noted that currently, computing is largely about retrieval—you simply ask your phone to retrieve a file from a server somewhere. In the future, he says computing will involve retrieval as well as generation powered by AI.
The 60-year-old computing revolution he’s referencing was led by Intel, the chip company that perfected mass-market CPU chips founded by Gordon Moore and Bob Noyce in 1958. Intel drastically increased humanity’s computing power by putting many small transistors on silicon computer chips. Jensen Huang sees Nvidia and GPUs at the center of the next revolution; the next Intel.
“You can’t solve this new way of doing computing by just designing a chip,” said Huang. “Everything from the networking to the switching, to the way the computers are designed, to the chips themselves, all of the software that sits on top of it, and the methodology that pulls it all together. It’s a big deal because it’s a complete reinvention of the computer industry.”
The shift Jensen is talking about lines up with what Sam Altman describes as the future of technology. At OpenAI’s keynote address, Altman describes a future where humans ask computers to do a task for you, instead of doing the task on the computer. They’re describing AI generation as not just a feature, but the core operating system of computers moving forward.
Huang sees himself as one of the first to recognize this new future, that involves new data centers, new computer design, and new coding languages to power it. Huang and Nvidia are betting the house on this vision, and hoping to be the technology company that builds the future of computing. In 5-10 years, Big Tech may have caught up to where GPUs are today, but that’s just the beginning according to Jensen Huang.
“It’s hard for people to wrap their heads around it,” said Huang. “But that was the great observation that we made. It includes a chip, but it’s not about that chip.”