A season for everything under the sun
Jensen Huang resigned today from Nvidia, the company he spent more than 30 years building and growing, and for a period of time the most valuable company in the world.
He was rocked by sex and espionage scandals, just days after the Republican party in the US had lost the midterm elections, and days before the planned IPO of Anthropic. One of his alleged lovers was a Managing Director of a Singapore-based company that was accused of smuggling GPUs to China in violation of US export controls, and another one was an employee who reported directly to him.
The allegations aside, this is a sad way to go for a man who pioneered and built the technology – GPUs – that have underpinned the ongoing AI revolution.
It is hard to imagine the AI boom arriving at this time without Jensen, and the AI era would not be the same without him at Nvidia’s helm.
For who can see what will happen next?
It was not obvious that Nvidia would grow to become the largest company in the world. Jensen started Nvidia as a maker of graphics cards, later known as Graphics Processing Units (GPUs), for use in processing graphics for video games, and later photo and video editing.
Early CPUs were relatively weak, and could not handle complex renders for video games. Hence, dedicated graphics cards were created that generated realistic graphics for games. Nvidia was a pioneer and leader in this field.
Similarly, at that point in time, sound for video games and other entertainment applications was off-loaded to dedicated sound cards. Singapore’s Creative Technology, creator of the Sound Blaster, was the leader in this field.
It turned out that graphics was much more compute intensive than sound, and while the sound card eventually got integrated into motherboards, integrated graphics never really took off for gaming applications, and standalone graphics cards continued to be needed for high end games.
I was very young back then, but I don’t think that any of us could have predicted that the fate of sound and graphics would have diverged so much, and that Nvidia would grow to become the fulcrum of potentially the biggest technology in the world while Creative would fall by the wayside.
I remember conversations about GeForce graphics cards long ago for first person shooters such as Counter Strike and Battlefield, or flight simulators, but not engaging heavily as the strategy games I favored did not require intensive graphics. So even amongst gamers, graphics cards were a niche segment.
Jensen survived all these years though, doing deals with Sega, Microsoft and TSMC. The road was not easy. Early on, Nvidia nearly went bankrupt while designing a chip for Sega, and Jensen had to fly to Tokyo to beg for release from the contract and an investment from Sega. Later, Jensen had to write a letter to Morris Chang from TSMC to build a manufacturing relationship, after his calls to TSMC were not returned. By all accounts, Jensen enjoyed his work and ran a tight ship at Nvidia, with as many as 60 executives reporting directly to him.
Jensen realized the power of GPUs in parallel processing many years ago, and steered Nvidia towards creating the CUDA architecture that was used in physics simulations, crypto mining, and machine learning research.
It was the crypto boom that led to a surge in orders for Nvidia’s GPUs to mine a variety of coins and Nvidia’s ascent to serious contender in the stock market, but it was the training of large language models powered by GPUs that propelled Nvidia to the top. Nvidia still holds a large percentage of the GPU market that powers training and inference for all the labs, closed and open source.
Perhaps we could have approached this historical moment in many other ways, but isn’t it amazing that the hardware we use to create and power intelligence came into being because of dreams of fighting on a simulated battlefield or flying aeroplanes in virtual worlds?
This was a very contingent future that came for us.
A time to play
“Papa, I don’t want to go to school! I just want to stay at home and play, play, play!”
My toddler complained, fiddling with his toy buses, toy cars, and toy aeroplanes, as I was trying my best to cajole, bribe, and threaten him to get out of the door, into my car and to childcare so I would be in time for my 9am meeting at work.
“You need to go to school so you can learn!”
I tried to explain to him, knowing that it was a partial lie. For one, his school focused on play-based learning so he would be playing with toys like magnetic tiles at school with a bunch of his friends anyway. For another, while he does learn things like phonics and counting and writing at school which we do not always have time to teach him fully at home, in my heart I do believe that he learns as well, imagining and running through various scenarios with his toy cars and buses and planes.
But I needed to get to work and I thought the socialization with kids his age was good for him, so I eventually managed to convince him to come to school with me.
—
“Why do you play video games so late at night?”
My wife asked me, irritated, as I headed to bed after another late night, squeezing in a game or two after putting my kid to sleep, chasing up on work, and doing my own reading and chores.
I had no good answer that I knew would satisfy her so I kept silent, but the joy of solving problems in an environment with no lasting stakes probably was part of it.
Human beings love to play. We play for fun. We play for excitement. And sometimes, when we play, we learn. Doing something that is of no identifiable value has often turned out to be valuable, and if anything, the GPU is a good second order proof of that.
I wondered in the past if the requisite for a future effective AI or AI civilization may be hunger or wanting, but now I wonder if the urge to play is as important to a future AI civilization.
Play is not just a hedonic trap, but a means of exploring new possibilities, pushing new frontiers, and iterating different scenarios. Something to remind myself of, next time my kid tells me he wants to “play, play, play”. And perhaps something for future AIs to wonder about as well, if they care to think about how new beginnings are generated.
There is nothing better for a person than to enjoy their work
Jensen is gone from Nvidia, and the ripples are just now spreading.
He was a key hero of the AI boom and an iconic figure in many ways – the only CEO that Nvidia had ever known, a strong supporter of open weights who pursued an acquisition of Hugging Face, a linchpin of the semiconductor supplier ecosystem in Taiwan and Korea, and perhaps the closest ally to President Trump who brought a different opinion to GPU export controls to China.
Jensen toiled for more than 30 years to build Nvidia into one of the most important companies in the world, and lost his place at its helm in a matter of days because of decisions that he made. With his departure, come key uncertainties to future compute development, open source AI, and even geopolitical contests.
What began as play is now deadly serious business.
Leave a Reply