Logo Logo

It only takes one fool - NVIDIA


There he is, dancing, alone. Such a fool. Everybody is looking at him. Following the rhythm of the music, the fool enjoys himself… Lunatic.

She looks at him. Maybe he is not that crazy. And she likes the music. Maybe she joins… Will she? She stops for a moment and takes a chance. And there are two: the fool and the follower.

And the others that enjoy the music start thinking: shall I join?

In a matter of moments, it is not crazy anymore, it is normal. The fool is the leader, there is a movement. And everybody dances.

NVIDIA was the fool. Not once, but three times. The CEO, Jensen Huang, has bet the company in gaming, crypto, and AI. NVIDIA faced a couple of near death experiences. But then it became the leader. When the AI market started to bloom, NVIDIA had a complete end-to-end solution ready for scaling, and in 2022, they captured 80% of the AI chip market. Jolted by ChatGPT, AI is expected to hit $200-$300B by 2030 and NVIDIA will keep on riding the wave while others stare from the shore. When should we expect another big bet?


NVIDIA was funded in 1993 by three engineers who realized that video games could benefit from faster calculations done in parallel outside the CPU. Despite the promising vision, the company entered a competitive low-margin undifferentiated market with a premium and complex product. Not facing enough demand, NVIDIA fired all but 35 of its staff just two years later.

To turn around, the company simplified its product, and invested in prototyping and testing, being able to deliver a new graphics chip at unprecedented rates - every 6 to 8 months. It also decided to outsource manufacturing (a low margin business) to TSMC (a taiwanese semiconductor manufacturer), focusing solely on design and marketing of its GPUs.

By 1999, NVIDIA went public. The stock price bumped 64% on the offering date, but its small revenues of $160M still made the company a small player in the semiconductor business. While revenues rose until 2005, reaching $2.4B in 2005, they dropped back to 7-8% a year in the following decade.

In 2006, NVIDIA decided to be the fool again. It launched the GeForce G80 with a new architecture - a higher number of simpler cores that could run in parallel at double clock-speed, able to perform any calculation required at any moment. While an interesting product, there was no usage for it in gaming and the increase in cost was substantial.

But NVIDIA was not targeting gaming. Instead, they aimed for scientists and programmers, universities, AI. The company sponsored conferences and workshops, promoting their new CUDA parallel programming model next to researchers at top schools.

The first follower came in 2012. Alex Krizhevsky, PhD student at the University of Toronto, who along Ilya Sutskever and Geoffrey Hinton used the graphics card to train AlexNet. It would win the ImageNet competition (a key competition in AI research), performing 41% better than competition. The moment bumped NVIDIA into AI leadership, and AlexNet was bought by Google.

So when big tech firms were ready to invest heavily into AI, NVIDIA had a multi-year lead and an end-to-end platform ready to scale. In 2017, their data centre business had a total annual revenue of $830M. In 2022, this number was $15B.


NVIDIA is clearly a fool. First with semiconductors, then with crypto, and now with AI. The company has been able to pinpoint and invest ahead of everyone in growth markets for the chip business. NVIDIA opportunistically bets on markets by design, not chance. Today, NVIDIA is on the trillion dollar group, and is the eighth largest company in the world by market cap. What would be the next market? Will the company foresee the next big thing, yet again?





Note: The company had near-death experiences, at least in market value term, in 2002 and 2008, losing more than 80% of its market value. In 2018, as the crypto craze passed and miners were reselling their chips, the company lost again more than 50% of its market value.


Catarina Pinto @catarinappinto