"The real story of who controls AI isn't unfolding in that courtroom, nor is it limited to those two men. It's taking shape in Abu Dhabi's meeting rooms, in the discreet offices of a fund in Hangzhou, and in the data centers rising in the Texas desert. There are many more players, but these nine individuals are quietly deciding far from press conferences and socialmedia fights how the technology that will change everything is built, financed, and governed."
"Jensen Huang, 62, born in Tainan, Taiwan, co-founded Nvidia in 1993 in a San Jose Denny's with $40,000 in capital. In 1996, the company was a month away from bankruptcy. He bet everything on an unproven chip the RIVA 128 and saved the business. Since then, he has turned that obsession with survival into a corporate philosophy: he tells his employees that the company is always 30 days away from going under."
"What no one anticipated was that chips designed for video games would be the perfect infrastructure for training artificial intelligence models. By the time the world realized this, Nvidia was worth $3 billion, and Huang had tattooed the company's logo on his left shoulder. Today, without its CUDA architecture, which is two decades ahead of any competitor, there would be no OpenAI, no Google DeepMind, no Anthropic, no DeepSeek."
AI control is being determined by influential people operating outside public courtrooms and viral headlines. Key decisions about how AI is built, financed, and governed are occurring in places such as Abu Dhabi offices, Hangzhou fund rooms, and data centers in Texas. Nvidia’s Jensen Huang exemplifies infrastructure power without direct involvement in AI products. Nvidia was nearly bankrupt in 1996, but Huang bet on an unproven chip, then developed a survival-focused corporate philosophy. Game-focused chips became the core infrastructure for training AI models. Nvidia’s CUDA architecture enabled major AI labs and systems, making many prominent AI organizations dependent on its silicon.
Read at english.elpais.com
Unable to calculate read time
Collection
[
|
...
]