If the Larrabee project had not been cancelled
Intel has decided to verbally attack Nvidia again. Recently, Intel CEO Pat Gelsinger said that the entire industry wanted to eliminate CUDA, and now the same Gelsinger said that all the current achievements of Nvidia — it's just luck.
Jensen [the head of Nvidia] worked very hard to increase computing power, primarily for graphics, and then he got extraordinarily lucky. They didn't even want to support their first AI project
Gelsinger also said that now the market could be completely different if Intel had not closed the Larrabee project at one time.
When I got kicked out of Intel 13 years ago, they shut down a project that could change the shape of AI
Can we say that Nvidia was just lucky? To some extent this is true, because until recently there were no mentions of artificial intelligence and GPU capabilities in this area from Nvidia at all. On the other hand, it seems that 10 years ago no one was thinking in this direction.
It can also be noted that Nvidia was supported by its current vice president for research in the field of applied deep learning, Bryan Catanzaro, who at one time worked at Intel, including on Larrabee applications.
In 2007, I worked at Intel on Larrabee applications. Then in 2008 I moved to Nvidia to work on ML. So, I was there at the time in both places and I can say: Nvidia's dominance was not a result of luck. It came from visioning the future and executing. What Intel was missing