Nvidia Engineer Responds to Intel CEO Saying It Got Lucky With AI Dominance

The employee, who worked at both Intel and Nvidia, says vision and execution are the reasons for its supremacy, not luck.

A few days ago, Intel CEO Pat Gelsinger raised a few eyebrows by attributing Nvidia's current dominance of the AI market to luck. He argued that if Intel hadn't killed its supercomputing GPU project in the 2000s, it would be where Nvidia is now. When Intel ditched the project, it also jettisoned Gelsinger, so he might be a bit salty about how things turned out. An engineer who worked on Intel's defunct "Larrabee" project but now works at Nvidia has responded to Gelsinger, saying it wasn't luck but Nvidia's vision and execution, which Intel lacks. Ouch.

The employee doing the smack talk is Bryan Catanzaro, whose Twitter bio says he is VP of Applied Deep Learning Research at Nvidia. We've seen him before in videos talking about the company's Deep Learning Super Sampling technology, or DLSS. He notes in his tweet that he was one of the engineers working on Larabee in 2007, and his LinkedIn profile shows he was an intern at Intel back then.

We previously wrote that Intel claimed Larrabee could do ray tracing in modern games at acceptable frame rates, so it was ahead of its time. Also, it seems there's some notable crossover between Intel's hopes for Larrabee in 2007 and what Nvidia produced with Turing in 2018.

Back then, there was an actual skirmish between Nvidia and Intel on the HPC processing front, with Nvidia promoting its Tesla cards and Intel promoting its Knights Ferry GPGPUs. We wrote then, "Having seen lackluster adoption of Nvidia's Tesla platform in HPC applications, Intel hopes that this level of cross-compatibility will ensure Intel's continued dominance in supercomputing."

Things did not turn out in Intel's favor, as Intel essentially killed the Larrabee project—choosing instead to go with its x86 processors for HPC tasks and focusing its graphics efforts on integrated GPUs. When Intel officially announced Larrabee in 2010, AnandTech wrote that Intel's decision "completely validates Nvidia’s Tesla strategy."

Catanzaro notes in his tweet that Intel was 10 times the size of Nvidia back then. Therefore, it figured it would crush Nvidia with Larrabee. Still, it lacked the vision and execution to see the project through to what Nvidia is doing now with its advantages in parallel computing. A quick Google search backs up these numbers, as in 2007, Nvidia's annual revenue was roughly $3 billion, while Intel's was $38 billion.

Fast forward to Q3 of 2023, and the tables have turned. Nvidia's earned $18 billion for the quarter, a 206% increase from a year ago. Meanwhile, Intel earned $14 billion, a decrease of 8% year-over-year. More important was where the companies made those dollars, with Nvidia pulling in $14.5 billion from its data center products, while Intel earned just $3.8 billion.

It remains to be seen if this war of words will continue. We doubt it will go on much longer as it doesn't make Gelsinger look particularly charitable toward one of Intel's partners. He noted in his previous comments that he's good friends with Nvidia's CEO and that they talk often, making us wonder if this topic has ever come up in their discussions.

Source