Intel would be wealthy now if that interfering corporation had not existed!
After introducing the company’s next-generation CPUs for both clients and servers, Intel CEO Pat Gelsinger has been in the news lately. During a recent press tour, he spoke with MIT engineering school students about the state of the semiconductor industry and made headlines when he said that Nvidia, not Intel, is currently leading the AI space because the company got lucky.
In response to a professor who inquired about Intel’s AI hardware efforts, Gelsinger summarized the company’s disastrous history with GPUs and “throughput computing” (as opposed to scalar). He pointed out that when Intel fired him 11 years prior, it also canceled its discrete GPU project, which was called Larrabee. Gelsinger claimed that had the project continued, Intel would currently be leading the AI industry, but Nvidia is leading, which he attributes to Nvidia CEO Jensen Huang’s “extraordinarily lucky.”
Gelsinger points out that while Nvidia gradually focused on “throughput computing,” Intel did nothing for fifteen years. He claims that Nvidia was always focused on graphics, which is true since it was always a gaming company that eventually branched out into AI and data centers. Gelsinger also notes that the company did not even want to support its first big AI project because you needed a supercomputer to run the big data sets they were using. When Gelsinger returned to the company in 2021, he immediately corrected this error by restarting the project, which is now known as its Arc graphics line of GPUs.
The site notes that while it seems true that Nvidia probably did not predict the current AI craze, it has relentlessly pushed the development of its platform forward and has arrived at just the right time with what most consider the industry’s best products, including hardware and its CUDA software. Is that blind luck, or just an educated bet? As PC Gamer points out, Gelsinger is not exactly wrong here. Plus, his comments seem to be more of a dig at Intel’s shortsightedness than anything about Nvidia.
Like everything else in the computing industry, companies are always making predictions about the future. Sometimes these predictions come true, and other times they blow up in your face. Intel has experienced more than its fair share of misses, but Gelsinger hopes to “course correct” with the entire company’s current offering. In terms of AI, Gelsinger recently announced that the Gaudi 3 AI accelerator will be competitive with Nvidia’s current H100 and AMD’s MI300 in 2024.