Intel CEO Says Entire Industry Is Motivated To Eliminate NVIDIA’s CUDA Dominance, Also Open To Manufacturing Competitors Chips

Intel's CEO, Pat Gelsinger, has called out NVIDIA's CUDA being a shallow moat and that the entire industry is motivated to end its dominance in the AI market.

At the "AI Everywhere" event, which brought the unveiling of Chipzilla's new Intel Core Ultra and Intel 5th Gen Xeon "Emerald Rapids" CPUs, Intel's CEO took the stage and made fairly controversial remarks on NVIDIA's humongous progress in the AI markets, saying that the industry is all set to "eliminate" the CUDA market, through the adoption of new training methods, and getting ahead in the "inference market".

You know, the entire industry is motivated to eliminate the CUDA market. We think of the CUDA moat as shallow and small. Because the industry is motivated to bring a broader set of technologies for broad training, innovation, data science, et cetera

Pat Gelsinger (Intel CEO) via Tom's Hardware

Intel believes that the future of AI lies in "inference" rather than in training models since that is where Intel lags behind a lot compared to its competitors. The statement by Gelsinger depicts that the firm sees NVIDIA's current success as a mere "bubble" that is destined to burst at any time.

The company's approach towards AI is framed in a way that they are prioritizing inference developments over model training mainly because it is much more resource-efficient and can adapt to rapidly changing data without the need to retrain a model. Intel took the opportunity to praise its OpenVINO model as well, claiming that it would be vital to transition toward next-gen markets.

As inferencing occurs, hey, once you've trained the model… There is no CUDA dependency. It's all about, can you run that model well? ....... fundamentally, the inference market is where the game will be at.

Pat Gelsinger (Intel CEO) via Tom's Hardware

Intel hasn't yet shown its true colors in the AI segment, or at least hasn't displayed the firm's utmost capabilities based on what they idolize. In the current day and age, NVIDIA is dominating the AI segment not because it has high computing power on its back, but because the CUDA platform has been framed in a way to leverage the performance of the company's AI accelerators. A prime example of this is how just recently, Team Green made a comparison with AMD's recently unveiled MI300X, claiming to have a much superior performance just because of its software stack around the CUDA ecosystem.

Intel's Gaudi accelerators have been a popular alternative to NVIDIA's AI solutions, offering very competitive perf/$ and the company teased its next-gen Gaudi 3 accelerator at the event which is aiming for a 2024 launch.

It is expected that NVIDIA, Intel, and AMD will continue to evolve their software ecosystems such as OpenVINO and ROCm, offering optimized performance and more features in the AI field. But for now, CUDA is the market standard and it will be interesting to see if that changes in the future.

Meanwhile, Intel's CEO also stated that while AMD and NVIDIA are their competitors, they would also love to have chips from both companies produced at its fabs. NVIDIA also hinted at using a third fab partner (Intel) to meet its supply needs if necessary.

We are traditionally in this market together but in other ways, very different. Our IDM 2.0 strategy is a crucial piece of why we differ. We manufacture at scale, we manufacture our chips but increasingly as you know we are becoming a manufacturer for everyones chips as well. And then you have the fabless model and that's companies like Qualcomm, NVIDIA, AMD and I have many products that I build as well so essentially, I am building two companies inside one, a major Foundry manufacturer at scale and a fabless company.

So in some ways, I compete and in some ways, I want to manufacture all of NVIDIA's chips, all of AMD's chips as well as all of Intel's chips going forward and we are rebuilding the world's supply chains for the most important resource of the future.

Pat Gelsinger (Intel CEO) via Yahoo Finance

Intel needs to do a lot more work if they are determined to dethrone CUDA's dominance in the AI markets, and even if they believe inference is the way to go, Team Blue needs to act quickly. It will be interesting to see how the markets evolve moving forward, but in its current state, NVIDIA will be leading the bandwagon going into 2024.

News Source: Tom's Hardware