Intel leads alliance that seeks to wrest dominance in artificial intelligence market from NVIDIA

Intel’s crusade continues. NVIDIA currently holds about 80% of the market for GPUs designed specifically for artificial intelligence (AI) applications, and has achieved such positive momentum that seems unstoppable. In fact, last week it overtook Apple to become the second most valuable company in the world after Microsoft. At this stage we have good reason to admit that the natural rivalry that existed between the two companies is a thing of the past. in a crescendo Over the past months.

Last week, during a speech at Computex, NVIDIA CEO Jensen Huang said that Intel processors are falling out of favor in the age of artificial intelligence. In this statement, he was referring to conventional processors, not GPUs, designed by Intel specifically for solving artificial intelligence problems. The response from Pat Gelsinger, CEO of Intel, was not long in coming. And what’s more, he did it with force.

“Contrary to what Jensen would have you believe, Moore’s Law is alive and well (…) I think of it the same way I saw the Internet 25 years ago. He’s so big. For us, AI hardware is fuel. “This will drive the semiconductor industry until it reaches a trillion-dollar market value by the end of this decade,” Gelsinger said.

Google, Qualcomm and OpenAI support Intel in its crusade against CUDA

During the Intel AI Summit 2024, which took place last week in Seoul, South Korea, Pat Gelsinger and other Intel executives worked to create alliances that would help them compete with NVIDIA and gradually undermine its dominance in the AI ​​market. Interestingly, the weakest link that could weaken Jensen Huang’s company in the medium to long term is CUDA (Compute unified device architecture), and Gelsinger is very clear about this.

In fact, CUDA technology plays an important role in NVIDIA’s business. Most artificial intelligence projects currently in development implemented on it. This technology combines the compiler and development tools used by programmers to develop their software for NVIDIA GPUs, and replacing it with another option in projects already underway is a challenge.

Huawei, which is aiming to capture a significant portion of this market in China, has CANN (Computational architecture for neural networks), which is an alternative to CUDA, but there are other options. “The entire industry is determined to eliminate CUDA from the market (…) We see this as a shallow and small moat, so we are motivated to offer a broader set of technologies, be it learning, innovation or data science.” Pat Gelsinger made this statement during the AI ​​Everywhere event, which took place in New York in mid-December.

Then this executive assured that Google and OpenAI are two companies with a large share in the artificial intelligence industry that want to leave CUDA behind. Some media defend this Qualcomm also joined to this initiative by Intel, although, curiously, AMD does not seem to have made a firm decision at the moment. However, it would be surprising if the company led by Lisa Su did not support the development of an open standard. It probably won’t take us long to find out if Intel has reached any firm agreements in this area during AI Summit 2024. We’ll wait.

Additional information | DigiTimes Asia

In Hatak | The AI ​​chip market is set to grow 250% in the coming years, and one company is rubbing its hands: TSMC