Intel’s interim CEO doesn’t believe Falcon Shores AI accelerators will be successful, but this is “the first step in the right direction”

Intel management does not believe that the company will soon be able to compete with Nvidia and AMD in the field of AI accelerators. That’s the impression, at least, following recent comments from one of the company’s interim chief executives, Michelle Johnston Holthaus, at the 22nd annual Barclays Global Technology Conference.

Image source: Intel

Let us remind you that Intel is developing Falcon Shores computing accelerators, which will be based on a graphics processor tailored for high-performance computing and data center tasks, and will complement the GPU with elements of current Gaudi AI accelerators. The project to develop this solution received an unexpected assessment from Holthaus: “We really need to think about how to move from Gaudi to our first generation of Falcon Shores GPUs. Will the new product be amazing? No, it won’t. But it will be the first step in the right direction.”

Holthaus reiterated Intel’s new pragmatic approach to developing hardware solutions to accelerate AI when she discussed product strategy: “If you drop everything and start creating a new product, it will take a very long time to develop it. It will take two to three years before anything appears. Instead, I would prefer to create something in smaller quantities, learn something new, and consistently improve in order to ultimately achieve my goals.”

Intel’s interim CEO acknowledged the enduring nature of the AI ​​market opportunity, highlighting the industry’s current interest in training AI models. However, Holthaus also highlighted the potential for greater opportunities in other areas: “It is clear that AI is here to stay. Obviously [AI] training is the focus today, but there are opportunities in other areas where there are also needs in terms of new hardware.” Apparently, she meant inference – the launch of already trained neural networks.

From the above, we can conclude that Falcon Shores will not become a miraculous lifeline for Intel, which will allow it to catch up with Nvidia in the GPU accelerator market. This is largely the first step towards developing a first-class product in the future. Intel’s next project after Falcon Shores should be Jaguar Shores. Its release is expected in late 2025 or early 2026 in the form of AI and HPC accelerators for data centers. However, before its appearance, the company has a lot of work to do to improve not only its hardware, but also its software.

Nvidia’s dominant position in the AI ​​market owes much to its CUDA hardware architecture, as competitors such as AMD offer comparable hardware performance. Intel faces a very difficult task. It will have to ensure ecosystem software development and seamless integration of its next-generation accelerators to ensure Jaguar Shores has a chance of catching up with the rest of the market.