Pat Gelsinger, former CEO of Intel, has admitted that the company underestimated the impact of artificial intelligence, a mistake that allowed competitors like NVIDIA and AMD to surge ahead in the AI chip market.

In a candid statement reported by Nikkei Asia, Gelsinger said, “I and pretty much everybody underestimated the impact of artificial intelligence.” Under his leadership, Intel prioritized AI inference—running trained models—while other companies poured resources into chips designed for training AI models, the foundation of modern generative AI systems.

Intel’s own AI accelerators, such as Gaudi, struggled to gain market traction as the industry rapidly shifted toward high-performance AI computing. Meanwhile, NVIDIA’s GPUs became the go-to solution for training large language models and other complex AI systems.

Gelsinger also pointed out that although AI chip performance has improved, energy efficiency has not kept pace, and Intel was slow to adapt to key industry trends.

This strategic misstep contributed to Intel losing ground in a market now valued in the hundreds of billions of dollars. The company also faced internal challenges, including the cancellation of the Falcon Shores project and a shift in leadership, with Lip-Bu Tan taking over as CEO.

Gelsinger continues to emphasize the importance of controlling semiconductor manufacturing, but critics argue that Intel’s delayed AI strategy and controversial IDM 2.0 foundry model have hurt the company’s ability to compete.

With new products like Jaguar Shores AI GPUs on the horizon, Intel is attempting to re-enter the AI race. Whether it can reclaim its position remains uncertain, as AI continues to reshape the global tech landscape.