At the prestigious Computex expo in Taiwan, Intel’s CEO Pat Gelsinger delivered a spirited defense of his company’s future in the AI era, directly countering claims made by Nvidia’s CEO Jensen Huang. Gelsinger asserted that Intel remains at the forefront of innovation and will play a pivotal role in the proliferation of AI, despite recent setbacks.
In his keynote, Gelsinger unveiled Intel’s latest offering, the Xeon 6 data-centre processors, which boast more efficient cores. These improvements, he noted, will enable operators to reduce the space required for tasks to a third of what was previously needed.
Gelsinger’s announcement comes at a critical time as Intel strives to reclaim market share lost to competitors like Nvidia, AMD, and Qualcomm.
“Unlike what Jensen would have you believe, Moore’s Law is alive and well,” Gelsinger declared, referencing the famous observation that the number of transistors on a microchip doubles approximately every two years, leading to continuous improvement in performance. He emphasized that Intel is poised to be a key player in the AI revolution. ”
I think of it like the internet 25 years ago, it’s that big,” he said, predicting that the semiconductor industry will hit $1 trillion by the end of the decade.
A highlight of Gelsinger’s presentation was Intel’s Gaudi systems, which package its processors into kits designed for generative AI training.
Partnered with Dell Technologies and Inventec, Intel will offer these kits at competitive prices—$65,000 for a kit with eight Gaudi 2 accelerators and $125,000 for a more powerful version with eight Gaudi 3 accelerators. Intel claims these kits are more affordable than those of competitors and offer superior performance.
Each Gaudi 3 cluster comprises 8,192 accelerators, with Intel estimating it can train AI models up to 40% faster than Nvidia’s H100 GPUs.
Moreover, Intel asserts that the Gaudi 3 is twice as fast as Nvidia’s H100 in executing AI inferencing tasks, particularly in popular models developed by companies like Meta Platforms and Mistral.
Despite these bold claims, industry analysts remain cautious about Intel’s ability to overtake Nvidia in the data-centre AI processing domain.
Leonard Lee, an analyst at neXt Curve, noted that “the performance of each individual accelerator is no longer the most important thing.” Nvidia’s integrated ecosystem and proprietary technology like NVLink provide it a substantial advantage, enabling massive, cohesive computing clusters.
Gelsinger’s confident presentation underscored his commitment to reversing Intel’s recent revenue decline. The company’s earnings have fallen sharply over the past two years, with projections indicating it will end 2024 with $20 billion less in revenue than it had in 2021.
Meanwhile, competitors Nvidia and AMD are expected to see significant growth, capitalizing on the booming AI market.
Gelsinger, who returned to Intel three years ago to spearhead its turnaround, has been investing heavily in new technologies and expanding Intel’s manufacturing capabilities.
“This is the most consequential time of our careers together,” he told the audience, stressing the importance of collaboration with partners to achieve Intel’s ambitious goals. “We were made for this moment.”