Archive — US — turtlesai
Former Intel CEO Pat Gelsinger Claims Nvidia is too expensive for AI inference -dropping Larrabee was a Mistake
Former Intel CEO Pat Gelsinger criticized Nvidia's AI GPU pricing as excessively high, claiming they are 10,000 times too expensive for inference tasks. During the GTC 2025 event, he attributed Nvidia CEO Huang's success to luck and noted Intel's shift from its 'Falcon Shores' line to 'Jaguar Shores'. Gelsinger believes Intel still has an opportunity, especially with emerging technologies like quantum computing.
Key points
- Pat Gelsinger criticized Nvidia's pricing for AI GPUs, saying they are excessively high for inference workloads.
- He claimed Nvidia's GPUs are 10,000 times more expensive than necessary.
- Gelsinger attributed Nvidia CEO Jensen Huang's success to luck in the AI sector.
- Intel has shifted its focus from the 'Falcon Shores' AI line to the 'Jaguar Shores' line.
- Gelsinger expressed skepticism about Nvidia's dominance and suggested the future lies in optimized inference solutions.
- Intel reported lower revenues and has struggled to keep up with competitors in the AI market.
Why it matters
- Nvidia's CUDA platform is seen as a competitive advantage, but Gelsinger's comments suggest a potential shift toward other AI frameworks.
- The trends in AI and quantum computing are significant as industries look for affordable and efficient technology solutions.
- Intel's struggles reflect broader challenges facing traditional tech giants in adapting to rapid advancements in AI.
Gelsinger’s criticism highlights the evolving landscape of AI technology and pricing. As Intel pivots strategy and explores quantum computing, it remains unclear if they can reclaim a strong position in the market dominated by Nvidia.