Former Intel CEO Pat Gelsinger, now leading his startup Gloo, has publicly endorsed DeepSeek’s new AI model, R1, over OpenAI.
DeepSeek recently unveiled R1, a high-performance reasoning AI trained on 2,000 Nvidia H800 GPUs in just two months at an impressively low cost of $5.5 million.
This open-source model rivals expensive closed systems like OpenAI and has sparked debate in the tech industry. Gelsinger lauded the model for making AI more affordable and accessible, allowing widespread adoption in devices like wearables, phones, and electric vehicles.
At Gloo, a messaging platform targeting churches, engineers are integrating R1 into their AI-powered service, Kallm, bypassing OpenAI due to R1’s openness and cost-effectiveness. Gelsinger believes this innovation proves that creativity in engineering can advance AI without relying solely on costly hardware.
However, critics question DeepSeek’s cost claims and its use of U.S.-restricted AI chips, raising concerns about transparency and the implications of working with a Chinese developer.
Despite skepticism, Gelsinger emphasized the benefits of open ecosystems, noting that DeepSeek’s approach challenges the increasing exclusivity of foundational AI models.
He dismissed concerns over privacy and data handling, pointing out that the West could learn from the success of open-source innovation. For Gelsinger, DeepSeek’s model represents a pivotal moment, showing that AI advancements can thrive under constraints and disrupt the status quo with affordable, high-quality solutions.
