DeepSeek’s new open-source AI reasoning model, R1, has sent shockwaves through the tech industry, triggering a sell-off in NVIDIA stock and catapulting its consumer app to the top of app store rankings.
Last month, DeepSeek revealed that it trained R1 using just 2,000 NVIDIA H800 GPUs over two months at a cost of $5.5 million — a fraction of what competitors spend on high-performance AI models. Last week, it published a paper showing R1 matches the performance of the world’s most advanced AI reasoning models, many of which rely on billions of dollars in AI chip investments.
Tech leaders reacted swiftly. Former Intel CEO Pat Gelsinger praised DeepSeek’s efficiency, calling it a reminder of computing’s core principles — lower costs drive wider adoption, constraints fuel innovation, and open-source models will challenge the increasingly closed AI ecosystem. Gelsinger revealed that his new startup, Gloo, is already shifting to R1 instead of paying OpenAI for access to its models.
While skeptics question DeepSeek’s reported costs and potential reliance on restricted AI chips, Gelsinger dismissed concerns, emphasizing that engineering creativity, not hardware brute force, is driving AI’s next leap forward. As OpenAI prepares to release its next-gen o3 model, the industry braces for a new wave of competition in AI affordability and accessibility.
