DeepSeek’s newly launched open-source AI reasoning model, R1, has created ripples across the tech industry, triggering a sell-off of Nvidia stock and catapulting its consumer app to the top of app store rankings.
The announcement follows DeepSeek’s disclosure that the R1 model was trained using a data center powered by approximately 2,000 Nvidia H800 GPUs over the course of just two months, at a remarkably low cost of $5.5 million. Last week, DeepSeek published a detailed paper confirming that R1’s performance rivals some of the most advanced reasoning models globally. This breakthrough is particularly striking as these high-performing models are typically trained in data centers investing billions in Nvidia’s cutting-edge AI chips.
Industry Reactions and Leadership Insights
The tech industry has responded with both excitement and skepticism to DeepSeek’s innovative approach, which combines high performance with affordability. Among the enthusiasts is Pat Gelsinger, former Intel CEO and now chairman of Gloo, a messaging and engagement platform for churches. Taking to social media platform X, Gelsinger expressed his admiration for DeepSeek, writing, “Thank you DeepSeek team.”
Gelsinger, a hardware engineer by background, emphasized three key lessons from DeepSeek’s success: the importance of lowering costs for widespread AI adoption, the ability of ingenuity to thrive under constraints, and the enduring power of open ecosystems. In his words, “DeepSeek will help reset the increasingly closed world of foundational AI model work.”
Gelsinger also revealed that his team at Gloo, which is developing an AI-powered chatbot and service platform called Kallm, has already started leveraging R1. The open-source nature of the model has allowed Gloo’s engineers to forgo dependency on closed-source alternatives like OpenAI. He remarked, “In two weeks, we’ll have rebuilt Kallm from scratch with our own foundational model that’s all open source. That’s exciting.”
Rethinking AI Accessibility and Performance
According to Gelsinger, DeepSeek’s work signals a shift toward more affordable and accessible AI, where advanced technology is integrated into everyday devices. “I want better AI in my Oura Ring, my hearing aid, my phone, my EV’s voice recognition. DeepSeek makes that possible,” he explained.
However, not everyone in the industry is as optimistic. Critics have questioned the feasibility of DeepSeek’s cost estimates and raised concerns about transparency, particularly given the company’s operations in China. Some suggest the model may have benefited from undisclosed high-end resources or other advantages. Others argue that upcoming models like OpenAI’s o3 could outclass R1, maintaining the current dominance of established players.
Gelsinger dismissed these doubts, emphasizing the broader implications of DeepSeek’s success. “It’s clear that R1 has achieved 10-50x lower training costs than OpenAI’s o1 model. That’s a testament to engineering ingenuity rather than brute computational force,” he said.
The Open-Source Revolution with a Global Twist
While the open-source nature of DeepSeek’s work has been celebrated, its origins in China have sparked debates over privacy, censorship, and geopolitical implications. Gelsinger acknowledged these concerns but remained optimistic, stating, “Having the Chinese remind us of the power of open ecosystems is maybe a touch embarrassing for our community, for the Western world, but it’s a necessary wake-up call.”
DeepSeek’s R1 model represents not only a technological leap but also a challenge to the status quo in the AI industry. As debates rage over transparency and global collaboration, one thing is certain: the future of AI just got a lot more competitive—and accessible.
