Topic: Deepseek

American companies are increasingly adopting Chinese AI models, such as DeepSeek's V4, because they are significantly cheaper than U.S. "frontier systems" like those from Anthropic and OpenAI. The hosts explain the concept of "token maxing" and how companies are exceeding their budgets due to the high cost of U.S. compute units. They argue that the "moat" around major AI labs is disappearing as cheaper, capable alternatives emerge.

Sam Altman discusses the rapid revenue growth of OpenAI and the eventual path to profitability, despite concerns over chip supply chains and competition from models like DeepSeek. The hosts label the current AI boom as the "longest demo in the history of Silicon Valley" and a potential hoax that could lead to a market collapse.

OpenAI is offering its Enterprise product to federal agencies for just one dollar, a move interpreted as a sign of a sales struggle. The company is also releasing "open weights" for some models to compete with Meta's Llama and China's DeepSeek. This shift suggests that the high cost of cloud operations is forcing AI companies to let users run models on their own hardware.

Treasury Secretary nominee Scott Bessent advised trading partners not to panic or retaliate, suggesting that current tariff levels represent a "ceiling" for negotiations. He noted that China faces higher rates due to the ongoing fentanyl crisis and the export of precursor chemicals. Bessent also dismissed recent Nasdaq declines as a "DeepSeek problem" related to specific AI developments rather than a broader failure of the administration's economic agenda.

Industry insights from a data center developer suggest a significant downturn in the AI infrastructure market, with companies like Microsoft reportedly canceling contracts. The emergence of the Chinese DeepSeek model has shifted expectations toward cheaper training methods, moving the industry focus from remote training centers to low-latency "inference" hubs. Many struggling data centers are being repurposed for Bitcoin mining, while major firms like KKR and BlackRock have already secured exits from these investments.

Testing of the Chinese AI model DeepSeek revealed significant "hallucinations," including a claim that Adam Sandler invented podcasting. When corrected, the AI's "reasoning" mode falsely attributed the invention to the BBC and NPR in the 1920s. Furthermore, the AI refused to quote Bible scriptures due to "cultural sensitivity," highlighting the ideological constraints programmed into the model by its developers.

OpenAI is reportedly losing money on its premium subscriptions, including the $200-a-month plan, due to the extreme inefficiency and compute costs of its models. Sam Altman has admitted the company needs more capital than initially imagined. The discussion notes that the majority of OpenAI's revenue comes from subscriptions rather than enterprise use of the models.

The hosts discuss the history of economic depressions in Silicon Valley, referencing the 1986 crash triggered by the collapse of the video game industry. They analyze the new trend of "agentic AI," which they dismiss as glorified "cron jobs"—automated tasks scheduled to run at specific times.

The release of DeepSeek, a Chinese AI model developed for under $6 million, caused a global slump in tech stocks, particularly affecting NVIDIA. Analysts suggest the low cost is misleading as it doesn't account for previous development iterations. The model's success is viewed as a failure of the Biden administration's export controls, as the Chinese firm utilized older or downgraded NVIDIA chips to train a competitive model.

The DeepSeek AI model is characterized as a "derivative scam" that likely uses model distillation to scrape data from competitors like OpenAI. Despite skepticism about its originality, the availability of open-source models that can run locally without internet access is noted as a significant development. A practical use case is described involving using AI to find specific Bible passages, such as those regarding respect for the elderly.

Chinese AI players DeepSeek and ByteDance have released models that rival OpenAI's performance at a fraction of the development cost. DeepSeek reportedly built its model for less than $6 million, challenging the American assumption that billions of dollars and massive GPU clusters are required for frontier AI. The hosts predict a market pivot toward "quantum" as the current AI bubble faces cost pressures.

China's DeepSeek releases an open-source AI model that reportedly matches OpenAI's performance at a fraction of the training cost. Meanwhile, Larry Ellison discusses the future of AI in healthcare, specifically the ability to design individual mRNA cancer vaccines within 48 hours. Critics argue these developments are therapies rather than traditional vaccines and question the role of government funding.