PANews, January 27 news, Castle Island Ventures co-founder NicCarter posted on the X platform that Deepseek's breakthrough technology has disrupted traditional perceptions in multiple AI fields, including that China will only develop closed-source technology, Silicon Valley is the center of AI development and has a huge first-mover advantage, OpenAI's moat cannot be shaken, the development of top-notch models requires hundreds of billions of dollars in investment, and the "fat model" hypothesis (the value will be concentrated in the model itself). He pointed out that these mainstream views have been shaken or even overturned, and the industry landscape is being reshaped.
In addition, Deepseek has greatly reduced the cost of inference, reducing the inference cost of OpenAI's most advanced model by 30 times, making AI-embedded applications much cheaper and more widespread. Nic Carter believes that while this innovation may change the proportion of capital expenditure in AI from training to inference, it will not weaken the equity value of companies in the infrastructure such as GPUs and data centers. On the contrary, the increase in inference demand will accelerate the transition from traditional technology to comprehensive AI embedding.
He also reiterated his doubts about the "fat model hypothesis", believing that model companies (such as OpenAI and Anthropic) will face severe capital consumption dilemmas due to the rise and intensified competition of high-quality open-source models, but this will not affect the value of companies at the AI infrastructure level. He expects the market to be highly volatile in the short term, but there is no need to panic in the long run.