The guest argues that investors are conflating the success of semiconductor manufacturers like NVIDIA with the viability of AI model and application companies. He posits that while the chip makers are profiting immensely, the companies building on top of that hardware lack sustainable, profitable business models.
AI model companies such as OpenAI and Anthropic are characterized as burning through billions of dollars with no clear path to profitability. The guest predicts that their eventual S-1 filings for IPOs will reveal devastating losses, shocking a market that currently believes in their growth story.
A key challenge for the AI industry is the inability of its enterprise customers to measure a clear return on investment. High, often subsidized, costs combined with intangible benefits are causing companies like Uber to question and cap their AI spending, signaling a potential slowdown in adoption as the true costs become apparent.
There is a significant lag, estimated at 6-12 months, just to install a single quarter's worth of GPUs, and major data center projects are reportedly far behind schedule. This indicates that the massive capital flowing into AI infrastructure is not translating into immediate operational capacity, creating a potential bubble in related 'neo-cloud' companies.
Hyperscalers are investing heavily in AI not from a position of strategic certainty, but because they lack other avenues for hypergrowth. In contrast, Apple is portrayed as being strategically cautious, avoiding massive capital expenditure and the associated risks, despite narratives that it is 'behind' in AI.
Keep pulling the thread on Ed Zitron.