The more successful AI is, the more likely investors are to lose money: a century of lessons from railways, oil to the fiber-optic bubble
Original headlineAI 越成功,投资者越可能亏钱:从铁路、石油到光纤泡沫的百年教训
Not investment advice. Muhabbit scores how much a story may matter, not which way to trade.
Veteran tech investor Packy McCormick warns that AI’s very success could make investors poorer, drawing a century of parallels from British railways and early oil to the fiber-optic bandwidth bubble, where capital stampedes into supply and prices collapse even as the technology itself wins. He argues the three standard rebuttals to “AI bubble” claims—that AI is genuinely useful, that leaders post real and fast-growing revenue, and that top labs are profitable—are precisely the first three steps of the classic commodity supply cycle, not evidence against one. The piece cites aviation as the cautionary tale: Buffett’s 1999 Sun Valley quip that Orville Wright did more damage to capitalists than Karl Marx, and Berkshire’s roughly $7–8 billion bet on four U.S. airlines in 2016 that was fully exited in April 2020. The stakes are framed around Anthropic’s IPO at a reported valuation of about $1.4 trillion and hundreds of billions of dollars in compute commitments across the major labs. McCormick says it is impossible to know whether Anthropic’s bull case—a winner-take-all model that keeps self-improving and expands across every industry—or a slower revenue path crushed by OpenAI, Google, Meta, Chinese open-source models and heavy compute contracts will prevail. What to watch next is whether demand catches up before the wave of committed capacity lands, and which survivors or outside acquirers consolidate the wreckage as the lowest-cost suppliers.
Muhabbit's AI wrote this summary in its own words from TechFlow's report and translated it from Chinese. It can get details wrong, so check the source.
Named in this story
Anthropic, Packy McCormick, OpenAI, Meta (META), Google (GOOGL)
A linked name opens that asset's sentiment page.