Mark Cuban on the AI Bubble: Who Actually Gets Wiped Out?

Cuban's opening claim is that this is not a dot-com rerun: there are no zero-revenue companies IPOing to retail buyers, so a bust would not hit most Americans directly. The exposure sits with VCs, PE and private credit funds who deployed at peak prices, and with market leaders funding hundreds of billions of capex through borrowing — 50-year bonds layered on top of an existing private-credit problem, which he calls planning for perfection. His fiber-optic analogy does the work: just as bandwidth breakthroughs stranded dark fiber that later sold for pennies on the dollar, a price-performance breakthrough in AI could leave data centers that can't get power turned on being converted into pickleball courts. His hedge on his own prediction: if he's wrong about the data centers, it will be because video and world models drive demand he isn't counting on.

The middle stretch is Cuban the operator. He tells portfolio companies to go public because stock is acquisition currency — a lesson from broadcast.com, where he bought five companies with stock, and from the famous Yahoo collar, which required Goldman Sachs to build a custom index of internet stocks he could short before a real collar was possible. He lost tens of millions on the short and considers it one of the great trades anyway. On AI itself he is bullish and dismissive at once: enterprise deployment is far harder than promised, and the fact that Anthropic, OpenAI, Microsoft and Palantir all need forward-deployed engineers is, to him, proof that AGI talk is premature — 'you should just be able to ask AI to do what I need you to do.' The opportunity is the gap: anyone with basic computer literacy can walk into a business and fix the failures AI leaves behind. He cites Lovable, where he's an investor, as the counter-case: 770,000 applications built per week, only 20% of users engineers.

The back half ranges wider. Health: he has ten years of his own blood-test data, uses Open Evidence to untangle a supplement-medication conflict, and sees self-directed healthcare as an early AI win that still won't replace doctors. Politics: DSA-style socialism is a local phenomenon, elections are decided by whoever controls the algorithm, and his stated hope is that truth-seeking LLMs — whose currency is correct answers, unlike social media's engagement — become the counterweight. He recounts calling the Berkeley economist behind Elizabeth Warren's wealth-tax model, who admitted it modeled a single year with no behavioral analysis. And he closes on the NBA: the second apron ends three-peats, the Knicks' title, and franchise valuations now riding on streaming subscriptions rather than attendance or ratings.