Speculative manias, like the AI boom, function like collective hallucinations. The overwhelming belief in future demand becomes self-fulfilling, attracting capital that builds tangible infrastructure (e.g., data centers, fiber optic cables) long before cash flows appear, often leaving lasting value even after the bubble bursts.

Major tech companies are investing in their own customers, creating a self-reinforcing loop of capital that inflates demand and valuations. This dangerous practice mirrors the vendor financing tactics of the dot-com era (e.g., Nortel), which led to a systemic collapse when external capital eventually dried up.

Today's massive AI company valuations are based on market sentiment ("vibes") and debt-fueled speculation, not fundamentals, just like the 1999 internet bubble. The market will likely crash when confidence breaks, long before AI's full potential is realized, wiping out many companies but creating immense wealth for those holding the survivors.

The current AI spending spree by tech giants is historically reminiscent of the railroad and fiber-optic bubbles. These eras saw massive, redundant capital investment based on technological promise, which ultimately led to a crash when it became clear customers weren't willing to pay for the resulting products.

Current AI investment patterns mirror the "round-tripping" seen in the late '90s tech bubble. For example, NVIDIA invests billions in a startup like OpenAI, which then uses that capital to purchase NVIDIA chips. This creates an illusion of demand and inflated valuations, masking the lack of real, external customer revenue.

Unlike typical consumer ads, San Francisco's outdoor advertising is dominated by niche B2B startups. They accept that 95% of viewers are irrelevant to reach a high concentration of VCs and tech talent, signaling a strategic return to immeasurable brand awareness over direct-response marketing.

In a late-stage bubble, investor expectations are so high that even flawless financial results, like Nvidia's record-breaking revenue, fail to boost the stock price. This disconnect signals that market sentiment is saturated and fragile, responding more to narrative than fundamentals.

The memo flags deals where money is "round-tripped" between AI players—for example, a chipmaker investing in a startup that then uses the funds to buy its chips. This practice, reminiscent of the 1990s telecom bust, can create illusory profits and exaggerate progress, signaling that the market is overheating.

Howard Marks distinguishes between two bubble types. "Mean reversion" bubbles (e.g., subprime mortgages) create no lasting value. In contrast, "inflection bubbles" (e.g., railroads, internet, AI) fund the necessary, often money-losing, infrastructure that accelerates technological progress for society, even as they destroy investor wealth.