Wall Street is betting against AI, and here's why: Uber, Amazon and the history of bets against tech

Wall Street’s biggest blind spot in AI isn’t the technology—it’s monetization. With AI spending potentially reaching around a trillion dollars in the next few years, investors expecting conventional returns (often framed as 5–10x) face a brutal math problem: that level of capital expenditure implies trillions in revenue must materialize. So far, the money is concentrating at the “picks and shovels” layer—especially chipmakers like Nvidia—while companies further up the stack that actually sell AI-powered products at scale remain comparatively scarce. That gap is widening as more firms pour in on hopes of future returns, turning AI into a high-spend, low-clarity bet.

The comparison to earlier tech cycles is central. Uber’s early growth was fueled by massive subsidies—about $30 billion over a decade—to reshape consumer behavior away from taxis. Once habits shifted, Uber could later rely on price increases and profitability drivers like Uber Eats, which became a major margin engine. The same pattern may be playing out in AI: heavy investment today could eventually unlock durable demand, but the revenue path isn’t obvious yet. The transcript frames AI as being in the “Amazon Web Services / Prime” phase—where costs and benefits weren’t initially legible to Wall Street, yet demand expanded once the pricing friction was removed.

Three historically plausible revenue sources are offered to close the “AI revenue gap.” First is breakthrough productivity driven by AI creativity. Large language models are portrayed as unusually strong at creative tasks, not just logical ones—opening the door to top-line gains rather than only cost-cutting. The example is a drug discovery breakthrough: if AI accelerates the creation of an entirely new drug class, the payoff could be enormous (hundreds of billions), directly attacking the revenue side that efficiency-only narratives often miss.

Second is efficiency gains—but with a twist. The transcript argues that efficiency doesn’t automatically translate into layoffs or lower spending. Instead, lower costs can expand demand and enable “cooler” or more ambitious work, meaning companies may get more done and cover more territory with the same workforce. That matters because it suggests productivity improvements might not show up as dramatic labor cost reductions; many workers operate with an “infinite to-do list,” so AI may help them reach 60% of tasks instead of 40%, without shrinking the overall workload. Anecdotal layoffs are described as mostly anticipatory and concentrated among already-struggling firms, rather than a broad, hard-edged productivity-driven wave.

Third is new markets and new devices. The transcript highlights that AI has mostly arrived through chat interfaces, but the next monetization wave could come from pairing language models with household robotics—an entirely new device category that could become as ubiquitous as smartphones once were. It also points to Apple’s AI push as a signal that AI may become embedded in consumer devices, potentially making language models an “operating system” layer that powers future hardware rollouts.

In the end, the core stance is cautious but actionable: the revenue gap is real, chipmakers are benefiting immediately, and the most credible monetization routes are (1) creativity-led breakthroughs, (2) efficiency that expands demand rather than just cutting jobs, and (3) device-driven new markets. Whether these threads add up to the implied trillion-to-multi-trillion revenue math remains uncertain—but the search for where the money will come from is now the defining question for AI investors.