Everyone Knows It's a Bubble. What Happens Now?

AI valuations are being propped up by a tightly interlocked financing and contracting network—while day-to-day workplace reality shows AI is mostly failing to deliver promised productivity gains. That mismatch is driving layoffs and “AI-driven” cost-cutting narratives, but it also raises the risk that investors eventually realize the returns don’t match the hype, triggering a broader reckoning.

The scale of the money is hard to ignore. Nvidia’s valuation has surged to about $5 trillion, and total global AI spending is projected to reach roughly $375 billion this year. Yet outside the AI buildout—especially data centers and processing—US GDP growth for the rest of the economy is described as only about 0.1% in estimates cited from Harvard economists. The result is a sense that “AI is the economy now,” which feels both politically and economically destabilizing.

A central thread is how corporate cash flows circulate through overlapping deals. A simplified chain is laid out: Nvidia sells chips; Oracle buys chips to run data centers; OpenAI pays Oracle to use those data centers. But the money doesn’t just move one way. A large portion of OpenAI’s payments is traced back to Nvidia through a roughly $100 billion deal, and Nvidia’s investments expand further—into OpenAI, Intel, and Coreweave. The same pattern is said to repeat across the ecosystem: companies hedge bets by investing in one another while also trying to stay dominant in hardware, servers, and models. The overlapping circles create a “bubble” dynamic where high revenues and rising stock prices can be sustained by continued dealmaking, even when underlying profitability remains elusive.

That fragility matters because workplace adoption doesn’t match the investment story. Layoffs are attributed to automation and AI, but the transcript argues that AI hasn’t replaced workers in a meaningful, economy-wide way. Instead, AI is portrayed as a managerial excuse: executives can justify layoffs while still rehiring a portion of workers soon after, or shifting remaining staff into more intense, error-prone roles. Multiple adoption findings are cited to support the gap between promise and performance—one study claiming GenAI fails in 95% of implementations, another finding that AI increased workload for a small minority of Danish workers, and research on programmers reporting longer coding times on average.

If AI is mostly a cost narrative rather than a productivity engine, the next question becomes what happens when investors lose patience. The transcript offers two competing possibilities. One is that the complex web of tech-focused players—Meta, Microsoft, and Amazon—could absorb losses because they hold cash reserves and are already profitable. The other is that newer, riskier financing structures could spread stress beyond tech. Meta is highlighted for using tradable securities backed by data-center leases, with hedge funds and additional debt entering the picture. If data-center renters can’t pay leases—because AI demand cools—losses could ripple into pensions and mutual funds, and potentially into banking.

The piece ends with a political-economic warning: since the system relies on hype and managers’ ability to fire workers at will, worker organizing is framed as the main counterweight. The claim is blunt—AI as deployed in the United States is described as a waste of money and resources—and the practical takeaway is that nothing about AI replacing workers is inevitable if labor power can limit the scam’s operating conditions.