Search Google Market
AI as a Factor of Production: Jackson Hole, Claude's Fine Print, and Burry's Warning
Federal Reserve Chair Kevin Warsh delivered his first Jackson Hole keynote this weekend, and the sentence generating the most sustained analysis was his suggestion that artificial intelligence could join labor, capital, and land as a factor of production in the models the Fed uses to gauge economic capacity. The implication is significant: if AI genuinely expands productive capacity in the way Warsh indicated, the economy could grow faster and longer without generating inflation than current models predict — a conclusion with direct and substantial implications for interest rate policy.
Warsh appeared to signal that the Fed is at least open to adjusting its models rather than waiting for lagging indicators to confirm productivity gains already visible in firm-level data — addressing what economists call the productivity paradox, the historical tendency for technology gains to take years to surface in aggregate statistics. For markets and tech investors, it validated arguments that AI productivity gains are real and large enough to change the macroeconomic math.
The timing coincided with a product announcement from Anthropic that illustrated the gap between AI's theorized future and its present operational reality. Anthropic announced that Claude Code users are receiving a permanent twenty-five percent increase in their weekly usage limits. What the fine print reveals is that the announcement coincides with the expiration of a summer promotion, meaning users operating on higher summer limits will see their available capacity drop seventeen percent on September fourteenth. It is mathematically accurate to describe the new baseline as a twenty-five percent increase over the pre-summer limit while simultaneously reducing what users had grown accustomed to. The reaction from the developer community was, predictably, pointed — and it illustrated the tension at the heart of the AI business model: compute costs are high enough that the companies building these systems are constantly managing supply against demand in ways that create friction with the very developer ecosystems they need.
Investor Michael Burry — who called the 2007 mortgage collapse when virtually every institutional investor was positioned on the other side of the trade — issued a warning this week that the AI investment thesis is dangerously concentrated. His argument: the entire trade depends on two private companies, OpenAI and Anthropic, delivering on their implied promises. Neither is publicly traded; neither has transparent financials. Billions of dollars in AI infrastructure investment — semiconductors, data centers, power generation — is predicated on a revenue model these two companies have not yet proven at scale. The counterargument holds that even if either lab stumbled, infrastructure capacity would be absorbed by Google, Meta, Amazon, or Chinese competitors, and that demand for AI capability is not dependent on any two specific labs. Burry's more targeted concern, however, is about the models themselves — the intellectual property, the training runs, the alignment research — and the switching costs and transition chaos a disruption to either lab would create for the enterprise software market that has built integrations around their products.
Google added its own layer to the ecosystem story by auto-expanding AI Overview summaries in search results, meaning users increasingly see AI-generated answers at the top of search pages with links to original sources buried further down or omitted entirely. The Search Engine Journal reported traffic drops of double-digit percentages from sites that had previously held stable search rankings. The legal framework governing this behavior — the Sherman Antitrust Act of 1890 — prohibits not simply having a dominant market share but using monopoly power to harm competition through exclusionary conduct. Regulators are arguing that Google is using its search monopoly to advantage its AI products over competitors; the legal question is not whether Google is large but whether it used search dominance to unfairly entrench its AI products in an adjacent market. That legal process is moving slowly while the market reality is moving fast.