Revenge of the AI bubble

The AI bubble debate has lurched through at least three frenzied phases in the span of three years:

1. Suspicion: Historic sums of capital poured into AI before anyone proved it could reliably automate work. A violent market correction felt inevitable.
2. Mania: Claude Code and autonomous agents made the early skepticism look outdated, fueling a corporate scramble to embed AI everywhere and maximize usage.
3. Reckoning: Companies discovered that AI can be extraordinary when aimed precisely — and ruinously expensive when treated as a universal productivity machine.

Why it matters: The first phase doubted the technology. The second phase worshipped it. The third phase — currently gaining steam across Corporate America — questions whether AI's immense power is worth the price.

Zoom in: The case against AI used to come from outsiders — Luddites, "doomers," short sellers betting on a crash. Its newest skeptics are emerging from inside the boom.

• Uber capped employee AI usage after burning through its annual Claude Code budget in four months. A top executive said the spending was getting "harder to justify," with no clear link between token use and more useful consumer features.
• Amazon shut down an internal token leaderboard after employees gamed it with throwaway tasks to climb the rankings. An Amazon executive told staff, "Please don't use AI just for the sake of using AI."
• GitHub moved Copilot, the AI coding assistant used by millions of developers, to usage-based billing as part of its effort to create a "sustainable" business. The change shocked users who were suddenly confronted with the true cost of heavy AI usage.
• Bain surveyed 951 large companies and found AI savings falling well below projections, even as most firms planned to spend more. "The technology worked. The value didn't arrive," the report concluded.

The intrigue: Even OpenAI CEO Sam Altman has acknowledged the new concerns, calling the question of whether AI spending will show up in revenue "the most fair criticism" of the moment.

Reality check: The companies sounding the alarm are the early adopters. Most of the economy is still at the starting line, while the pioneers are the ones absorbing the cost shocks, wasted tokens and employee backlash.

• AI is already creating real value for chipmakers, model labs and some power users. The harder question is whether that value spreads across the companies paying to deploy it.

By the numbers: Wall Street got a fresh reminder Friday of how much AI optimism is baked into markets.

• The Nasdaq plummeted 4.2%, its worst day in more than a year, while the Philadelphia Semiconductor Index plunged 10.3%, its worst day in more than six years.
• One culprit was Broadcom: The chipmaker reported explosive AI growth, but failed to raise its longer-term AI revenue outlook — disappointing investors looking for signs that demand was still accelerating.

The bottom line: AI can make the right worker dramatically more productive, but those gains depend on knowing exactly where and how to apply it. The real bubble may have been the assumption that AI could be sprayed across companies, employees and workflows and reliably pay for itself.