AI Spending Has Eclipsed the Dotcom Bubble, But Corporate Budgets Are Cracking
Corporate investment in AI and related computing equipment has exceeded the peak of the late 1990s dotcom boom relative to the US economy, according to Jefferies. Yet companies are now pulling back on spending as they confront the true cost of widespread AI adoption.
US spending on information processing equipment and software reached 4.91% of gross domestic product in the first quarter of 2026, surpassing the 4.46% peak during the 2000 dotcom collapse. AI spending alone contributed 1.34 percentage points to annualised US economic growth of 2% in that quarter.
The warning signs are mounting. Microsoft has cancelled internal licences for Claude Code, an Anthropic coding tool, and switched back to its own GitHub Copilot product. Uber's chief technology officer reported internally that the company burned through its entire 2026 AI budget in four months.
The Tokenmaxxing Problem
The culprit is what Jefferies calls "tokenmaxxing"-a reference to how AI models charge for output in units called tokens. Companies encouraged staff to use AI freely through internal leaderboards, and employees began running up bills on unnecessary tasks to inflate their scores.
The result: rising costs with no productivity gains. Cloud service providers are raising prices roughly 30% year-on-year, compressing corporate budgets at a time when per-token AI costs are falling faster than underlying computing costs.
Anthropic's annualised revenue run rate surged from $9 billion at the end of 2025 to over $44 billion by early May 2026, illustrating both the scale of adoption and the cost burden spreading across the economy.
Capital Destruction Ahead
Jefferies said a significant portion of capital will ultimately be destroyed in the current AI cycle, drawing parallels to the dotcom boom and 19th-century railway mania. The bank invoked Amara's Law-the observation that people overestimate a technology's short-term impact while underestimating long-term consequences.
Semiconductor stocks remain the clearest winners. SK Hynix and Micron both surpassed $1 trillion in market capitalisation this week, with the Philadelphia Semiconductor Index up 80% year-to-date and trading 64% above its 200-day moving average.
For finance professionals, the implications are clear: AI spending decisions require the same scrutiny applied to other capital investments, and strategic planning around AI adoption must account for actual productivity gains, not usage metrics alone.
Your membership also unlocks:
