Semicon Alpha

From Token Maxxing To Token Rationing In Just Three Months. What's Going On?

William Martin Keating's avatar
William Martin Keating
Jul 01, 2026
∙ Paid

Just a little over three months ago, we wrote about a newly emerging phenomenon dubbed Token Maxxing:

The Token Maxxing Delusion

William Martin Keating
·
Apr 22
The Token Maxxing Delusion

Earlier this month, Fortune reported that Meta’s 85,000 employees were competing on an internal AI leaderboard called “Claudeonomics”, details here.

Read full story

Meta had an internal leaderboard called “Claudeonomics” where 85,000 employees competed for titles like “Token Legend.” Jensen Huang told the All-In Podcast that if his $500,000 engineers weren’t burning $250,000 of tokens a year, he’d be “deeply alarmed.” Job candidates were reportedly negotiating their token budgets before signing offer letters.

We labelled this trend a delusion: without a clear correlation between token consumption (spending) and return on said investment, this was never going to end well. Over the course of the ensuing three months since then, things have changed dramatically.

The companies that told employees to consume as many tokens as possible are now capping them. The executives who bragged about astronomical consumption are now asking, on the record, whether any of it produced anything. Token maxxing is over. Welcome to token budgeting. Let’s dig in…

Semicon Alpha is a reader-supported publication. To receive new posts and support my work, consider becoming a free or paid subscriber.

User's avatar

Continue reading this post for free, courtesy of William Martin Keating.

Or purchase a paid subscription.
© 2026 William Martin Keating · Privacy ∙ Terms ∙ Collection notice
Start your SubstackGet the app
Substack is the home for great culture