Companies Impose Token Caps as AI Spending Spirals Out of Control

From unlimited to rationed

Pylon CEO Marty Kausas faced a stark choice this month: cap employee AI spending or accept a $1.4 million bill. As his software company approached 150 employees on its Anthropic plan, costs threatened to more than triple. His response signals a broader shift in corporate AI strategy — the era of unlimited token spending is over.

Kausas is now working with his VP of finance to set token caps for non-technical employees, joining companies like Coinbase, Walmart, and Amazon in pulling back from unrestricted AI access. The reversal comes just months after many of these same organizations encouraged workers to maximize their AI usage through leaderboards and incentive programs.

OpenAI CEO Sam Altman noted the dramatic change in tone. At the start of 2026, companies were comfortable with their AI spending. Now, according to Altman, costs have become "a huge issue."

Why it matters

This spending whiplash creates a new workplace dynamic where engineers must advocate for the compute resources they need to remain competitive. Companies that restrict AI access too aggressively risk falling behind technologically and losing talent to competitors offering better token budgets. The shift also reveals a fundamental tension: AI tools have become essential for productivity, yet their costs scale unpredictably as adoption grows.

The tokenomics squeeze

Max Kan, a tokenomics analyst at data provider SemiAnalysis, builds token models for hedge funds and hyperscalers. In May, he argued that $10,000 in tokens could double the productivity of a $100,000-per-year employee. Those calculations came during the "tokenmaxxing" phase, when companies pushed engineers to consume AI resources aggressively.

The word "tokens" appeared in 129 earnings calls during Q2 2026, up from 57 the previous quarter, according to analysis by AlphaSense performed for Business Insider. Within weeks, the belt-tightening began. Amazon shut down its internal token leaderboard. Multiple enterprises implemented spending caps.

Kan worries about the message this sends to workers: "My boss is adamantly pushing me to do one thing, then I did that thing, and now I'm getting yelled at because I did that thing too well."

New hiring dynamics

Token access has become a recruiting tool and a negotiation point. Job candidates now ask detailed questions during interviews about AI budgets, model tiers, and lab partnerships, according to AI advisor Allie K. Miller. Some hiring managers guarantee token allocations to attract top engineering talent.

Max Christoff, CTO of legal tech company Everlaw, advocates for giving engineers token caps while allowing them to negotiate for larger budgets when justified — similar to how cellular data worked before unlimited plans became standard. He wants to eliminate mindless AI usage while ensuring engineers can access resources when they deliver real value.

Yet average AI spending continues rising. Technology and media companies spent $66.29 per employee on AI in May, up from $58.84 in April, according to Ramp's AI Index. Lead economist Ara Kharazian expects this metric to keep climbing, though he sees early signs of cost management through tools like model routers.

Career consequences

Brock Simon, who advised companies on AI for Bain & Company before founding startup Native, watched some organizations restrict AI tool access or slow adoption. The result: employees fell behind the technology curve. "It really hurt some people's careers," Simon said.

Kausas acknowledged the stakes at Pylon. Telling engineers they cannot use AI products would drive them away, he said. "It would feel like you were in the Stone Age." His challenge now is finding return on investment while avoiding engineer backlash — a balance many companies are struggling to strike.

These details were first reported by Business Insider.