Rising costs of the token consumption race

Rising Costs of the Token Consumption Race

August 27, 2026 – 8:08 am

The Cost of Tokenmaxxing

Credit: Vwalakte via Magnific

Lately, we’ve witnessed a surge in a practice called Tokenmaxxing. More and more companies are tracking their employees’ productivity using tokens spent on AI usage. Jensen Huang, CEO of Nvidia, summarizes the sentiment clearly:

“If that $500,000 engineer did not consume at least $250,000 worth of tokens, I am going to be deeply alarmed.”

Critics might argue that Jensen has an incentive to advocate for higher token usage, as it directly translates into higher revenue for Nvidia. This sentiment is echoed in corporate America across various product companies and startups. Databricks CEO Ali Ghodsi mentioned a single engineer who spent over $7000 in AI tokens, while startups like Sendbird have leaderboards tracking each employee’s token expenditure.

The push for increased token consumption isn’t limited to tech companies; it’s happening across the board. Legal tech startup Harvey’s token spend has increased by roughly 12X to 12 trillion tokens per month. It’s clear that everyone wants to increase their token spend, but understanding the reason behind this trend is crucial.

More tokens equal more productivity, or so the claim goes from AI enthusiasts and early adopters of tokenmaxxing. There’s a race to spend as many tokens as possible, and companies are readily paying the price. The adoption of large language models (LLMs) represents a new way of working across sectors, and those that don’t evolve quickly risk being left behind.

The LLM Bill is Here, and It’s Massive

Since the initial fervor of tokenmaxxing in early 2026, we’ve seen many companies implement caps on token usage as costs have grown exponentially. One of the most notable stories came when Uber CTO Praveen Neppalli Naga revealed in an interview with The Information that they had burned through their annual AI budget in just four months and were reassessing their strategy.

As Sarah Perez reported, Uber is not alone in facing an AI budget reckoning. Meta’s Adam Mosseri envisions a future with token limits on each employee, having already implemented policies to reduce token consumption. The AI spend leaderboard that made headlines earlier this year has been shut down. Microsoft cancelled Claude code licenses and consolidated everyone under Copilot.

The cost of tokenmaxxing is here, and it’s essential to delve into the root causes behind this massive increase in enterprise balance sheet costs. Several factors contribute to the rising costs associated with LLM usage:

Widespread Adoption: In 2026, companies created internal dashboards to track AI spend and encouraged employees to use more AI in their daily tasks.
Model Selection: One of the primary reasons for high token cost is model selection. Top-tier models are 5 to 10 times more expensive than their optimized counterparts.

The following table offers a closer look at Claude’s optimized and top-tier models (pricing from July 2026):