What is Tokenmaxxing & Why is it Already Fading?

What started as a niche term among AI enthusiasts has quickly evolved into one of the most controversial workplace trends of 2026. At its core, tokenmaxxing refers to maximising the number of AI tokens consumed by employees or teams, often as a way of measuring AI adoption and, in some organisations, productivity.

The idea sounds straightforward. AI models such as ChatGPT, Claude, Gemini, and coding assistants process text in units called tokens. The more tokens a user consumes, the more extensively they are interacting with AI systems.

As companies spend billions of dollars on AI, many executives have begun looking for ways to measure whether employees are actually using these tools. Token consumption emerged as an easy metric because it is objective, quantifiable, and readily available from AI providers.

But the rise of tokenmaxxing has sparked a fierce debate over whether AI usage should be confused with productivity.

What Exactly Is Tokenmaxxing?

Tokenmaxxing borrows its name from internet slang such as "looksmaxxing" and "fitnessmaxxing," where individuals attempt to optimise a specific trait to the extreme.

In the workplace context, tokenmaxxing means maximising AI usage by increasing the number of tokens consumed through chatbots, coding assistants, AI agents, and other generative AI tools.

Now, to understand tokenmaxxing, it is important to understand what a token is. In AI, a token is a small unit of text that an AI model processes. A token can be a whole word, part of a word, punctuation mark, or even a space, depending on how the model breaks down language.

For example, the sentence "AI is transforming work" may be split into several tokens before being processed by an AI system. AI companies typically charge customers based on the number of input and output tokens used, making tokens a key measure of AI consumption.

The practice gained prominence as companies rolled out AI-powered coding agents capable of autonomously performing tasks for hours at a time. Rather than using AI occasionally, employees began orchestrating multiple agents simultaneously, resulting in enormous token consumption.

In some firms, internal dashboards and leaderboards were introduced to rank workers by AI usage. Notable among them were Meta and Amazon.

Supporters argue that high token usage demonstrates a willingness to experiment with new technologies and adapt to an AI-first workplace. Critics counter that it rewards activity rather than results.

Why Companies Embraced the Metric

For executives under pressure to justify massive AI investments, token usage offers a simple answer to a difficult question: Are employees actually using the tools the company is paying for?

Unlike productivity, which can be difficult to measure across different roles, token consumption is visible and measurable.

Some organisations have reportedly introduced AI adoption targets, while others have celebrated employees who consume the most tokens.

Among the most vocal supporters is Sequoia Capital partner Sonya Huang, who told The Wall Street Journal:

“We all should be tokenmaxxing.”

Interestingly, Sequoia also has its own leaderboard.

The Growing Backlash

Despite the enthusiasm, opposition to tokenmaxxing has grown rapidly. Most recently, Sridhar Ramaswamy, Snowflake CEO, called tokenmaxxing a terrible idea.

He argues that high token usage alone doesn't signal AI-driven productivity, but zero usage suggests a lack of engagement. Instead, he advocates measuring outcomes, calling it “impactmaxxing.”

Using more AI does not necessarily mean producing better work, solving more problems, or generating greater business value. Several analysts have compared tokenmaxxing to older, flawed workplace metrics such as counting lines of code written by software developers.

Amazon eventually shut down an internal AI token leaderboard after concerns emerged that employees were using AI for the sake of usage rather than solving customer problems.

As one Amazon employee reportedly observed:

“When they track usage it creates perverse incentives.”

In some companies, workers allegedly began running AI tools for unnecessary or repetitive tasks simply to increase their usage numbers.

Dave Treadwell, Amazon Sr. Vice President, told staff after scrapping the leaderboard:

"Please don't use AI just for the sake of using AI. Use AI to help you solve customer problems, to help you solve business problems, to innovate."

Meta also shut down an internal dashboard that tracked and ranked employees' token usage.

When More AI Becomes More Expensive

Another challenge is cost. Unlike traditional software subscriptions, many AI services operate on usage-based pricing models. Every token consumed contributes to the overall bill.

As employees and AI agents generate billions of tokens, costs can escalate quickly. At Uber, tokenmaxxing led to the company burning through its entire 2026 AI budget in four months.

Uber COO Andrew Macdonald became one of the highest-profile critics after questioning the connection between AI usage and productivity.

“If you’re not actually able to draw a direct line to how [many] useful features and functionality you’re shipping to your users, that trade becomes harder to justify,” he said.

His comments helped fuel a broader debate throughout Silicon Valley about whether tokenmaxxing has become more of a status symbol than a legitimate performance metric.

Microsoft, too, has dropped Anthropic's Claude Code and shifted engineers to GitHub Copilot CLI as rising AI costs and growing employee preference for Claude over Microsoft's own tools have raised concerns.

Perhaps the strongest criticism comes from leaders who believe tokenmaxxing distracts organisations from what actually matters.

Cognizant CEO Ravi Kumar recently described the obsession with token counts as vanity exercise. He said:

“For the last two years, how you consumed tokens, how much tokens you consumed was a vanity metric. I don’t think you should equate this to the number of paid hours. I don’t think you should equate this to productivity.”

Kumar argued that companies should focus on outcomes rather than usage metrics.

His comments reflect a growing consensus among management experts who warn that any metric can become distorted when employees are rewarded for maximising it.

The phenomenon closely resembles Goodhart's Law, a famous principle in economics and management, which states that when a measure becomes a target, it ceases to be a good measure.

What Comes After Tokenmaxxing?

Many experts believe the future lies in measuring outcomes rather than inputs. Instead of tracking token consumption, organisations are increasingly exploring metrics such as cost per completed task, code quality, customer satisfaction, sales conversions, bug reduction, and overall business impact.

The goal is to determine whether AI is helping employees achieve better results rather than simply generating more activity. Some industry observers have already coined a new phrase: "outcome maxxing."

Outcome maxxing shifts the focus from how much AI employees use to what they actually accomplish with it. Instead of tracking token consumption, companies adopting this approach measure tangible business results such as faster product development, higher sales conversions, improved customer satisfaction, fewer software bugs, or reduced operating costs.

If the past months were defined by the race to maximise tokens, the coming months may be about maximising impact—suggesting that tokenmaxxing, once Silicon Valley's favourite AI metric, is already beginning to fade.