First up was tokenmaxxing. Now it’s time to maximize efficiency.
Silicon Valley has encouraged employees to use AI, including through gamified internal leaderboards that measure the number of tokens (units of data processed by AI) employees use.
For some, playtime is over.
There is a big debate in the tech industry about whether token maxing, the use of large amounts of AI tokens to increase productivity, has gotten out of control. Executives also wonder when they will start seeing a return on investment.
Last week, the Financial Times reported that Amazon shut down an internal dashboard that tracks AI usage after some staff members performed tasks just to move up the leaderboard.
“Don’t use AI just for the sake of using AI,” Dave Treadwell, Amazon’s senior vice president, told employees. “We use AI to help solve customer problems, solve business problems, and innovate.”
An Amazon spokesperson told Business Insider that employees set up the unofficial dashboard several weeks ago to “raise awareness” of how AI can accelerate their work, and that it was “never intended to promote the use of AI for that purpose.”
Amazon is not alone in recalibrating its use of AI. Uber Chief Operating Officer Andrew McDonald said in an interview published in late May that the company has yet to see improvements directly tied to increased AI spending.
As OpenAI and Anthropic aim to go public this year, some AI skeptics see them as signs of a bubble.
“When other companies report the same thing, the bubble bursts,” Gary Marcus, an AI researcher and professor emeritus at New York University, wrote in X last week about Uber’s experience.
Tech giants want to reduce AI costs
For executives concerned about the cost of AI, the technology industry offers mixed signals about how AI will unfold.
This week, GitHub Copilot, Microsoft’s AI-powered coding assistant, is moving from fixed monthly payments to pay-as-you-go billing. GitHub said in an April blog post that it has so far absorbed much of the technology’s cost increases and that its fixed-price model is “no longer sustainable.”
Like other companies like Anthropic and OpenAI, we are moving away from flat rate seats to usage-based billing for our business customers.
While the price changes have frustrated some developers, investors say they are a necessary move and a sign that the market is maturing.
“Over the past year, many AI products have effectively assisted in their use in the race for growth, and we are now seeing the cost of that,” Barry Downs, managing partner at investment firm Sure Valley Ventures, told Business Insider.
“This is a healthy transition and not a red flag,” he added.
Uber COO Andrew Macdonald’s comments sparked a discussion about tokenmaxxing. Bloomberg/Getty Images
On the other hand, costs are likely to fall over time as AI companies develop more efficient models and use them to gain a competitive advantage.
Google, for example, says its latest Gemini 3.5 Flash model, as well as Anthropic’s latest Opus 4.8 model, rivals Frontier products at a lower price point. As Business Insider’s Hugh Langley reports, Google is in a strong position to reduce costs because it owns the full stack: chips, data centers, cloud, models, and many large-scale applications.
AI labs have been competing on intelligence for years, but now they’re competing on intelligence per dollar, with OpenAI, Anthropic, and others offering smaller, more efficient models.
Tokenmaxxing faces ‘necessary reality check’
Oded Tahori, founder and CEO of Jeen.ai, continues to transition away from tokenmaxxing as his startup helps organizations manage AI adoption and token spend.
He told Business Insider that the token maxing craze stems from “FOMO” and companies not understanding the full challenges of building with AI. The focus has now shifted to linking spending to results, he said.
“I think you should have goals for your budget,” he said. “And the reason to spend money is that if you want to spend $1 million on a token, you should do something good with it in your business.”
He proposed rewarding employees who use AI to build things that benefit the company with larger token budgets, and reducing the budgets of employees who could “harm” the tokens.
Some companies have already implemented incentive systems that prioritize performance over performance. For example, Business Insider previously reported that Visa rewards teams who use AI to enhance their work with internal “points” that can be used to buy things like coffee makers.
Tim Mills, managing partner at ACF Investors, told Business Insider that the token maxing discussion feels more like a “sensible practicality check” than a sign of an impending bubble.
“Tokenmaxxing is artificially distorting usage statistics for counterproductive reasons, so the crackdown is a constructive step forward,” he said. “This serves as a necessary reality check, reminding organizations that deploying these AI tools comes with very real infrastructure costs.”