At the beginning of April 2026, the US portal The Information revealed that Meta was running an internal competition between employees to see who could use the most AI tokens. In total there are said to have been around 60 trillion tokens within just 30 days.
Tokenmaxxing: High costs, low output
The person at the top of the ranking had used over 280 billion tokens alone, according to The Information. Converted to US dollars, this can mean costs ranging from several hundred thousand to a few million, depending on the API price.
Editorial recommendations
${content}
${custom_anzeige-badge}
${custom_tr-badge}
${section}
${title}
In mid-May, Openclaw founder Peter Steinberger proudly announced that he had used AI tokens worth around $1.3 million in 30 days. In his case, his new employer OpenAI covers the costs.
Meanwhile, the problem in many tokenmaxxing cases is that simply increasing the consumption of AI tokens does not have any correspondingly positive effects. This means that the trend is often simply expensive, but does not produce the desired results that would allow the money to be recouped.
Tech companies are slowing down the token trend
Meta stopped the internal token consumption competition a few days after it became known. Amazon, where – like Google and OpenAI – similar internal competitions are supposed to take place, also put a stop to this at the end of May. Microsoft had withdrawn access to Claude Code from external engineers and switched them to GitHub Copilot CLI, as Trendingtopics writes.
After all, companies don’t want the corresponding costs to rise to unexpected heights. This happened to one of his customers, as an AI consultant explained to the Axios portal. This customer was suddenly presented with an invoice for half a billion US dollars. The problem: He had not set any limits on AI token usage for his employees.
Uber: annual budget consumed in 4 months
And this is not an isolated case. Uber CEO Dara Khosrowshahi had to admit that output fell short of token consumption, as Gizmodo reports. For many developers, it seems as if AI is free. The consequence: Uber claims to have used up its annual budget for AI tokens in the first four months of 2026.
This is how Deepseek works – and what makes it different from other AI models
According to a Wall Street Journal report, many companies are now putting the brakes on their AI token releases. So much so that it could “complicate AI’s triumph throughout the economy.”
AI expert: Token minimization as a new trend
According to Trendingtopics, Eugene Cheah, CEO and co-founder of Featherless.ai, provides an explanation for this: “Token usage is a metric, but extreme consumption in the guise of tokenmaxxing is, in most cases, not a sustainable business model and an inaccurate way to capture real value.” According to Cheah, he assumes that the next phase of AI use in companies will be characterized by token minimization.
See more in Google News:
Top Article
${content}
${custom_anzeige-badge}
${custom_tr-badge}
${section}
${title}
