Out-of-control employees are blowing AI budgets alarmingly fast

Yahoo Finance ·

Out-of-control employees are blowing AI budgets alarmingly fast James Titcomb Wed, 27 May 2026 at 1:00 am GMT-5 5 min read AMZN DIS META Forget showing up to the office early or working through your lunch break. If you want to impress your boss in 2026, you need to be “tokenising”. As executives seek to turbocharge productivity, and demonstrate to shareholders that they are embracing artificial intelligence, companies including Amazon , Disney and Meta have been tracking their workers’ use of AI tools. At Meta, staff competed to top a leaderboard showing the social media giant’s 250 most prolific employees, measured by how much they were spending on tokens, a unit of AI usage. Disney has a similar “AI adoption dashboard”, while Amazon has been pushing AI on staff to the extent that some employees are reportedly inflating scores by getting AI systems to carry out unnecessary tasks. Tokenmaxxing allows staff to demonstrate that they are using AI as much as possible, even if the results are less quantifiable. One techie compares it to measuring sawdust on a building site – without it, you can tell nobody is cutting wood. But this performative productivity does not come cheap, and as tokenmaxxing staff blow through IT budgets, companies are grappling with the astronomical costs of the AI boom . After a flurry of excitement around rapidly developing AI tools, executives are now facing a financial reckoning. Praveen Neppalli Naga, Uber’s chief technology officer, revealed in April that the ride-hailing giant had burned through its entire annual AI budget in less than four months as staff swarmed to programming tool Claude Code. Andrew Macdonald, Uber’s chief operating officer, admitted last weekend that the returns from this splurge had been less than certain. “It’s very hard to draw a line between one of those stats and, ‘OK, now we’re actually producing 25pc more useful consumer features’,” he told the Rapid Response podcast. One chief financial officer “fell off his chair when he saw the first bill”, says Danny Quilton, of tech consultancy Capacitas, which counts retailers and private equity firms among its clients. Most consumers experience AI through bots such as ChatGPT, which are free or accessed through a monthly subscription. But large companies are increasingly seeing their AI use metered, paying by the “token” – typically a word or fragment of a word. A chatbot answering a simple question such as “What is the capital of France?” might only take a handful of tokens. Generating a 500-word essay will cost hundreds, while programming a simple video game might be a few thousand. Each of these requests costs pennies – today’s most advanced AI models cost $5 (£4) for one million tokens – but engineers are finding ways of blowing through billions. Some engineers and product developers have constructed swarms of “agents” to analyse data through evenings, nights and weekends, which are often themselves managed by other AI agents. Boris Cherny, the head of Claude Code at AI company Anthropic, says he has hundreds of agents running at any one time, writing and testing code. Peter Steinberger, who sold his AI tool OpenClaw to OpenAI, recently revealed that he spent $1.3m (£1m) in a month on tokens, with the cost borne by his employer. Jensen Huang, the chief executive of chip giant Nvidia , has said that a software engineer on a $500,000 salary should be expensing at least half that again on tokens. “If that $500,000 engineer did not consume at least $250,000 worth of tokens, I am going to be deeply alarmed,” he said. Nvidia’s AI chips process these tokens, so Mr Huang stands to benefit from giant IT budgets. While AI is undoubtedly making some workers more productive , its efficacy is difficult to measure. “If you say to everyone ‘burn a bunch of tokens’, then you’re going to burn a bunch of tokens. But it’s not a valid measure of productivity,” says James Governor, the founder of tech advisory firm RedMonk. “Most organisations are just not in a world in which they can practically spend those sorts of sums on productivity that is not proven.” Meanwhile costs are ballooning. According to research company Silicon Data, the cost of tokens has almost doubled since the start of January 2026 and risen 26pc since the start of May as developers demand access to the latest and most expensive systems. Agents, which are designed to automate swathes of white-collar work , also consume vastly more computing resources than the more primitive chatbots many consumers use. These costs could continue to rise as heavily loss-making AI companies attempt to show investors they can make a profit. Both OpenAI and Anthropic are preparing for stock market listings as soon as this year. According to analysts at Goldman Sachs, AI spending associated with engineering roles is approaching 10pc of what it costs to pay a human, but at current growth rates, the two could be on par within months. Some companies are starting to rethink this heavy spending. Andrea Zimmerman, an executive at US retailer Target, said on Tuesday that heavy AI costs were forcing the company to “re-evaluate our strategy”. Luis von Ahn, the chief executive of language learning app Duolingo, recently made an about-turn on plans to evaluate staff based on how much they were using AI, saying he would not “force” employees to use the technology. AI companies are also encouraging use of their cheaper, less-capable systems to frugal customers. But others are trying to free up cash for more AI spending, often at the expense of employees . Meta cut 8,000 staff last week – about a tenth of its entire workforce – in part so it could spend more on AI. Dara Khosrowshahi, Uber’s chief executive, told investors in May that the company was “offsetting” its AI spending by slowing down recruitment. Bola Rotibi, an analyst at CCS Insight, says this fervour may not last. “Some organisations may initially be tempted to re-prioritise spending away from hiring and towards AI,” she says. “But over time, that balance will be revisited as leaders demand clearer links between AI spend and proven outcomes.” Ultimately, if the cost of tokenmaxxing continues to rise, the wisdom of replacing humans may start to be questioned. “In some cases, it’s going to make more sense to hire a human than it is to use AI,” says Mr Governor.

AI 시장 분석

News reports indicate that AI budgets are being depleted alarmingly fast due to out-of-control employees. This suggests that companies are facing significant challenges in budget management and efficiency during their AI adoption processes.

상승 영향

하락 영향

AI가 생성한 분석으로 투자 자문이 아닙니다.

DYAX Investor Sentiment

Bullish (Long) 55% · Bearish (Short) 45%

468 participants

Related News

원문 보기 — Yahoo Finance