“One chief financial officer fell off his chair when he saw the first bill."
Forget showing up early or working through lunch. In 2026, career success is increasingly tied to how aggressively workers use artificial intelligence.
Executives at companies including Amazon, Disney and Meta are now tracking employee AI activity through internal dashboards, leaderboards and usage metrics.
At Meta, staff have competed on rankings tracking the 250 most active employees, measured by how many AI “tokens” they consume. Tokens represent fragments of text processed by AI systems.
Disney operates a similar “AI adoption dashboard”, while Amazon has encouraged widespread AI use, to the point where some employees are reportedly inflating activity by assigning unnecessary tasks to AI tools.
The emerging practice, dubbed “tokenmaxxing”, has become a performative measure of productivity, even when outputs are difficult to quantify.
One tech worker compared it to tracking sawdust on a building site, saying it signals activity rather than meaningful output.
But the strategy is proving expensive. As firms push workers to maximise AI usage, IT budgets are coming under strain from rapidly rising compute costs linked to large language models.
The financial pressure is already visible at major firms.
Uber’s chief technology officer Praveen Neppalli Naga revealed in April that the company exhausted its annual AI budget in under four months, driven by demand for coding tools such as Claude Code.
Andrew Macdonald, Uber’s chief operating officer, later said the returns were uncertain, adding:
“It’s very hard to draw a line between one of those stats and, ‘OK, now we’re actually producing 25% more useful consumer features’.”
Danny Quilton, of tech consultancy RedMonk partner Capacitas, said the reaction inside some firms has been extreme, saying:
“One chief financial officer fell off his chair when he saw the first bill.”
Although consumer AI tools like ChatGPT are often free or subscription-based, enterprise systems are typically billed per token. A single query may cost pennies, but at scale, usage can reach millions or billions of tokens.
More advanced tasks dramatically increase costs. A short question might use a handful of tokens, while generating reports or code can consume thousands per request.
According to analysts, even models priced at around £4 per one million tokens are becoming costly at enterprise scale, particularly as engineers deploy automated “agent” systems that run continuously in the background.
Some teams now run swarms of AI agents that analyse data overnight or write and test code autonomously.
These systems are often themselves managed by additional AI agents, multiplying usage further.
Boris Cherny, head of Claude Code at AI firm Anthropic, says he personally runs hundreds of agents simultaneously for coding tasks.
Meanwhile, Peter Steinberger, who sold his AI tool OpenClaw to OpenAI, reportedly spent £1m in a single month on token usage, with costs covered by his employer.
Nvidia chief executive Jensen Huang has argued that AI usage should scale with salaries:
“If that $500,000 engineer did not consume at least $250,000 worth of tokens, I am going to be deeply alarmed.”
Nvidia, whose chips power much of this infrastructure, stands to benefit from rising compute demand across the industry.
However, experts warn that token usage is a poor proxy for productivity.
James Governor, founder of tech advisory firm RedMonk, said:
“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.
“Most organisations are just not in a world in which they can practically spend those sorts of sums on productivity that is not proven.”
Costs are also rising quickly.
Research suggests token pricing has almost doubled since the start of 2026, increasing 26% since May as demand for advanced models grows.
Agent-based systems are a major driver of this growth, as they require significantly more compute power than standard chatbot interactions.
These rising expenses are forcing some companies to reconsider their approach.
Andrea Zimmerman, an executive at Target, said heavy AI costs were prompting the retailer to “re-evaluate our strategy”.
Duolingo chief executive Luis von Ahn has also reversed earlier plans to evaluate staff based on AI usage, saying he would not “force” employees to use the technology.
At the same time, some firms are attempting to offset AI spending by cutting costs elsewhere.
Meta recently reduced its workforce by around 8,000 roles, while Uber has slowed recruitment to balance rising AI expenditure.
Analysts at Goldman Sachs estimate that AI spending linked to engineering roles is approaching 10pc of the cost of human labour, with parity potentially close if current trends continue.
Bola Rotibi, analyst at CCS Insight, said:
“Some organisations may initially be tempted to re-prioritise spending away from hiring and towards AI.
“But over time, that balance will be revisited as leaders demand clearer links between AI spend and proven outcomes.”
Ultimately, the question for many firms is whether “tokenmaxxing” represents real productivity or simply expensive signalling in a rapidly inflating AI economy.
Loading ...