AI token costs overtake human salaries as automation spending spirals, tech execs admit

A quick look at the computer bill in many organizations reveals that AI is increasingly more expensive than human labor.

There's something ironic in a boss discarding employees to cut costs, only to find the computer bill draining the budget faster than the people they replaced. When will tokens outpace a human employee's salary? This is the question that tech investor Jason Calacanis also found himself pondering on a recent episode of the All-In podcast.

When Jensen Huang infamously stated that he would be “deeply alarmed” if his $500,000-a-year engineer did not consume at least $250,000 in tokens, it sparked a global debate over such a dubious metric.

Bizarrely, the comments from the leather jacket-donned tech maestro paved the way for tech workers to turn their token usage into a pissing contest, the trend known as tokenmaxxing, is seeing workers racking up bills of $150,000 a month without breaking a sweat.

Uber CTO Praveen Neppalli Naga admitted that its reliance on Anthropic's Claude Code has already maxed out its full-year AI budget just 4 months into 2026. Elsewhere, even Nvidia vice president Bryan Catanzaro admitted that AI compute is now costing more than the employees using it. But how did we get here?

When AI’s savings promises start to look like a familiar cost trap

Forget the shiny tech buzzwords – boardrooms and shareholders have always been motivated by the bottom line. The brief will always look something like, "Let's automate tasks rather than paying employees to do it.”

But an Axios report revealed the inconvenient truth: that they were actually spending more on AI compute than on the employees they were meant to be displacing.

Token costs are beginning to follow a familiar script: affordability at the start gives way to rising costs once dependence is locked in. The story typically begins with securing investor capital to build a great new service, lower its price, get enterprises addicted, and then raise the price later.

Uber and AWS are great examples of making their amazing new tech affordable and accessible. It quickly took over our lives, both in and out of the office. Once everyone was on board, prices started to rise as they had to work out how to deliver on the promises made to investors who were waiting in the wings, expecting a healthy return.

We are now seeing the same story play out with AI companies that are not trying to save humanity or solve the world's big problems. They're just trying to get really rich and become bigger and more powerful companies, and even more valuable in terms of their market caps.

Predictably, CEOs driven by fomo are playing along to the same old tune, as Gartner forecasts global IT spending to grow 13.5% in 2026, totaling a phenomenal $6.31 Trillion. But many have forgotten the warning from a 2018 MIT study, which suggested that AI automation is economically viable in only about 23% of jobs. Ultimately, humans are still cheaper in the remaining 77%.

When AI Tokens cost more than human employees

We seldom see headlines about how autonomous agents are evolving into always-on, token-burning machines. Although this might be ok in pilot phases, the math doesn't add up when a Fortune 500 company attempts to scale up.

Digital employees can work 24/7 without breaks, holidays, or sick days. But the cost of continuous usage quickly adds up. Short-sighted enterprises that have replaced humans with AI could find themselves paying 5 times as much to maintain the tech. The only winners in this scenario are the tech overlords who often talk like villains from a James Bond movie.

The conversation on this topic is live. Join in the discussion.

Sam Altman has previously discussed the benefits of introducing a universal basic income to better manage economic inequality. The idea is that everyone receives a no-strings-attached cash payment regardless of their wealth or employment status.

Without a flicker of irony or self-awareness, he has since expanded his thinking to "universal basic compute." Once again on the "All-In" podcast, Altman suggested that everyone could get a slice of GPT -7's compute and choose to use it, resell it, or donate it to someone for cancer research.

When participation in society hinges on access to centrally distributed computational power, it feels as if we are sleepwalking into an Orwellian nightmare.

Rich white dudes throwing a few crumbs at displaced people feels slightly reminiscent of when Dutch political commentator Eva Vlaardingerbroek suggested that if everyone got individual carbon credits, rich people who want to go on holiday more often could buy personal carbon credits from people who can't afford plane tickets or meat.

Many business leaders have been blindsided by AI looking so cheap at the prompt level. But many are finding out the hard way that the bill increases with heavy token usage from AI agents running continuously. Then there are the infrastructure, cloud, and hardware costs.

Why AI now has to prove its worth in the real world

With Meta and Amazon both experiencing high-profile AI incidents, the good news is that human workers are needed more than ever to monitor, correct errors, and review AI-generated code. Shiny demos are not enough to move the needle.

There has been a noticeable shift at every tech conference I've attended this year, with a focus on how AI is delivering on its promises to improve business outcomes with clear ROI. More client stories and big before-and-after stats rather than yet another product announcement that's coming soon.

Make no mistake, AI won't be judged by tokenmaxxing or by how many jobs are cut. Businesses are finally returning to their senses and demanding to see examples of the measurable difference it will bring to their business.

There is no hiding the fact that future success will be built on a balance between the cost of compute and the value of human contribution.

Unlock more exclusive Cybernews content on YouTube.