The Shovels Cost More Than the Miners

Nvidia’s own VP just admitted something that should terrify every CFO who approved a layoff this year.
There’s a term going around tech circles right now: tokenmaxxing. It means exactly what it sounds like. Employees burning through as many AI tokens as possible because leadership told them heavy usage equals productivity.
Amazon built an internal leaderboard for it. A literal ranking of who used the most AI. Employees figured out the game fast and started running pointless tasks just to climb it. Amazon quietly took the leaderboard down once that became obvious. Meta built something similar and had to walk it back too.
Here’s the part that made me stop scrolling. Bryan Catanzaro, a VP at Nvidia, the company that makes the chips powering all of this, said on record that the cost of compute for his own team now exceeds what the company spends on the employees using it.
Read that again. The company selling the shovels is telling you the shovels cost more than the miners.
And yet Jensen Huang, Nvidia’s own CEO, said a $500,000 engineer should be spending at least $250,000 a year on tokens to be considered productive. Not efficient. Productive. As if the amount you spend is the same thing as the value you create.
A pitch I know well
I’ve spent fourteen years negotiating vendor contracts. I have sat across the table from people selling me exactly this kind of story. Big number, urgent tone, spend now, figure out the return later. Every procurement person in the world has a name for that pitch, and it is not usually a compliment.
Companies obliged anyway.
- Big Tech has announced $740 billion in AI capital spending this year, a 69 percent jump from last year
- Uber burned through its entire annual AI budget in four months and had to cap spending per employee just to survive its own experiment
- Microsoft is now quietly steering its own engineers toward cheaper internal tools because the good ones got too expensive to run at scale
- MIT’s own research this year found that 95 percent of enterprise AI pilots showed no measurable impact on the bottom line. Not underperformed. No measurable impact.
Meanwhile, back home, engineers were let go so the company could “reallocate toward AI.” Nobody in that townhall slide mentioned that the thing they were reallocating toward might currently cost more than the salary it replaced.
What I’m actually against
I am not against AI. I use it every day, and I have built parts of my own business around it. What I am against is the certainty. The confidence with which a company will fire a person over a technology whose own top vendor cannot yet prove pays for itself.
We didn’t get replaced by something better. We got replaced by something expensive that hasn’t finished proving itself yet, and somebody upstairs decided that was a risk worth taking with someone else’s salary.
Have you seen a cost-cutting decision at your company that, when you actually looked at the numbers, wasn’t cutting any cost at all?


