The 'Tokenmaxxing' Trend Gains Traction in Corporate America

• From niche flex to corporate scoreboard
• The dashboards arrive, and the incentives get weird
• The CFOs: flying blind on a new cost center
• AI sprawl: shadow IT on steroids
• Startups join the race—carefully, or not at all
• The backlash: “stupid” and doomed?
• Tokenmaxxing as culture war inside companies
• Where this goes next

The ‘Tokenmaxxing’ Trend Gains Traction in Corporate America Human Human coverage portrays tokenmaxxing as a controversial and potentially short-lived trend in which corporates and startups overspend on metered AI usage to signal AI adoption, sometimes setting quotas or incentives that encourage gaming. While acknowledging that some leaders see AI tokens as a force multiplier and competitive necessity, these reports foreground critics who argue the practice is financially reckless, culturally distorted, and likely to be replaced by more disciplined, cost-effective approaches. @7dlt…clgf Corporate America has a new status symbol, and it isn’t headcount, patents, or office campuses—it’s how many AI tokens you can burn through before the finance team notices.

§From niche flex to corporate scoreboard

The practice now dubbed “tokenmaxxing” started as a Silicon Valley in‑joke: if AI vendors bill you per token, then the real power users are the ones torching the most compute.

It didn’t stay a joke for long. AI companies price access to models based on tokens—a rough unit of text and computation—and “Corporate America’s new favorite flex is the number of AI tokens it burns.”¹ What began as engineers quietly experimenting with model APIs has turned into a visible, tracked, and sometimes gamified race to prove AI adoption inside big firms.

By early 2026, major players like Visa were bragging about the pace of that burn. The company “proudly shared how it doubled its token usage from 1 trillion in February to 2 trillion in March,” a jump framed less as a cost spike and more as proof of being out front in the AI race.¹

At the same time, large financial and entertainment companies were wiring this metric deep into their internal dashboards. JPMorgan and Disney now track employees’ AI usage, with Disney’s internal “AI Adoption Dashboard” slicing and dicing who is using what, and how often.¹ In other words, tokens have gone from obscure billing detail to performance proxy.

§The dashboards arrive, and the incentives get weird

Once executives started asking, “How many tokens are we using?” the next question was obvious: “Which teams are pulling their weight?” The result is a wave of internal leaderboards, targets, and KPIs.

Employees at companies like Meta reportedly “competed on a token leaderboard before it was taken down,” turning experimentation with AI tools into something closer to a sales contest.² JPMorgan and Disney’s dashboards serve a similar function, even if couched as “adoption tracking” rather than outright competition.¹

The logic from the top is clear: if AI is the next general‑purpose technology, a company that isn’t using a lot of it probably isn’t moving fast enough. For tech executives convinced they “need to spend big to reach their AI goals,” tokenmaxxing fits hand‑in‑glove with a growth‑at‑all‑costs mindset.¹

But dialing up that pressure introduces a classic incentive problem. When workers know they’re being scored on usage, they’ll find ways to score, whether or not the work actually needs AI. As one analysis put it bluntly: “Still, incentivizing workers to do something that you know will run up your bill feels like a disaster (and a CFO’s nightmare).”¹

§The CFOs: flying blind on a new cost center

Behind the scenes, finance leaders are trying to impose order on a cost category that barely existed two years ago. The issue, says veteran CFO Amy Butte, now a strategic adviser at Navan, is structural: “a lack of standards for finance professionals to work from.”¹

Without agreed‑upon KPIs for AI development and usage, companies are using raw token burn as a crutch metric. Tokens are easy to count, but hard to interpret: is a 50% jump a sign of productivity gains, or just a pricey science project?

That uncertainty becomes more dangerous in organizations that explicitly reward high usage. If bonuses, manager praise, or promotion narratives lean on AI adoption, employees have little reason to ask whether the model call was necessary—or whether a cheaper tool would have sufficed.

§AI sprawl: shadow IT on steroids

Then there’s the systems angle. The push to “democratize” AI—letting anyone in the org build their own tools—plays well in innovation decks. In practice, it collides with a familiar nightmare for CIOs and CTOs: shadow IT.

“If John in accounting builds an application that speeds up his work, that’s great,” one analysis notes. “If John, Jim, Jane, and Jessica in accounting all build separate tools that sort of do the same thing but don’t talk to each other, that’s a big problem.”¹

Tokenmaxxing turbocharges that tendency. When employees are explicitly encouraged to spin up AI projects, the result can be dozens of brittle, siloed tools with overlapping functionality and inconsistent data handling. Even some of the most sophisticated companies, like Amazon, “are struggling” with the sprawl that follows.¹

§Startups join the race—carefully, or not at all

While Big Tech and large corporates can absorb larger AI bills, startups face a harsher trade‑off. Here, tokenmaxxing turns from flex to existential bet.

Some young companies are leaning hard into the burn. Business Insider found “startups take different strategies with token spending, from hard budgets to minimum quotas,” with some leaders arguing that big token bills have tangibly helped them succeed.²

Kavitta Ghai, the 29‑year‑old cofounder of ed‑tech startup Nectir, didn’t want her engineers dabbling in AI; she wanted them all‑in. She “started setting minimum quotas for Claude Code use. First it was at least $100 in tokens a week, then $200. Now, the expectation is that her engineers each spend a couple thousand in AI tokens a month.”²

The internal culture flip was real. Some of Nectir’s senior engineers, initially skeptical of AI coding tools, now call the system their “army of coders,” according to Ghai.² Yet she publicly distances herself from the hype label, insisting Nectir isn’t really part of the “tokenmaxxing” crowd: “We don’t really play into the Silicon Valley trends,” she said. “We live in our own world, and we’re competing against ourselves.”²

Other founders are more openly evangelical. Aron Solberg, the 44‑year‑old cofounder of Risotto, doesn’t want an explicit token leaderboard, “but he does want the mindset behind it,” seeing heavy token spending as a “force multiplier” for a small team.²

Their argument is straightforward: in a world where the competitor down the street is pairing every engineer with what amounts to an infinite‑patience copilot, not leaning into that leverage is self‑sabotage.

§The backlash: “stupid” and doomed?

On the other end of the spectrum, a growing camp of founders and operators thinks the whole fad is reckless.

“Some startups are tokenmaxxing. Others tell us it’s a ‘stupid’ trend that will die out,” one report summarized, after speaking with leaders who prefer “lower‑cost subscriptions” over open‑ended token burn.²

These skeptics frame tokenmaxxing as a classic arms race that primarily benefits vendors and the biggest, cash‑rich platforms. They ask who really wins when small companies on tight runways turn compute into a vanity metric. For them, capped subscription plans and tightly scoped AI use cases are safer, saner paths—especially when the ROI of each extra million tokens remains hazy.

§Tokenmaxxing as culture war inside companies

Beneath the spreadsheets, tokenmaxxing is morphing into an internal culture marker. On one side: leaders who see aggressive AI use as the clearest signal that their org won’t miss the next big wave. On the other: cautious operators worried they’re recreating the worst excesses of cloud‑spend sprawl, just faster.

The language used to describe the trend reflects that split. One story asks if we’re “tokenmaxxing too close to the sun,” capturing the mix of awe and dread around AI’s new cost center.¹ Another foregrounds the skepticism in its headline: “Some startups are tokenmaxxing. Others tell us it’s a ‘stupid’ trend that will die out.”²

Even among believers like Ghai, there’s a quiet effort to rebrand the practice as something more sober than a spending contest. Her emphasis on “competing against ourselves” is a subtle rejection of the leaderboard mentality—an attempt to decouple intensity of use from performative excess.²

§Where this goes next

In the short term, expect more of everything: more dashboards, more subtle (and not‑so‑subtle) quotas, more teams bragging about their token graphs in investor decks.

But as finance chiefs demand proof that all this compute is doing more than padding case studies, the arms race will likely evolve. Raw token burn is too blunt an instrument to survive as a primary success metric; the companies that last will be the ones that trade sheer volume for smarter measures of productivity, quality, and risk.

For now, though, one thing is clear: in the contest to look like an AI‑first company, nothing says commitment quite like watching your token counter spin and hoping you don’t discover, too late, that you’ve been maxxing out on noise.

§1. Are we ‘tokenmaxxing’ too close to the sun? — “Corporate America’s new favorite flex is the number of AI tokens it burns”; details on Visa, JPMorgan, Disney, AI sprawl, and CFO concerns.

§2. Some startups are tokenmaxxing. Others tell us it’s a ‘stupid’ trend that will die out. — Startups split between big token spend, quotas, and critics calling tokenmaxxing a “stupid” fad that will die out.

Story coverage