Uber Capped AI Spending at $1,500 Per Employee

Enterprise AI's spend-at-all-costs era ended the week OpenAI and Anthropic filed for IPOs: on June 26, CNBC reported that Uber capped employee AI tools at $1,500 a month after its CTO admitted the company burned its entire 2026 AI budget in four months — while Lindy's CEO moved 100% of traffic off Claude to DeepSeek and a D.A. Davidson analyst warned the labs' fastest revenue growth is already behind them.

What tokenmaxxing was: For most of 2025 and early 2026, Silicon Valley treated AI token consumption as a proxy for productivity. Employers built internal leaderboards. Developers were rewarded for burning through frontier models on every task.

• Meta's employee-built "Claudeonomics" dashboard tracked 60 trillion tokens across 85,000 staff in 30 days, per The Information in April.
• The top individual user averaged 281 billion tokens — potentially millions of dollars at public API rates.
• Meta took the leaderboard down within days of media coverage. Uber capped spending. Duolingo reversed a policy tying performance reviews to AI usage.

Goodhart's Law did what it always does: optimize the metric, lose the outcome.

The reckoning: Uber's crackdown is the template.

• In April, CTO Praveen Neppalli Naga told The Information the company exhausted its full-year 2026 AI budget in four months.
• COO Andrew Macdonald called it a "head-exploding moment" in a May Business Insider interview.
• Uber now caps some AI coding tools at $1,500 per employee per month, with higher tiers available on request, CNBC reported June 26.

Macdonald's verdict on whether the spend translates to consumer value: "That link is not there yet." He added: "It's very hard to draw a line between one of those stats and, 'Okay, now we're actually producing 25% more useful consumer features.'"

The exodus: Flo Crivello, CEO of 25-person AI startup Lindy, switched his entire company off Anthropic's Claude models to DeepSeek this month.

"We did it, and you could see that cost curve go down, like, crash to the ground," Crivello told CNBC. He expects the move to save millions within months — though AI costs still exceed Lindy's payroll.

"It's a matter of survival for the business," he said. "That's all it is."

Crivello spent nearly five years at Uber. He knows how enterprise budgets work when the invoice arrives.

Why IPOs should care: OpenAI and Anthropic built their revenue run rates on the tokenmaxxing wave.

• Anthropic last reported a $47 billion annualized run rate in May, up from roughly $10 billion for all of 2025.
• OpenAI's run rate was pacing near $25 billion earlier this year, per CNBC, up from $13.1 billion in 2025.

Both filed confidential IPO paperwork in early June. Both are targeting valuations approaching $1 trillion.

D.A. Davidson analyst Gil Luria told CNBC: "Current growth rates for Anthropic and OpenAI are the fastest they will ever be, which is mostly a matter of basic math."

His read on timing: "That is a good reason to go public now, as is the concern that some of their largest enterprise customers may start limiting their out-of-control token spend."

Luria added: "There has to be some period of time in the future where there's some rationalizing of spend by companies, and that may be a blip ahead for Anthropic and OpenAI."

On June 25, The New York Times reported OpenAI is leaning toward delaying its IPO until 2027 rather than cut its valuation target. The reported delay came the same week enterprise customers started cutting checks.

The routing revolution: Roughly 95% of enterprise AI usage still runs on expensive frontier models, Glean CEO Arvind Jain told CNBC — even as companies like Lindy route bulk workloads to cheaper open-weight alternatives.

Microsoft CEO Satya Nadella warned in a June essay against concentrating AI value in "a few models that eat everything they see." Microsoft, Amazon, and Google are all pushing lower-cost model tiers.

Ramp co-CEO Eric Glyman said most CFOs "not only didn't plan for this in their annual plans — the steep growth — but don't have great tools to manage this." Token spend, he said, is "not a clean area of spend."

What to watch:

• Q2 earnings from hyperscaler-backed enterprises for AI line-item discipline
• Whether OpenAI's GPT-5.6 Terra pricing at half GPT-5.5 rates accelerates the race to the bottom
• DeepSeek V4-Pro adoption by Microsoft Copilot and other Western platforms
• OpenAI and Anthropic S-1 disclosures on customer concentration and churn risk

Convina's view: Tokenmaxxing was the venture-capital version of a sugar high — real productivity gains buried under a metric nobody should have worshipped. The labs that benefited most from the binge are now filing to go public just as their best customers install budget caps and route traffic to Chinese open-weight rivals. Wall Street is still pricing trillion-dollar AI platforms. Enterprise finance departments are pricing $1,500 a month. One of those numbers is going to win.