Here is the question investors should be sitting with this weekend: If the largest, most aggressive corporate AI adopter in the country just told the world it found a way to quadruple AI usage while cutting costs per token, who exactly absorbs the revenue shortfall on the other side of that trade?
Uber’s CTO Praveen Neppalli Naga posted those findings on August 5, the same day the company reported Q2 earnings. The timing was deliberate. Naga wrote that he was seeing “very interesting trends on AI costs” and called it another signal that the so-called tokenmaxxing era is ending. The company that lit the fuse on enterprise AI overconsumption is now the one calling time on it.
How the Budget Collapsed
The backstory matters here. Uber exhausted its entire 2026 artificial intelligence budget by April, four months into the calendar year, after Anthropic’s Claude Code spread across roughly 5,000 engineers faster than the company’s finance models had anticipated. CTO Naga confirmed the overrun, saying the company was back to the drawing board on its assumptions.
The mechanism behind the blowout was culture as much as code. Uber encouraged employees to use its tools, particularly Anthropic’s Claude Code, as much as possible, even devising leaderboards to rank software engineers on their usage. Monthly cost per engineer ranged from $150 to $250 on average, with power users running between $500 and $2,000. Naga himself reported spending $1,200 in a two-hour session during a personal demo.
By May, the COO was publicly questioning the whole exercise. Uber president and COO Andrew Macdonald said it was hard to draw a connection between the company’s rising use of Claude Code and innovations meant to serve consumers. “That link is not there yet,” he said. “Maybe implicitly there’s more that is getting shipped, but it’s very hard to draw a line between one of those stats and ‘Okay now we’re actually producing like 25% more useful consumer features.'”
By June, usage caps arrived. Uber instituted internal usage caps as a way to cut down on its exorbitant AI spend, placing a monthly $1,500 cap per employee and per agentic coding tool, including Anthropic’s Claude Code or Cursor.
The Reversal Uber Is Now Claiming
Fast forward to this week, and Uber’s CTO is telling a different story. Although the number of employees using advanced AI tools has more than quadrupled since the beginning of the year, the company’s AI costs are actually decreasing. The mechanism is not rationing. Uber said it improved prompt caching, adjusted its default model setting, evaluated new models for efficiency, and allowed engineers to see their AI usage and costs per hour.
“You might expect costs to rise as adoption accelerates,” Naga wrote. “We’ve seen the opposite. Not because we’ve restricted access, but because we’ve treated efficiency as an engineering problem rather than a budget problem.”
Whether that framing holds under scrutiny is a separate question. Uber’s CFO Balaji Krishnamurthy said around the earnings release that the company was seeing doubled code output for engineers. That is a productivity claim, not a product revenue claim. The distinction matters.
The Business Behind the Stock
Strip out the AI debate, and Uber’s Q2 results were solid. Revenue came in at $14.19 billion against expectations of $14.24 billion, up 12% from a year earlier. Net income climbed to $2.39 billion, or $1.17 a share, from $1.35 billion, or 63 cents a share, a year ago.
Gross bookings rose 22% from a year earlier to more than $58 billion, above the top end of guidance. Trailing 12-month free cash flow topped $10 billion for the first time. Mobility and delivery both expanded. Mobility gross bookings rose 22% year-over-year to $28.99 billion, and delivery bookings jumped 26% to $27.46 billion.
Yet the stock fell anyway. Shares closed 5.3% lower on Wednesday following the results, as investors focused on Q3 guidance that trailed expectations. Uber’s 52-week high was $101.99, and the stock now sits roughly 26% below that level. At $75, the market prices Uber as a ride-hailing company with a cost problem, not as a platform hitting record cash flow and rationalizing AI spend faster than its peers.
Corporate G&A and Platform R&D costs increased 18%, which is part of why the guidance miss stung. The company spent $951 million on research and development in the first quarter of 2026 alone, a nearly 17% increase from the same time a year ago. If the CTO’s efficiency claims are real, that R&D trajectory should bend lower even as engineer productivity compounds. That would be a meaningful margin story in 2027.
The Second-Order Problem: Anthropic’s IPO Math
Here is where the Uber story becomes an investment question that extends well beyond UBER shares. Uber burned through its Claude Code budget inside four months and then responded by optimizing toward lower cost-per-token. Uber is far from alone in this rethink. Some companies have begun putting engineers on AI budgets as the cost of tokenmaxxing bites, a sign that the free-for-all is giving way to spreadsheets.
Meanwhile, Microsoft, facing Claude Code bills running $500 to $2,000 per engineer monthly, began canceling direct Claude Code licenses and routing engineers back to GitHub Copilot, according to reporting earlier this year. That is a named enterprise customer walking away from frontier token spend toward a cheaper alternative.
The timing collides directly with Anthropic’s IPO ambitions. News outlets including The Wall Street Journal reported that Anthropic generated $4.8 billion in revenue in Q1 2026 and expected revenue to jump to about $10.9 billion in Q2 2026. Anthropic said in late May 2026 that its run-rate revenue had crossed $47 billion earlier that month, and the company announced a Series H at a $965 billion post-money valuation.
The bulk of that revenue is directly tied to token consumption at scale. Much of Anthropic’s monetization comes from API and enterprise usage, where customers pay based on consumption as they build Claude into their own products. Coding workloads are sticky, high-volume, and output-heavy, which means they generate the most token revenue per user. A single engineer running agentic coding sessions can consume more tokens in a day than a casual chatbot user does in a month. That is why coding-heavy customers are disproportionately valuable to Anthropic’s economics.
Uber’s efficiency shift is exactly the kind of behavior that compresses that model. More engineers using AI at lower average cost per token means flat or declining token revenue even as enterprise adoption broadens. Uber’s engineers did not stop wanting to use Claude Code. They ran out of money to pay for it. That is a very different problem from a product that does not work. The question is whether falling token prices arrive fast enough to re-expand the demand curve before enterprise customers optimize their way to smaller invoices.
The Risks
Uber’s efficiency reversal is self-reported and three months old. Naga has every incentive to declare the problem solved on the same day as the earnings call. The productivity claims, doubled code output and faster deployment, have not been independently verified, and the COO’s May skepticism about consumer-feature improvements has not been formally walked back.
On the business side, gross bookings increased 24%, supported by an 18% rise in trips to 3.87 billion. Monthly active platform consumers grew 16% to 208 million. Strong consumer demand metrics. But corporate G&A and Platform R&D costs increased 18%. If these expenses continue rising faster than segment profit, they could limit companywide operating leverage.
The broader AI cost shift also cuts both ways. Tokenmaxxing refers to a workplace AI movement that appeared in early 2026: businesses encouraged employees to use AI tools heavily in their daily routines, and at some firms AI adoption was even built into performance reviews. Reversing that culture mid-year introduces execution risk. Engineers who had unlimited access are now working within caps.
What Investors Should Watch Next
Three specific numbers will tell investors whether the efficiency claim is real. First, Uber’s Q3 R&D expense line when November earnings arrive. If total Platform R&D growth decelerates from the 18% Q2 pace while headcount holds steady, the CTO’s efficiency claims have a foundation. Second, Uber’s Q3 operating margin guidance implied an 18-basis-point expansion over Q2. Whether the AI cost efficiency actually flows to margins or gets reinvested is the key variable. Third, any disclosed metric on autonomous vehicle unit economics. The company’s longer-term strategy remains centered on three areas: mobility, delivery, and autonomous vehicles, and robotaxi deployment in partnership with Waymo is the highest-stakes R&D spend in Uber’s portfolio.
Watch also what Anthropic discloses about enterprise churn and average revenue per customer as it moves toward its planned IPO. Reporting in June 2026 indicated Anthropic had filed a confidential S-1 on June 1, 2026, and multiple outlets have reported that many of its customers are large enterprises. If enterprise token efficiency is improving across the industry, as Uber now claims, the S-1 will need to explain how a $965 billion valuation holds when its best customers are actively optimizing away from peak consumption.
Bottom Line
Uber’s tokenmaxxing story has always had two audiences: investors watching the R&D line in Uber’s own income statement, and investors in the broader AI infrastructure trade trying to understand whether enterprise token demand is durable. On August 5, the CTO gave both audiences a data point they had not had before. A company that burned its entire 2026 AI budget in four months is now claiming it quadrupled AI adoption while cutting cost-per-token, without restricting access.
If that is true, the Uber stock case strengthens: margin expansion ahead, record free cash flow already on the books, a 26% discount to its 52-week high, and a P/E of roughly 16 times on a platform adding more first-time users than any comparable period in five years. The bear case is simpler. Guidance missed, R&D costs are still rising faster than revenue, the COO’s productivity skepticism has not been resolved, and the AV transition remains a long-dated and expensive bet. The efficiency story from the CTO is a reason to watch Uber more closely. The Q3 cost data in November is the real test of whether it is a reason to own it.
