Uber, Nvidia question AI spending

Published May 27, 2026 by The Daily Scout

- Uber and Nvidia executives said on May 27 that AI spending is getting harder to justify as compute and token bills rise faster than clear returns. - Uber's Andrew Macdonald said “that link is not there yet,” while Nvidia's Bryan Catanzaro said compute costs exceed employee costs on his team. - Uber's AI spending comments came on the Rapid Response podcast; Nvidia's earlier remarks were reported by Axios in April.

Why it matters

Uber and Nvidia executives are now saying in public what many finance and product teams have been arguing in private: the economics of AI are getting harder to defend. On May 27, coverage of remarks from Uber President and COO Andrew Macdonald and Nvidia Vice President of Applied Deep Learning Bryan Catanzaro focused on the same pressure point — rising AI infrastructure costs without a clear line to near-term returns. ### Why are Uber and Nvidia both talking about AI costs now? Andrew Macdonald said on the “Rapid Response” podcast that Uber has struggled to connect rising AI usage to more useful customer features. “That link is not there yet,” NewsNation reported, citing the interview, and Macdonald said AI consumption was becoming “harder to justify.” (ndtvprofit.com) Bryan Catanzaro told Axios in April that, for his Nvidia team, “the cost of compute is far beyond the costs of the employees,” a remark that has since circulated widely because Nvidia sits at the center of the AI hardware boom. That comment did not come from an outside critic or short seller; it came from a senior executive at one of the biggest beneficiaries of AI infrastructure spending. (newsnationnow.com) ### What exactly is Uber saying the money is buying? Uber’s concern appears to center on token consumption and software tooling rather than a single capital project. Multiple reports said the company burned through its planned 2026 AI budget within roughly four months, with spending tied in large part to Claude Code usage. (finance.yahoo.com) Macdonald’s complaint was not that engineers dislike the tools. It was that the company cannot yet point to a proportional increase in shipped consumer features or measurable user benefit. That is a narrower and more concrete test than broad claims about productivity. ### Why does a quote from Nvidia matter so much? (finance.yahoo.com) Nvidia’s business has been built on the surge in demand for AI compute, so Catanzaro’s comment lands differently than similar skepticism from a budget-conscious customer. When a senior Nvidia executive says compute costs on his team exceed labor costs, it suggests the pricing pressure is visible even inside companies most aligned with AI expansion. (finance.yahoo.com) That does not mean Nvidia is pulling back from AI. It means even heavy users are distinguishing between technical capability and economic efficiency. The debate has moved from whether models can perform a task to whether running them at production scale is worth the bill. That framing was reflected in May 27 coverage aggregating comments from Uber, Nvidia and other tech companies. (finance.yahoo.com) ### Is this about a broader pullback across tech? Microsoft, Uber and Meta were all cited in recent coverage about rising AI costs and uneven returns, though the details differ by company. Reports described token pricing, GPU spending and inference costs as the common source of pressure, especially when companies try to move from experiments to broad internal or customer-facing deployment. (ndtvprofit.com) Analysts and commentators quoted in follow-up reports have cast the shift as a move from “can we build this” to “should this live in production at this cost.” The practical effect is that product teams are being pushed to prove retention, efficiency or revenue gains before expanding usage. (cnbctv18.com) ### What should readers watch next? Uber’s next public test will be whether it can show customer-facing gains from the AI tools driving its token bills, after Macdonald’s May comments on “Rapid Response.” Nvidia’s next signal will come from whether executives keep describing compute as a larger internal cost center even as demand for its chips remains strong. (newsnationnow.com) (ndtvprofit.com)

Key numbers

Uber and Nvidia executives said on May 27 that AI spending is getting harder to justify as compute and token bills rise faster than clear returns.
On May 27, coverage of remarks from Uber President and COO Andrew Macdonald and Nvidia Vice President of Applied Deep Learning Bryan Catanzaro focused on the same pressure point — rising AI infrastructure costs without a clear line to near-term returns.
Multiple reports said the company burned through its planned 2026 AI budget within roughly four months, with spending tied in large part to Claude Code usage.
That framing was reflected in May 27 coverage aggregating comments from Uber, Nvidia and other tech companies.

What happens next

On May 27, coverage of remarks from Uber President and COO Andrew Macdonald and Nvidia Vice President of Applied Deep Learning Bryan Catanzaro focused on the same pressure point — rising AI infrastructure costs without a clear line to near-term returns.
That framing was reflected in May 27 coverage aggregating comments from Uber, Nvidia and other tech companies.
(cnbctv18.com) What should readers watch next?

Sources

Quick answers

What happened in Uber, Nvidia question AI spending?

Uber and Nvidia executives said on May 27 that AI spending is getting harder to justify as compute and token bills rise faster than clear returns. Uber's Andrew Macdonald said “that link is not there yet,” while Nvidia's Bryan Catanzaro said compute costs exceed employee costs on his team. Uber's AI spending comments came on the Rapid Response podcast; Nvidia's earlier remarks were reported by Axios in April.

Why does Uber, Nvidia question AI spending matter?

Uber and Nvidia executives are now saying in public what many finance and product teams have been arguing in private: the economics of AI are getting harder to defend. On May 27, coverage of remarks from Uber President and COO Andrew Macdonald and Nvidia Vice President of Applied Deep Learning Bryan Catanzaro focused on the same pressure point — rising AI infrastructure costs without a clear line to near-term returns. Why are Uber and Nvidia both talking about AI costs now? Andrew Macdonald said on the “Rapid Response” podcast that Uber has struggled to connect rising AI usage to more useful customer features. “That link is not there yet,” NewsNation reported, citing the interview, and Macdonald said AI consumption was becoming “harder to justify.” (ndtvprofit.com) Bryan Catanzaro told Axios in April that, for his Nvidia team, “the cost of compute is far beyond the costs of the employees,” a remark that has since circulated widely because Nvidia sits at the center of the AI hardware boom. That comment did not come from an outside critic or short seller; it came from a senior executive at one of the biggest beneficiaries of AI infrastructure spending. (newsnationnow.com) What exactly is Uber saying the money is buying? Uber’s concern appears to center on token consumption and software tooling rather than a single capital project. Multiple reports said the company burned through its planned 2026 AI budget within roughly four months, with spending tied in large part to Claude Code usage. (finance.yahoo.com) Macdonald’s complaint was not that engineers dislike the tools. It was that the company cannot yet point to a proportional increase in shipped consumer features or measurable user benefit. That is a narrower and more concrete test than broad claims about productivity. Why does a quote from Nvidia matter so much? (finance.yahoo.com) Nvidia’s business has been built on the surge in demand for AI compute, so Catanzaro’s comment lands differently than similar skepticism from a budget-conscious customer. When a senior Nvidia executive says compute costs on his team exceed labor costs, it suggests the pricing pressure is visible even inside companies most aligned with AI expansion. (finance.yahoo.com) That does not mean Nvidia is pulling back from AI. It means even heavy users are distinguishing between technical capability and economic efficiency. The debate has moved from whether models can perform a task to whether running them at production scale is worth the bill. That framing was reflected in May 27 coverage aggregating comments from Uber, Nvidia and other tech companies. (finance.yahoo.com) Is this about a broader pullback across tech? Microsoft, Uber and Meta were all cited in recent coverage about rising AI costs and uneven returns, though the details differ by company. Reports described token pricing, GPU spending and inference costs as the common source of pressure, especially when companies try to move from experiments to broad internal or customer-facing deployment. (ndtvprofit.com) Analysts and commentators quoted in follow-up reports have cast the shift as a move from “can we build this” to “should this live in production at this cost.” The practical effect is that product teams are being pushed to prove retention, efficiency or revenue gains before expanding usage. (cnbctv18.com) What should readers watch next? Uber’s next public test will be whether it can show customer-facing gains from the AI tools driving its token bills, after Macdonald’s May comments on “Rapid Response.” Nvidia’s next signal will come from whether executives keep describing compute as a larger internal cost center even as demand for its chips remains strong. (newsnationnow.com) (ndtvprofit.com)