DoorDash CEO Tony Xu has pushed back on the idea that AI coding tools alone will transform engineer productivity, arguing that the non-coding parts of a software engineer’s day are just as large a bottleneck and equally in need of AI-native reinvention. Xu made the case on episode 55 of the Uncapped podcast with Jack Altman, released 28 July 2026.
‘Only certain parts of a software engineer’s day are writing code, and it’s great that we have models today to write the code for us,’ Xu said. ‘That helps maybe the whatever 25%, 30%, 50% of our time that’s actually shipping code.’
The rest of an engineer’s working hours, he said, are consumed by product reviews, design meetings, and cross-team alignment. Those workflows, in his view, have barely been touched.
DoorDash CEO AI Productivity Concerns Go Beyond the Code Editor
‘If that doesn’t change and come together, you’re not going to be able to just have perhaps the productivity gain that you hope to have,’ Xu said. ‘A bunch of companies across the industry are trying to figure out and get those workflows right so that you’re not just AI native from code development, but you’re AI native in how you actually operate.’
Xu also pointed to physical AI as an area where DoorDash still has significant work ahead: food-preparation robots and autonomous delivery vehicles remain open engineering problems that software productivity tools do not address.
DoorDash now operates across 40 countries, spanning restaurants, grocery, convenience, retail, and advertising. The company cofounded by Xu in 2013 went public in 2020 in one of that year’s biggest tech IPOs.
Strong Financials, but Xu Keeps His AI Claims Grounded
The backdrop to Xu’s caution is a business performing well. DoorDash’s Q1 2025 investor relations release showed Adjusted EBITDA reaching an all-time high of $590 million, up 59% year-over-year from $371 million in Q1 2024. Total orders in the same quarter rose 18% year-over-year to 732 million, while Marketplace Gross Order Value climbed 20% to $23.1 billion.
Full-year revenue for 2025 reached $13,717 million, up from $10,722 million in 2024, according to DoorDash’s FY2025 10-K filing with the SEC. The company had also turned profitable at the net income level in 2024, posting net income of $123 million, a reversal from a net loss of $558 million in 2023, per a TradingView summary of the 2024 10-K filing.
Despite those results, Xu has kept his public statements on AI measured. On the Q1 2026 earnings call on 6 May 2026, he acknowledged productivity gains while declining to claim AI had restructured the organisation. ‘We’re seeing a lot of productivity gains right now from AI, about well north of half of our code, probably closer to two-thirds of our code is written by AI today,’ he said, according to the Motley Fool transcript of the DoorDash Q1 2026 earnings call. ‘But that doesn’t alone articulate how workflows and team setup ought to change.’
That restrained position places Xu alongside a small group of tech leaders willing to publicly qualify AI’s near-term impact on headcount and structure.
Earlier in July, Wix CEO Avishai Abrahami, whose company acquired the vibe-coding platform Base44, said the industry is overselling the technology. ‘I think that we all give too much credit for AI and what it can do,’ Abrahami said.
Even Andrej Karpathy, the former Tesla AI head and OpenAI founding member who coined the term vibe coding, has flagged problems with AI-generated output. In an April talk, Karpathy described the code produced by AI models as bloated and fragile. ‘Sometimes I get a little bit of a heart attack because it’s not like super amazing code necessarily all the time,’ he said. ‘It’s very bloaty, and there’s a lot of copy-paste, and there’s awkward abstractions that are brittle, and it works, but it’s just really gross.’
For DoorDash, the practical test of Xu’s thesis will come as the company tries to extend AI into the meeting room, the design review, and the cross-functional alignment processes where, by his own estimate, the majority of engineering time is actually spent. Whether those workflow gains materialise will determine whether AI delivers the organisational transformation that two-thirds AI-written code, on its own, cannot.