Airbnb, DoorDash and Revolut have little in common on the surface: one rents out spare rooms, one delivers dinner and one looks after people’s money. But all three now use AI models built by Chinese labs, and all three point to cost and performance as the reason.
A growing number of global tech companies are turning to alternative AI platforms for day-to-day operations. Airbnb runs customer service queries through Alibaba’s Qwen, DoorDash handles software reviews with Moonshot AI’s Kimi, and Revolut has migrated its developers to GLM by Z.ai.
Revolut stands out as both the most recent convert and the only fully regulated bank among them. At a conference on 3 November, David Abitbol, who oversees AI operations at the company, said its AI coding bill is now about a tenth of what it paid before.
The Economics Of The Migration
Revolut says about 2,000 engineers now use a self-hosted GLM model as their default coding assistant, after moving over from Claude Code and OpenAI Codex. Abitbol put the blended cost at about 30 cents per million tokens. According to Revolut’s own figures, from a sample of its engineers, coding time fell 60 per cent and feature lead time, from inception to release, dropped 70 per cent.
The financial comparison pits variable commercial API costs against owned infrastructure, meaning it measures operational structure as opposed to pure model quality. Claude Code and Codex are still part of the stack for some tasks, while image and design work goes to other models such as Gemini.
More from News
- Nvidia, Jensen Huang And A Potential Below-The-Belt Security Scandal: Are We Equating Tabloid Fodder With Genuine Corporate Risk?
- Sanctioned At Home: A New Raw Docuseries, Forensically Dissecting The West’s Most Destructive Policy Weapon
- The Democrats Have Taken Congress. What Does This Mean For America’s Startups And SMEs?
- Say No To AI, Lose Google Search – Is That A Fair Choice For Schools To Face?
- Where Does $6.4 Billion In Stolen Crypto Go And What Does It Take To Track Down The Money?
- Experts Comment: Did The Autumn Budget Give UK Startups And Scale-Ups What They Asked For?
- Microsoft Is Reportedly Offering To Pay Staff To Leave While Pouring Billions Into AI – What Does It Say About Big Tech’s Priorities?
- Tairex Launches Multidisciplinary Human-Agent Collaboration Ecosystem To Advance AI Adoption in Real-World Clinical Settings
What Open Weights Make Possible
The key detail here is open weights. The models’ weights are published, so any company can download them and run the model on its own hardware without going through the lab that built it.
Revolut runs its model on about 56 Nvidia B200 chips on US infrastructure. Abitbol framed the main concern as reliance on a China-hosted API, and said self-hosting solves it.
Airbnb chief executive Brian Chesky shared a similar perspective. He emphasised that Airbnb isn’t an Alibaba customer and that Chinese companies don’t receive or access its data.
How Has The US Responded?
Z.ai, also known as Zhipu AI, was added to the US Entity List in January 2025. More recently, House committees opened a joint inquiry on 29 April 2026. Lawmakers wrote to Airbnb and Cursor’s parent company Anysphere about their use of Chinese models, with the Airbnb letter asking for details on its reliance on Alibaba’s Qwen.
DoorDash received a letter of its own on 31 July about its use of Kimi in code review, citing concerns about alleged model distillation, where one model is trained to copy the capabilities of another. So far, the response has been congressional scrutiny, and there’s been no ban on companies using these models.
Why A Bank Adds A Layer
Revolut’s case differs from the other two because of what it is.
The company received its full UK banking licence on 11 March 2026, after a long wait, which means it answers to banking regulators as well as to any lawmakers who take an interest, as they have in Airbnb and DoorDash. UK banks also face rules on operational resilience and on managing the risks of relying on outside suppliers, which gives regulators their own reasons to ask where a bank’s tools come from.
As reported, Revolut uses the model for coding work by its own engineers, with nothing customer-facing involved, which keeps it further from customers than Airbnb’s Qwen-powered service agent. So far, there has been no mention of a letter or regulatory comment aimed at Revolut. That leaves open how a banking regulator might view the choice, since nobody has commented publicly yet.
Cheap Models, Close Attention
The savings Revolut reports are large enough to have drawn in a bank, even as US lawmakers push back. Each company has so far described its choice as one about cost and control, and each has kept other models for the work that needs them.
As businesses continue to benchmark self-managed infrastructure against commercial cloud subscriptions, the scrutiny aimed at Airbnb, DoorDash and Anysphere is likely to follow the companies that adopt them.
