OpenAI buys Odessia, a one-year-old AI travel startup

A one-year-old AI travel startup was bought by OpenAI and shut down the same day. The startup is Odessia. It had raised a $6 million pre-seed from Sequoia. Its AI travel agent had been live for about five months. TechCrunch reports the sale. OpenAI has not confirmed it.

What Odessia was

Odessia let a traveller plan a trip through an AI chat. Founder Francis Davidson said he built the agent by directing teams of AI agents. He said this on social media. The home page now carries one line: "OpenAI has acquired Odessia." A second line says the service is no longer available. Anyone with a booking is sent to support@odessia.com. The page shows no shutdown date.

Who the founder is

Francis Davidson started Sonder in 2014, while in college. Sonder rented serviced apartments by the night. It was like Airbnb for upscale stays. It became a unicorn. Revenue reached about $143 million. The company went public through a SPAC. The pandemic hurt it badly. A deal with Marriott did not save it. Davidson stepped down as chief executive in June 2025. The company went bankrupt. He started Odessia after that.

Davidson said on social media that he is now at OpenAI. He said he will work on the future of travel. TechCrunch, which broke the story, calls the deal likely an acqui-hire. The price was not disclosed. OpenAI declined to comment.

Why OpenAI is buying a travel agent

OpenAI had already named travel as a job for a ChatGPT agent. It gave that answer in July 2025. It has partnerships with Booking.com, Expedia and Kayak for the booking step. What it did not have was an AI travel agent built as its own product. Buying one is faster than starting from zero. It is also picking up the founder.

What it means for anyone building an AI assistant

The pattern matters past travel. A lab picks a category. A small team that sold a product in that category gets one buyer. Odessia had a Sequoia round. It had a founder with a public track record. It had a product in the market. It still chose this exit at one year old. A smaller team on the same path will find the same door open. No other door will be as wide.

If you build a consumer AI assistant on top of a lab's model, that lab is now a plausible competitor. It is also a plausible buyer. Travel is the first sign in this quarter. The next category a lab picks will likely produce the same shape of deal. Teams building a general chat product should pick: get bought by a lab, get bought by a rival, or pick a category a general assistant cannot cover. A workflow that needs a human expert next to the model is one such category. A general chat is not a substitute for it.

Source

Reported by TechCrunch. Odessia's own announcement is at odessia.com. The Information also covered the deal in a briefing.