Ghost Autonomy

What did it do?

Ghost Autonomy was a Mountain View-based start-up developing autonomous driving software for carmakers. It first promised a motorway autonomy kit for privately owned cars, then turned to collision-avoidance technology and finally to applying large language models to driving. In 2023 it received investment from the OpenAI Startup Fund.

Why did it shut down?

The consumer kit the company promised for 2020 did not arrive on time, and Ghost changed strategy several times. Its $55 million round in 2023 was a down round. Commercialising autonomous driving required many years and a great deal of capital, while the investment climate had turned harsh. In April 2024 the company, with around 100 employees, ceased operations.

The story

In late 2023, when Ghost Autonomy announced that it had received $5 million from the OpenAI Startup Fund, along with early access to OpenAI systems and Microsoft's Azure resources, it was arguing that the next phase of autonomous driving would come through large language models. Five months later, in April 2024, a short note appeared on the company's website announcing that its worldwide operations had ended.

The company was founded in 2017 as Ghost Locomotion. When it went public in 2019, it had raised a total of $63.7 million from investors including Keith Rabois of Founders Fund, Vinod Khosla of Khosla Ventures and Mike Speiser of Sutter Hill Ventures. The plan was to develop a kit that individuals could fit to their own cars to enable autonomous driving on the motorway. The company said it would deliver this product in 2020.

When the target date came, there was no kit. In 2021, with a $100 million Series D led by Sutter Hill and joined by Founders Fund and Coatue, the company turned towards universal collision-avoidance technology. The approach of co-founder and CEO John Hayes relied on tracking the movement of clusters of pixels in the image rather than first recognising objects. In 2023 came a $55 million down round in which early investors also took part; a filing with the SEC confirms the amount. The same year the company declared its new focus to be applying large language models, able to understand both text and images, to autonomous driving. Experts viewed this approach with scepticism.

At the time of its closure Ghost had around 100 employees and operations in Mountain View, Dallas and Sydney; in total the company had raised around $220 million. The note on its website said that, given the current funding environment and the long-term investment autonomous driving requires, the path to profitability was uncertain. Hayes told TechCrunch that they had completed the motorway product and were moving on to last-mile delivery in cities, but that the years needed to bring the product to market could not be funded.

Ghost's seven years show that in deep-tech start-ups every missed delivery date spawns a new narrative, and each new narrative is harder to fund than the last. Changing technological focus while the valuation falls gives investors the impression of a search, not a roadmap.

Lessons for founders

• In fields with long development cycles such as autonomy, put an interim product that will bring in the first commercial revenue at the front of the roadmap.
• If you miss a publicly announced delivery date, talk openly with investors about what was delayed and why before moving on to a new story.
• Turning to a fashionable technology does not solve a funding problem if it does not shorten the product's time to market.

Sources

1. OpenAI-backed Ghost Autonomy shuts down (TechCrunch, 2024-04-03)
2. Form D/A: Ghost Autonomy Inc. (U.S. Securities and Exchange Commission, 2023)

This section is compiled from public sources (Wikidata, news archives and company announcements). Figures are given as they appear in those sources; unknown fields are left blank. If you spot a mistake in an entry, you can send us a correction.

Is something in this entry wrong? Request a correction → Method