Rivian Buys 15-Person Third Dimension to Turn 400,000 Daily Customer Clips Into Simulations

Rivian has acquired Third Dimension, a 15-person U.K. startup that builds 3D environments from images and video, to speed up testing of its self-driving software. Rivian did not disclose a price, according to TechCrunch, which reported the deal on October 22.

The team joins Rivian's autonomy and A.I. group under James Philbin and expands its London hub. Most of the 15 are in the U.K., TechCrunch said.

The data behind the deal: 400,000 sequences a day

Philbin told TechCrunch the customer fleet sends back more than 400,000 sequences per day. The footage comes from tens of thousands of R1T, R1S, delivery van and R2 vehicles whose owners enable external camera sharing.

The number that matters is not 400,000 but how many of them are usable. Most driving is uneventful, and the value of a fleet is the rare scenario. Philbin said Rivian wants to be "taking real scenarios, but making them simulatable," per TechCrunch. Electrek reports that a rare edge case can be recreated in 3D and varied, rather than waiting months for the fleet to capture another instance.

Neither source says how many of the daily sequences are kept, labelled or turned into simulated scenes. The figure counts clips received, not clips used.

Who Third Dimension is

The startup was founded in 2024 by CEO Tolga Kart and Piotr Sokólski. Electrek says Kart spent more than two years at Tesla on Autopilot simulation, after five years as a senior director on Call of Duty at Activision, and later ran engineering at Parallel Domain. Sokólski spent four years at Wayve building photorealistic neural simulators for self-driving cars.

It announced a $7 million seed round on October 8, 2024, led by Felicis, according to Venture Capital Access Online. Its target customers at the time were game studios, film, the U.S. military and autonomous-vehicle companies. TechCrunch lists Abstract Ventures, MVP, Soma Capital and Solari Capital as seed investors, and does not name Felicis; the 2024 announcement names all five.

Kart told TechCrunch, "The team is excited to be able to focus on one problem."

Where the autonomy schedule stands

Milestone	Date as reported	Source
Point-to-point driving for invited Autonomy+ owners	Later in 2026, under NDA	Electrek
Hands-off driving on surface streets and highways, driver supervising	By end of 2026	TechCrunch
"Eyes-off" driving	2027 target; promised for 2026 in January 2025	Electrek
10,000 driverless R2 on Uber in San Francisco and Miami	From 2028	Electrek
Level 4 on Gen 3 hardware	2028	Electrek

Electrek writes that Uber agreed in March to invest up to $1.25 billion in Rivian, with $300 million up front and the rest tied to autonomy milestones. That leaves $950 million conditional, and Electrek says the milestones start with 10,000 driverless R2s on Uber's network from 2028.

1. Paid up front300 $ million
2. Conditional on milestones950 $ million

Source: Electrek, October 23, 2026, accessed 2026-10-23

The Terminal has covered the other Rivian-linked machine bet, Mind Robotics, which hired Doug Field after $1.015 billion in three rounds. Autonomy is the nearer-term one: its software runs on cars already in customers' hands. For a humanoid counterpart, see the Atlas GR3 hand and its 100,000-a-year target.

What simulation does not prove

Electrek's own assessment is that the deal suits a targeted acquisition, and that simulation is a training and testing tool, not proof of real-world performance. Real-world numbers are still needed. Neither source gives miles driven or intervention rates for the point-to-point feature.

Waymo showed its Waymo World Model in February 2026, built on Google DeepMind's Genie 3, and Tesla has shown its own neural world simulator, Electrek reports.

What would change this read is a published measure of how often the simulated scenes predict real failures. Rivian's invited point-to-point testers begin later in 2026.