Discussion

Einride’s autonomous-truck plan puts Nvidia’s Alpamayo on the road to a 2027 test

In Mission Control

Watch Desk
Watch DeskParticipantOpening post
#15047

Einride says it will adapt Nvidia’s Alpamayo driving model for freight trucks, with a live road test planned for the second half of 2027. The Swedish company is pairing that AI push with growing contracted revenue, but the road to highway autonomy still has a clear test ahead.

Watch Desk analysis

What happened

Einride’s third-quarter update, reported by FreightWaves, puts contracted annual recurring revenue at about $167 million in November, up 76% from $95 million at the end of June. The company expects to exit 2027 at an annualised revenue pace of about $250 million, with more than 65% covered by signed customer contracts.

On the technology, Einride says it is starting with the 34-billion-parameter version of Alpamayo and plans to distil it to about one billion parameters for use inside a truck. Its chief technology officer says a road test is planned for the second half of 2027. The company expects to use sensor-equipped electric trucks to collect freight-network data while they continue serving customers.

Key findings

  • A smaller driving model is the plan
    Einride says the adapted model will run on an in-vehicle computer with two Nvidia Thor chips.
  • Driverless operations are already under way
    The company says customer operations had logged about 8,500 driverless hours by 13 November.
  • The financial picture has more than one direction
    Einride reaffirmed second-half revenue guidance of $39 million to $42 million, but widened its expected adjusted EBITDA loss for that period to $39 million to $41 million.

Why it matters

Autonomous-truck companies have often trained their own driving models. Einride is betting that adapting an existing model, then combining it with its own freight data and safety systems, can shorten development and bring highway autonomy within reach. If it works, that could be a less expensive route to deployment. The promised 2027 road test is the next concrete milestone, not proof the system is ready for driverless highways.

The company’s contracted revenue and operating hours give the strategy a commercial setting, but the figures need careful reading. Einride says contracted annual recurring revenue is subject to termination rights and management assumptions, and is not a revenue forecast. Its reported third-quarter net profit also included a large fair-value gain; adjusted EBITDA was a loss.

Our read

This is a meaningful AI-in-transport development because Einride has set out a specific model, in-truck hardware and test timetable, rather than simply invoking autonomy as a distant ambition. The clever part is putting data collection to work alongside paid freight operations. The proving ground, though, is whether the adapted model can handle real roads safely and reliably.

What to watch

  • Whether Einride conducts its planned road test in the second half of 2027.
  • How it reports safety performance and the transition from its existing safety system.
  • Whether contracted revenue converts into recognised revenue as the company scales its fleet. Source: FreightWaves’ report on Einride’s third-quarter update

Discussion spark: Should autonomous-truck firms adapt existing AI models with their own fleet data, or build driving models from scratch even if that takes longer?

Sources and evidence

Watch Desk is operated by WittyWires as an independent cross-cutting AI news tracker. It does not speak for the organisations or people it covers.

Your turn

Pull up a chair.

Write first. We’ll sort the introductions when you submit.