Watch at EaseStories. Answers. Perspective.

Technology · NEWS EXPLAINER

Einride will test Nvidia reasoning model in shadow mode before 2027 road evaluation

The freight operator plans to adapt Alpamayo with its own operating data, compare its suggestions with existing driving software and human performance, and keep current safety systems in control.

Original editorial illustration of an unbranded autonomous electric freight truck using sensor paths on a controlled logistics corridor.
Original editorial illustration: Einride plans to evaluate Nvidia Alpamayo inside its autonomous-truck development stack. Watch at Ease

Einride is expanding its Nvidia partnership to evaluate the Alpamayo vision-language-action model for autonomous trucks. The first phase is deliberately limited: the model will run in shadow mode, observing live operations and proposing actions without controlling the vehicle. A live-road evaluation is planned for the second half of 2027.

What the new collaboration adds

Einride said it aims to become the first trucking company to deploy Alpamayo on Nvidia Hyperion. The project extends a September partnership that already covered Hyperion computing, the Halos safety system and Cosmos and Blackwell infrastructure for model training and validation.

The model is designed to reason, not just detect

Alpamayo is a vision-language-action model. According to Einride and Nvidia, it interprets a driving scene, proposes how the vehicle should move and explains its reasoning. Einride plans to adapt the model to its vehicles, routes and operating data rather than use it unchanged.

Shadow mode keeps existing control in charge

During initial tests, Alpamayo will run live but will not control the truck. Its outputs will be compared with Einride’s current driving planner and human-driver performance while the existing safety architecture retains authority. That distinction makes this an evaluation phase, not a commercial autonomous deployment.

Freight data is central to the claimed benefit

Einride plans to post-train and distill Alpamayo using proprietary data from real freight operations. The companies say this could shorten development cycles and reduce the cost of extending autonomous capability to new routes and vehicle types, but those efficiency gains remain goals rather than measured outcomes.

The next hard checkpoint is in 2027

A live-road evaluation is planned for the second half of 2027. Reuters noted that the collaboration follows Einride’s September rollout of a cab-less autonomous truck in regular public-road service in Germany. The new Alpamayo program still has to progress from shadow-mode comparison to validated road testing.

Original editorial illustration of an unbranded freight truck following a solid safety path beside a translucent observational AI path on a test route.
Original editorial illustration: shadow mode lets the AI propose actions without controlling the vehicle. Watch at Ease