Einride to adapt Nvidia's Alpamayo AI model for autonomous freight

Einride will test Nvidia's Alpamayo, a reasoning-based AI driving model, inside its autonomous driving system, with a live road evaluation planned for the second half of 2027. The Swedish company announced the move Wednesday, saying it will adapt the model to its own vehicles, routes and operating data. Neither Einride nor Nvidia disclosed financial terms.

The project extends a collaboration announced Sept. 21, under which Einride will build its next-generation driving system on Nvidia's Hyperion platform. Einride said adapting Alpamayo is designed to shorten development cycles and reduce the cost of extending the Einride Driver — its autonomous driving stack — to new vehicle types and routes.

The company tied the decision directly to its operational track record. Einride had logged more than 5,400 driverless hours in contracted customer operations across six autonomous deployments in the U.S. and Europe as of June 30, according to its first-half 2026 results.

"Our advantage has always been running autonomous freight in the real world, for real customers," said Henrik Green, Einride's chief technology officer.

What does Alpamayo actually do?

Nvidia introduced the Alpamayo family at CES in January. The company described Alpamayo 1 as a 10-billion-parameter vision language action model that uses video input to generate driving trajectories with reasoning traces.

The Alpamayo models "serve as large-scale teacher models" that developers fine-tune and distill into smaller models for their own driving systems, Nvidia said. In May, the chipmaker launched Alpamayo 2 Super, a 34-billion-parameter teacher model built to be distilled into compact models that run on DRIVE AGX Thor, Hyperion's in-vehicle computer. Einride's release does not say which Alpamayo model it will use.

Einride will post-train and distill the model on its proprietary operational data. The initial testing will run in shadow mode: the model operates live without controlling the vehicle, and its outputs are compared against Einride's driving planner and human driver performance while existing safety systems stay in full command.

Who else is adapting the model for trucks?

Einride is not alone in betting on Alpamayo. PlusAI said in March it is adapting the model for heavy-duty trucking on Level 4 trucks built on Nvidia's DRIVE AGX Hyperion platform. PlusAI expects fully driverless commercial operations on those factory-built trucks to begin in 2027.

Einride nonetheless plans to be the first trucking company to deploy Alpamayo on Hyperion. That claim turns on what counts as deployment. Green framed the company's position in narrower terms.

"Being the first trucking company in this engagement gives us the opportunity to help shape the model for freight," he said.

Nvidia, for its part, framed the work as exploratory rather than a production commitment. Rishi Dhall, Nvidia's vice president of automotive, said the partnership combines Alpamayo and Nvidia's Halos safety system with Einride's real-world freight data and safety engineering on Hyperion to explore how reasoning-based AI can help trucks handle complex driving conditions and accelerate development for new routes and vehicle types.

What comes next on hardware?

New vehicle platforms are already on Einride's schedule. The company, truck manufacturer DAF and research organization TNO are defining and testing how to integrate Einride's autonomous system into DAF's electric truck platform through 2026. Software integration and commissioning are scheduled for 2027 — the same year as the Alpamayo road evaluation.

The commercial logic for shippers and carriers is cost. Einride is betting that an AI model trained on its own freight data will cut the expense of porting the Einride Driver onto new trucks and routes, which has historically been one of the slowest and most expensive steps in autonomous freight deployment. Until the second half of 2027, the model will run only in shadow mode and will not control any vehicle.