Autonomous trucking requires fleet and safety capital before labour savings arrive. Amazon's smart glasses can be layered onto an existing delivery network, so operating evidence comes earlier and at a different balance-sheet cost.
“Physical AI” can describe two very different capital structures.
Einride said on September 21 that it will adapt Nvidia's DRIVE Hyperion architecture for heavy-duty autonomous trucking. The company operates hundreds of electric trucks across the U.S., Europe and the Middle East and expects its broader fleet to reach roughly 1,500 to 2,000 vehicles by 2028. It says about 80% of customer demand captured on its platform is suitable for automation in the medium term.
The Nvidia collaboration standardizes part of the compute, sensor and safety architecture. Einride still designs and operates the autonomous-driving system, handles validation and regulatory approval, and has to put vehicles into contracted freight operations before the labour economics show up.
Its accounts make that timing visible. Einride ended June with SEK748 million, about $77 million, of cash. It reported an H1 2026 net loss of SEK1.12 billion, including substantial non-cash and listing-related charges, and is targeting cash-flow breakeven in 2028. Management links the expansion toward 1,500 to 2,000 trucks to conversion of roughly $800 million of potential long-term ARR in joint business plans.
Amazon is buying evidence inside a network it already owns
Amazon plans another $1.9 billion of investment in its Delivery Service Partner programme in 2027. That total covers pay, safety and technology and should not be assigned to smart glasses alone. Within the programme, Amazon expects more than 20,000 Smart Delivery Glasses to be deployed by the end of 2027.
The glasses augment an existing driver, van, route and dispatch system. Amazon can measure navigation time, phone interactions, scan accuracy and safety before it attempts to replace labour. The marginal technology arrives after the logistics network is already funded.
Einride sits further up the capital ladder. Full autonomy can remove a larger recurring labour cost on suitable routes, but the company has to finance vehicles, redundant sensing, compute, remote operations and safety validation before it captures that saving.
The technologies are operationally different, which is the reason the comparison works. Worker augmentation can prove incremental productivity on an installed network. Autonomous freight has to prove that a larger upfront capital stack can be repaid by structurally lower operating cost.
Sources
Einride, “Einride Enters Strategic Collaboration with NVIDIA to Advance its Autonomous Trucking on NVIDIA Hyperion,” September 21, 2026 · Einride, “Einride Reports First Half 2026 Results,” August 18, 2026 · Business Wire, “Amazon Invests Another $1.9 Billion in the Delivery Service Partner Program with New Safety Initiatives and AI-Powered Technology,” September 21, 2026