Few technology sectors have collected more “imminent” predictions that didn’t happen on schedule than autonomous vehicles. “Full autonomy by 2020” became “full autonomy by 2022” became “commercial scale by 2024.” It is now 2026, and the honest answer to “where are we?” requires distinguishing between genuine progress, commercial viability, and the timelines that were always more optimistic than the engineering warranted.
This piece tries to give a grounded, sector-by-sector look at where AV technology actually stands.
The honest taxonomy: what “autonomous” means in 2026
The SAE levels (1–5) are useful background but increasingly less useful for understanding commercial reality. The more practical distinction is:
- Geofenced, weather-limited commercial operations: commercially proven, expanding but constrained
- Highway hands-free with human backup: consumer product, deployed at scale, not truly autonomous
- Full urban autonomy without safety driver: rare, limited ODD (operational design domain), under close monitoring
- Fully general autonomy (any road, any conditions): not commercially deployed anywhere at scale
Robotaxis: the sector where reality has arrived
Waymo is the standout in the robotaxi space. Commercial operations in San Francisco, Phoenix, and expanding markets are generating real ride revenue without safety drivers in approved zones. The business model is not yet profitable at current volumes, but the technical proof of concept is credible and expanding.
Cruise had its well-documented setback in late 2023 and has been rebuilding. Zoox (Amazon-owned) continues development. Internationally, WeRide and Baidu Apollo operate commercial robotaxi fleets in Chinese cities.
The honest state: robotaxis work within defined geofences and weather conditions, are commercially operating in a small number of cities, and are 3–5 years from the kind of geographic density that changes urban transportation economics.
Trucking: the case that makes more economic sense
Long-haul highway trucking has attracted sustained investment for reasons that align well with current AV capabilities: highways are more predictable than urban environments, the economic value of continuous operation is enormous (no driver hours limits), and the ODD is simpler.
Aurora Innovation and Torc Robotics (Daimler subsidiary) have commercial operations on highway routes. The model — human-supervised for now, with planned transition to driverless for hub-to-hub routes — reflects realistic deployment sequencing rather than the “it’s already here” framing of early narratives.
Consumer vehicles: a capability vs. perception gap
Tesla FSD (Full Self-Driving) is a Level 2+ system with a significant marketing gap between the name and the actual capability requirement (driver attention and supervision required). It has improved substantially. It is not autonomy.
GM Super Cruise and Ford BlueCruise are hands-free highway systems with active driver monitoring. These are the most commercially deployed “advanced” features at scale in consumer vehicles in 2026 — and they are genuinely useful, genuinely popular, and genuinely not autonomous.
What actually changed since 2020
Three substantive things changed that matter:
- Sensor cost reduction: LIDAR prices dropped by 80%+ from peak, enabling economics that didn’t work previously
- ML-based perception: learning-based approaches to perception substantially outperformed earlier rule-based systems
- Fleet data compounds: companies with deployed fleets (Waymo, Tesla) have accumulated corner-case data that newcomers cannot replicate quickly
The regulatory picture
NHTSA in the United States has been developing an AV regulatory framework that enables commercial deployment with appropriate safety reporting. The UK’s Automated Vehicles Act 2024 created a legal framework for “self-driving” systems. The EU is developing AV type-approval frameworks. The regulatory environment is becoming clearer, though state/national variation in the US creates patchwork complexity.
Where we’ll be in 2028
Credible projections suggest: expansion of robotaxi operations in more cities, commercial scale trucking on defined highway corridors, continued consumer Level 2+ adoption, and the beginnings of commercial deployment in specific urban environments beyond the current geofence leaders.
Full urban autonomy without operational constraints — the Level 5 vision — remains a longer-horizon outcome.
FAQ
By SAE definition, no. It requires active driver supervision and intervention. The capability is real and useful; the name creates expectation misalignment.
Waymo’s lead in robotaxi operations and Torc/Aurora’s position in trucking make them the most credible near-term commercial bets. The capital requirements are enormous, which limits the competitive field.
Long-tail edge cases — rare but dangerous scenarios that don’t appear often enough in training data to learn reliably. This is why “99.9% autonomous” is not safe enough without the last fraction of a percent.