Autonomous vehicles were supposed to be ubiquitous by now. Predictions from 2015-2018 suggested we’d have robotaxis everywhere by 2020. We don’t.

But dismissing autonomous vehicles as vaporware is also wrong. Real progress is happening—just slower and more constrained than originally promised.

The Current State

What actually exists in 2025:

Waymo: Operating commercial robotaxi services in Phoenix, San Francisco, and Los Angeles. Millions of miles driven. Real customers paying fares.

Cruise: Resumed operations after 2023 pause. More cautious expansion. Still proving commercial viability.

Tesla FSD: Supervised self-driving requiring driver attention. Wide deployment but not truly autonomous.

Chinese players: Baidu Apollo, Pony.ai, and others operating services in Chinese cities.

Trucking: Aurora, Kodiak, and others piloting autonomous trucks on specific routes.

Autonomous vehicles are real but limited in scope and geography.

What Works

Where autonomous vehicles succeed:

Geofenced areas: Controlled environments with mapped streets and predictable conditions.

Specific routes: Fixed paths that can be thoroughly tested and optimized.

Good conditions: Clear weather, good visibility, maintained roads.

Slow speeds: Lower speeds reduce decision time requirements and crash severity.

Low complexity: Environments with few edge cases and unusual situations.

This is why robotaxis work in Phoenix (sunny, grid streets) better than Boston (snow, confusing intersections).

What Doesn’t Work Yet

Remaining challenges:

Edge cases: Unusual situations that training data doesn’t cover adequately.

Weather: Rain, snow, and fog remain challenging. Sensor limitations in adverse conditions.

Construction zones: Temporary changes that aren’t in maps.

Human interaction: Communicating with pedestrians, cyclists, and other drivers.

Remote areas: Anywhere without detailed mapping and infrastructure.

Cost: Current systems too expensive for most use cases.

The “long tail” of driving situations is genuinely hard.

The Technology Evolution

How AV technology has progressed:

Sensors: Better cameras, lidar, radar. Lower costs. Higher resolution.

Compute: More powerful onboard processing. Faster decisions.

AI/ML: Better perception, prediction, and planning algorithms.

Simulation: Vastly more simulated miles training and testing systems.

Maps: More detailed, more frequently updated mapping.

Connectivity: V2X communication supplementing onboard sensing.

The technology is much better than five years ago—but the goal is also harder than expected.

Business Models

Where commercial viability might emerge:

Robotaxis in select markets: Waymo has a working business in limited geography. Expansion is the question.

Long-haul trucking: Highways are simpler than cities. Driver shortage creates economic pull.

Delivery: Lower stakes, lower speeds, can pause if uncertain.

Controlled environments: Airports, ports, campuses, industrial sites.

Luxury feature: Premium pricing for advanced driver assistance.

Full L5 autonomy everywhere remains distant; commercial viability in niches is nearer.

Investment Implications

For investors and strategists:

Longer timelines: AV investing requires patience. Most early bets lost money.

Consolidation: Many AV startups have failed or will fail. Winners take most.

Supplier opportunities: Components, software, and services for OEMs.

Adjacent technologies: Sensors, compute, simulation tools with broader markets.

Regulatory awareness: Policy will shape where and how AVs can operate.

Expert guidance: Working with an AI consultancy like Team400 that understands the technology trajectory can help assess investment timing.

The ADAS Path

Most vehicles are taking a different path:

Advanced driver assistance (ADAS): Features like lane keeping, adaptive cruise control, emergency braking.

Gradual improvement: Each generation more capable than the last.

Human in the loop: Driver responsible, technology assisting.

Volume economics: Every car can have ADAS; only a few can be robotaxis.

This “evolutionary” approach may ultimately deliver more value than the “revolutionary” robotaxi path.

What’s Coming

AV evolution ahead:

Geographic expansion: Robotaxis in more cities, though slowly.

Trucking commercialization: Autonomous trucks on specific highway routes.

ADAS improvement: More capable driver assistance in consumer vehicles.

Regulatory development: Clearer rules for AV deployment.

Cost reduction: Technology becoming more affordable over time.

New entrants: Apple, Xiaomi, and others with AV ambitions.

The Bottom Line

Autonomous vehicles are real but limited. The technology works in constrained conditions but not universally.

The pattern is familiar: initial hype, disappointment, and then gradual progress toward real capability. We’re in the “gradual progress” phase.

Full autonomy everywhere remains distant. Useful autonomy in specific contexts is here or approaching.

---

Tracking the actual state of autonomous vehicle technology and deployment.