Tesla Faces Challenges with Its Camera-Based Autonomous Driving Strategy

Waymo co-chief executive Dmitri Dolgov has laid out the most detailed technical case yet against camera-only autonomy, arguing that weak sensing hits a safety ceiling well before it reaches superhuman performance . Speaking at Y Combinator's Startup School in early August, Dolgov never named TSLA, but camera-only is Tesla's entire bet. His specific claims: cameras degrade in darkness, glare, dust, and low contrast, radar penetrates fog, rain, and snow, and a single leaf on a lens can stop a camera-only vehicle.

Tesla removed radar in 2021 and ultrasonic sensors in 2022, committing fully to Tesla Vision. The strategic logic was cost and scale: cameras are cheap enough to ship on every vehicle, which is how Tesla accumulates training data no lidar-equipped fleet can match. Dolgov's argument is not that the approach fails, but that it tops out. In his framing, cameras alone can build a strong driver-assist product yet cannot reach the safety level required for unsupervised robotaxis carrying passengers.

The operational comparison is unflattering at present. Waymo runs roughly half a million driverless rides a week using cameras, lidar, and radar together. Tesla's robotaxi service has logged about 14 crashes across roughly 800,000 miles with a paid safety monitor in the front seat, and operates inside geofenced, speed-limited zones that critics characterize as a demonstration rather than scaled autonomy .

This is a valuation question, not an engineering debate. A meaningful share of Tesla's market capitalization rests on robotaxi economics that require removing the human from the vehicle. Watch three markers: whether Tesla removes safety monitors from any commercial route, the geographic and speed expansion of the Austin service, and whether regulators grant unsupervised operating approvals. Until a monitor comes out, the sensor debate stays unresolved in Waymo's favor.

Related Stocks

• TSLA
• AMZN
• SPCX