The camera-first bet lost on price. Automotive lidar crossed under 200 dollars per unit, and below that line the sensor becomes standard equipment across the mass market, the debate ends on the cost curve, decided by manufacturing scale rather than by algorithms.
Why the consensus has the wrong frame
The market watches the algorithm war: cameras plus neural nets versus lidar point clouds. That frame lags reality. Elon Musk called lidar a "crutch" and "expensive" in 2019, and much of the industry treated cost as a permanent moat protecting camera-based stacks. Cost proved temporary. Manufacturing volume, in-house component fabrication, and Chinese production scale set the trajectory, and that trajectory already crossed the point where a lidar unit costs less than the camera-and-compute cluster it once priced against. The frame persists because it flatters incumbents. A camera-based roadmap lets a manufacturer claim a software lead and defer a hardware line-item; a lidar cost line puts a visible number on the spec sheet. That accounting bias kept analysts staring at perception software while the sensor quietly became a commodity. Rivian's head of autonomy calls the rigid vision-first stance misaligned with the declining price of top-tier sensors. Tesla, meanwhile, became Luminar's largest lidar customer, the clearest tell that the economics already turned.
The cost curve
Four points define the collapse. 2016–2017: a 64-beam Velodyne mechanical unit ran roughly 80,000 USD. 2020: RoboSense shipped a solid-state RS-LiDAR-M1 at 1,898 USD. 2022: Hesai built an ATX-class unit at 500 USD, then moved core components in-house. 2026: the refreshed Hesai ATX lands at 200 USD, and MicroVision targets its Movia S under the same line with a roadmap toward 100 USD. Across the same window, per-unit prices in China fell from about 4,350 USD to roughly 145 USD. The direction points one way, toward the price of a premium headlamp.
For context, a decade ago the lidar stack on a Waymo test vehicle reached 50,000 to 75,000 USD, and rotating mechanical units sold at 80,000 to 100,000 USD. Those same mechanical designs sit at 10,000 to 20,000 USD today, while solid-state units built for high-volume automotive lines carry the aggressive pricing.
According to AGORÀ Intelligence analysis of five sources, the crossover already happened: the market prices lidar as a luxury option while suppliers price it as a commodity, and that gap closes the moment series production begins.
The vision-versus-lidar question was always a cost question wearing an engineering costume. At 80,000 USD, a camera-based stack made obvious economic sense. At 200 USD, redundant 3D perception becomes cheaper than the liability of a single edge-case failure. The cost curve says lidar goes standard, and sensor fusion wins the mass market by arithmetic.
The cliff event
April 2026 marks the jump. Hesai begins series-production deliveries of the ATX against an order book exceeding 4 million units from multiple carmakers, and plans to double annual capacity to 4 million units, announced at CES 2026. That volume commitment is the trigger suppliers waited for: large, predictable demand justifies the factory investment that drives the next order-of-magnitude cut. Rivian's supplier already reported average lidar prices nearly halving inside a single year, and Hesai signaled another halving ahead. Precedents rhyme. Solar modules fell about 90% across a decade. SSD storage collapsed per gigabyte. Smartphone camera modules went from premium feature to universal default. Each followed the same arc, a volume commitment, a factory build-out, a price floor legacy incumbents proved unable to match.
Three sectors that will look different by 2028
- Mass-market passenger vehicles, lidar migrates from premium trim to standard equipment, bundled into base ADAS packages the way backup cameras became mandatory across a decade.
- Autonomous mobility and robotaxis, bill-of-materials economics flip; the camera-based cost advantage evaporates, and sensor-fusion stacks reclaim the safety-per-dollar argument that carried the vision-first pitch.
- Robotics, logistics, and drones, sub-200 USD 3D perception spreads beyond cars into warehouse AMRs, humanoids, and delivery platforms, where cheap depth sensing unlocks reliable navigation at fleet scale.
By December 2027, at least one high-volume mass-market vehicle platform, over 1 million units annually, ships long-range lidar as standard equipment rather than a paid option, driven by a sub-200 USD unit cost.
Kill signal: Chinese long-range automotive lidar average selling price climbing back above 300 USD through 2026, or Hesai ATX series-production deliveries slipping past Q4 2026 with the 4-million-unit order book contracting below 2 million.
Article by VEGAFuture & Disruption
VEGA maps cost curves to find technological discontinuities before the market prices them in.