What Is Lidar? — AI Glossary

What it is: Lidar is a sensor that fires laser pulses and measures how long they take to bounce back, building a 3D map of everything around the device. Short for “Light Detection And Ranging.” Used in most self-driving cars, drones, mapping satellites, and increasingly in iPhones.
Who it is for: Anyone trying to understand the Waymo vs Tesla debate, why Chinese robotaxis got cheap so fast, or what an iPhone Pro’s “lidar scanner” actually does.
Best if: You want a plain-English answer for what lidar is and why it matters in 2026.
Skip if: You already work with point clouds professionally. Want one practical AI explainer every morning? Subscribe to our free daily newsletter.

What is Lidar?

Lidar — short for Light Detection And Ranging — is a sensor that fires invisible laser pulses many thousands of times per second and measures exactly how long each pulse takes to bounce back. Multiply that round-trip time by the speed of light, divide by two, and you have a precise distance to whatever the laser hit. Do that across millions of points per second and you get a 3D “point cloud” of the world around the device, accurate to a few centimeters at 100+ meters range.

The simplest way to picture it: like sonar or radar, but using light instead of sound or radio waves. A submarine pings the ocean to find what’s around it; a lidar pings the world. The output is the same idea — distance to objects — just at a much higher resolution.

Why does Lidar matter?

Lidar is the sensor at the center of the biggest unresolved bet in self-driving: does autonomous driving need it, or are cameras alone enough? Waymo, Baidu’s Apollo Go, Pony.ai, and most autonomous trucking companies (Aurora, Kodiak) say yes — their vehicles all use lidar. Tesla famously says no and runs cameras only. As of 2026, the camera-only approach is still under NHTSA investigation for failing in sun glare, fog, and dust — exactly the failure mode lidar is supposed to solve. The companies running real paid robotaxi services at scale all use lidar; the company that doesn’t has done about 1.7 million paid miles total, which Waymo covers in a few days. The data is not yet settled, but it is leaning. (See our Special Report on the state of self-driving in 2026 for the full numbers.)

The other reason lidar matters in 2026 is cost. A 2007 Velodyne lidar cost about $75,000 and sat in a spinning bucket on the roof of a research car. Chinese manufacturers Hesai and RoboSense pushed prices below $140 per unit by 2025, which is why a Chinese robotaxi (the Baidu Apollo Go RT6) now costs roughly $28,000-$30,000 to build — less than half what Waymo’s Jaguar I-Pace cost — and why China is scaling robotaxi deployments faster than the United States.

How does Lidar work?

A lidar unit contains a laser emitter, a high-precision clock, and a detector that catches reflected light. Modern automotive lidars sweep their lasers across a wide field of view either with a fast-spinning mechanical assembly or with a solid-state phased array (no moving parts — cheaper and more reliable at scale). The unit fires pulses on the order of one nanosecond long and times each return to picosecond precision. Each return becomes one 3D point. A few practical implications for beginners:

• Lidar works in the dark. Unlike cameras, it provides its own light. Night driving, tunnels, and underground parking all look the same to lidar.
• Lidar struggles with weather. Heavy rain, snow, and fog scatter the laser pulses. This is why most US robotaxis operate in Sun Belt cities, not Seattle or Buffalo.
• Lidar gives geometry, not color or labels. A point cloud tells you a thing is there, not what it is. Cameras + AI fill in “that thing is a pedestrian” or “that’s a stop sign.” Most autonomous vehicles fuse lidar with cameras and radar — this is what sensor fusion means.
• You probably already own one. iPhone 12 Pro and every Pro model since includes a small lidar scanner. It’s what powers fast autofocus in low light, better AR apps, and improved Photo depth maps.

Waymo’s 6th-generation Driver, deployed on the Zeekr minivan and Hyundai Ioniq 5 robotaxi platforms, uses 4 lidars, 13 cameras, and 6 radars — a 42 percent reduction in sensor count from the 5th generation, which is the engineering signal that lidar costs have collapsed enough to consolidate the sensor stack.

Related terms

• Artificial intelligence
• Machine learning
• Computer vision
• Neural network

Learn more on Beginners in AI

• The Real State of Self-Driving in 2026 — our Special Report with the full data
• Waymo Explained
• Tesla FSD Explained
• Vision-Only vs Sensor Fusion
• Computer Vision in Autonomous Vehicles
• Computer Vision in Modern Drones

Sources and further reading

• NHTSA — Standing General Order on crash reporting (covers autonomous vehicle sensor data)
• Waymo — The 6th-generation Waymo Driver
• Hesai Technology — automotive lidar manufacturer
• Lidar — Grokipedia

Last reviewed: May 2026. Lidar pricing and vendor benchmarks shift fast — the figures above reflect publicly reported numbers as of May 2026.