Waymo car cost estimate from ChatGPT

Breaking down the estimated cost of a Waymo autonomous car (approximately $250,000), the majority of the expense comes from its sensors, computing power, and custom modifications. Here's a rough breakdown based on available data and common pricing for autonomous vehicle components:

  1. LiDAR Sensors: Waymo's 6th-gen cars use 4 LiDAR sensors. Depending on the type (solid-state or mechanical), these could cost between $1,000 and $75,000 per unit. Assuming mid-range solid-state LiDAR sensors, the total cost for all 4 could range from $4,000 to $20,000.

  2. Cameras: Waymo’s 6th-gen system has 13 cameras. Automotive-grade cameras typically range from $100 to $500 per unit. The total cost for cameras would likely be $1,300 to $6,500.

  3. Radar Sensors: There are 6 radar sensors on each vehicle. High-quality automotive radar sensors range from $100 to $1,000 each. For 6 units, the total would range between $600 and $6,000.

  4. Computing Systems: The onboard computing system handling AI and sensor data processing is a significant expense. A high-performance computer system for autonomous vehicles, including GPUs or specialized chips, could cost around $10,000 to $20,000.

  5. Vehicle Platform: The base vehicle (like the Jaguar I-Pace or Zeekr EV) itself costs about $70,000 to $100,000, depending on the model.

  6. Development and Integration: Beyond hardware, much of the cost includes software development, sensor integration, testing, and custom modifications to the vehicle platform. These development costs, along with maintenance and operational costs, account for a significant portion of the overall price, estimated to be around $50,000 to $100,000.

Estimated Breakdown:

  • LiDAR Sensors: $4,000 to $20,000
  • Cameras: $1,300 to $6,500
  • Radar Sensors: $600 to $6,000
  • Computing Systems: $10,000 to $20,000
  • Vehicle Base: $70,000 to $100,000
  • Development and Integration: $50,000 to $100,000

This results in a total cost ranging from $136,900 to $252,500 depending on the specific components used

所有跟帖: 

其中激光雷达成本比例并不高,大规模生产成本会大幅度下降 -study169- 给 study169 发送悄悄话 (0 bytes) () 10/16/2024 postreply 08:43:20

3 -Hongmei20- 给 Hongmei20 发送悄悄话 (0 bytes) () 10/16/2024 postreply 08:56:41

3 -Hongmei20- 给 Hongmei20 发送悄悄话 (0 bytes) () 10/16/2024 postreply 09:00:33

没看懂你想说啥。小米汽车用了3颗毫米波雷达、12颗超声波雷达、1颗激光雷达。 -lanyin0314- 给 lanyin0314 发送悄悄话 (0 bytes) () 10/16/2024 postreply 09:21:05

三心网友说waymo car cost的问题,就了解了一下。感觉cost上规模以后可能不是问题。小米车自动驾驶应该不如 -study169- 给 study169 发送悄悄话 (14 bytes) () 10/16/2024 postreply 09:24:58

我跟你的观点一样,雷达并不贵。 -lanyin0314- 给 lanyin0314 发送悄悄话 (0 bytes) () 10/16/2024 postreply 09:28:50

和洛基亚不搞SMART PHONE一个道理, 我有朋友从洛基亚跳到apple, 他说其实这些都有POC, 但因为内部原因 -testmobile- 给 testmobile 发送悄悄话 (0 bytes) () 10/16/2024 postreply 09:58:04

和GOOGLE打不过openAI 一个道理,AI的技术能力google第一,但因为产品方向问题,所以搞不出CHATGPT -testmobile- 给 testmobile 发送悄悄话 (0 bytes) () 10/16/2024 postreply 10:02:07

有scaling cost 的 estimate 吗? 可否share 一 个 link? -三心三意- 给 三心三意 发送悄悄话 (86 bytes) () 10/16/2024 postreply 09:30:19

我做投资只从工程师的角度看技术前景,做不到你那样详细的分析,抱歉! -study169- 给 study169 发送悄悄话 (0 bytes) () 10/16/2024 postreply 09:33:38

Not to mention W has to map out hi-def image for the route -三心三意- 给 三心三意 发送悄悄话 (71 bytes) () 10/16/2024 postreply 08:43:24

不能拿L3及以下技术成本跟L4/L5比 -study169- 给 study169 发送悄悄话 (0 bytes) () 10/16/2024 postreply 08:45:43

有道理。如果特斯拉不能突破L4,股价50以下。如果能,5年后1000 -三心三意- 给 三心三意 发送悄悄话 (0 bytes) () 10/16/2024 postreply 08:56:00

一龙已经铁了心要走纯视觉道路,所以成本不会变太大,但必须技术突破 -三心三意- 给 三心三意 发送悄悄话 (0 bytes) () 10/16/2024 postreply 09:00:24

特斯拉的机器人其实更有希望早日突破,可以是限制场景和用途。以一龙的执行力,能跟他竞争的不多。 -study169- 给 study169 发送悄悄话 (0 bytes) () 10/16/2024 postreply 09:00:25

我一直怀疑这个雷达需要hi-def地图是Google自己加的护城河,地图是导航用的,GPS上的就够,雷达只要能识别其他车 -testmobile- 给 testmobile 发送悄悄话 (0 bytes) () 10/16/2024 postreply 08:51:32

纯视觉FSD的上限到什么位置在到达之前谁也不知道 -study169- 给 study169 发送悄悄话 (275 bytes) () 10/16/2024 postreply 08:58:21

3 -Hongmei20- 给 Hongmei20 发送悄悄话 (0 bytes) () 10/16/2024 postreply 09:06:28

3 -Hongmei20- 给 Hongmei20 发送悄悄话 (0 bytes) () 10/16/2024 postreply 09:08:11

请您先登陆,再发跟帖!