I saw this reposted elsewhere but thought I'd comment here directly and get your feedback if any. I thought your analysis was VERY GOOD.
* The vehicle was originally designed cooperatively with Waymo back in 2020
* The original hard point vehicle was adapted by Zeekr into the Zeekr MIX in 2023
* The MIX is a very high content vehicle and baselines in CHN @ $38K.
* The decontented RT for Waymo is likely significantly less cost perhaps closer to $32-$33K
* The automated seat rails, leather throughout and lounger seats with massage in the back for example are not included
* Furthermore, because of the Biden-era ruling about Russian and Chinese cars, much of the add-on cameras and other tech are also removed
* Waymo applied for a Commerce Dept exemption from the tariffs as a B2B purchase and a hardship since Jaguar plant was torn down. Who knows how that turns out
* You should probably consider the shipping cost to get it to the US from Ningbo. It could get loaded on shipments for Polestar & Volvo which are also Geely properties
* The Zeekr RT is REPORTEDLY plug-n-play
* Not having to do retrofits for wiring, cabling, power, etcetera is the big Kitting reduction
* The designed in drive by wire makes adaptation to Robotaxi service trivial which is a BIG SAVINGS in integration.
Thanks for your comments! This is really interesting, and implies the actual cost to waymo could be a huge overestimate on my end.
The note on this not actually being based on the Zeekr MIX is something I definitely missed in my analysis. So if they can get their tariff exemption, then that would put it at $70k, and it sounds like I may have overestimated the integration cost by good amount as well.
So the Zeekr became the go-to market vehicle for Waymo. Only they would know when all the other todos would predictably converge. The original LiDAR for example was from Velodyne and were $75K full retail on the early FCA Pacificas. Waymo built their instruments thereafter and supposedly reduced the price to $7500 in R1 & closer to $1500 in r2. The bigger challenge was to create a family design that would work on a Semi-truck which they felt needed 500m field of view instead of the taxi 300m. The 'process' for the Pacificas was horrible and very expensive. The Jaguars were hopefully going to be better with involvement of Magna on the original build in Austria and the Kitting plant in Phoenix. Still a bad experience as the vehicles were far from plug n play. More like an RV conversion. Waymo knew what they wanted and couldn't get a custom build in the US. Geely had EV and taxi experience. The plan was great until Biden ban of communication tech on Russian and Chinese vehicles. He also raised the tariffs. The challenge was the vehicle was already done. I presume Waymo for a custom design had to commit to a minimum. Zeekr, to their credit and ingenuity, repurposed the vehicle to make the MIX. If you look online you will find that the RT and the MIX have the exact same height, wheelbase, length and even retained the novel suspension and the steer-by-wire tech (and dynamic B-pillar). The weird look to the ORIGINAL RT design was the double A-pillars for US crash worthiness.
Now one place where the RT was costlier (at least in theory) was the pathways for the sensor pods to the compute. That should have been built into the stampings with cable routing considered. The pods are crazy large especially on the front for the Zeekrs. The pods are designed for natural ventilation, heating, cooling and power for the cleaning systems. All of that wiring and cabling would be a NIGHTMARE. If you look closely at photos of the vehicles at POE you can see the protruding mounts above the vinyl shipping. The cost and time building the Jaguars is all of the cabling, wiring, power routing, etcetera. Supposedly the design was to include all of the cabling from pod to compute so that termination and testing is easy, fast and reliable. I have no knowledge about the compute though. I would guess water-cooled TPUs if that is possible to make it work. It should not be underestimated that EVERY Western LLM is using NVidia for inference except Alphabet. That is why the per token cost is SO MUCH CHEAPER on Gemini. The TPUs are simply a better mousetrap.
The shipping should not be underestimated. A ro-ro ship can transport a vehicle ~$2K China to POE. Gotta move it to Phoenix, do the integration and then get the vehicles to the deployment cities. It all adds up.
Hah i used to work for zoox. Before amazon. It would be interesting to see that, but i have trouble believing they can overcome the lead that waymo and tesla have!
How are you factoring in tariffs on Chinese electric vehicles on the price of Zeekrs? I’m pretty sure the current tariff rate on Chinese EVs is 100%. Wouldn’t that double the cost of the car
Its a good question, and I'm not sure how to factor this in. They might get an exemption, and some amount of these vehicles were already purchased. I think this is the biggest risk factor though for my estimate being too *low* -- if the actual price is higher, I am going to blame tariffs.
Fair enough. In that case I do think your estimate is too low. However, Zeekr owner Geely owns European brands and may be able to shift production to other countries to avoid the 100% tariff (may raise msrp price though). Or, they could sell the cars at a loss and with tariffs still have a very cheap car. E.g., 10k msrp plus 100% tariff = $20k which is still cheap. China has a track record of selling at a loss to gain market share. Though Geely is publicly traded so that may impact its ability to sell for a loss assuming investors would penalize them for it. Idk.
My only quibble with the piece is that prices might still fall quite a bit, even if Waymo were a monopoly (assuming the true cost per mile is a lot lower). Waymo is also in competition with other forms of transit / no transit. Assuming that customers are fairly cost sensitive and Waymo can create as many vehicles as they need in the long run, you'd expect the profit maximizing price (margin x volume) to be significantly lower than it is for existing rideshare.
Yeah i could definitely see costs ending up a lot lower than this per vehicle. Right now customers aren't overly price sensitive- i think the competition is still Uber, and the driverless experience appears worth a premium to many riders.
I saw this reposted elsewhere but thought I'd comment here directly and get your feedback if any. I thought your analysis was VERY GOOD.
* The vehicle was originally designed cooperatively with Waymo back in 2020
* The original hard point vehicle was adapted by Zeekr into the Zeekr MIX in 2023
* The MIX is a very high content vehicle and baselines in CHN @ $38K.
* The decontented RT for Waymo is likely significantly less cost perhaps closer to $32-$33K
* The automated seat rails, leather throughout and lounger seats with massage in the back for example are not included
* Furthermore, because of the Biden-era ruling about Russian and Chinese cars, much of the add-on cameras and other tech are also removed
* Waymo applied for a Commerce Dept exemption from the tariffs as a B2B purchase and a hardship since Jaguar plant was torn down. Who knows how that turns out
* You should probably consider the shipping cost to get it to the US from Ningbo. It could get loaded on shipments for Polestar & Volvo which are also Geely properties
* The Zeekr RT is REPORTEDLY plug-n-play
* Not having to do retrofits for wiring, cabling, power, etcetera is the big Kitting reduction
* The designed in drive by wire makes adaptation to Robotaxi service trivial which is a BIG SAVINGS in integration.
Thanks for your comments! This is really interesting, and implies the actual cost to waymo could be a huge overestimate on my end.
The note on this not actually being based on the Zeekr MIX is something I definitely missed in my analysis. So if they can get their tariff exemption, then that would put it at $70k, and it sounds like I may have overestimated the integration cost by good amount as well.
Maybe $65k per robotaxi then?
So the Zeekr became the go-to market vehicle for Waymo. Only they would know when all the other todos would predictably converge. The original LiDAR for example was from Velodyne and were $75K full retail on the early FCA Pacificas. Waymo built their instruments thereafter and supposedly reduced the price to $7500 in R1 & closer to $1500 in r2. The bigger challenge was to create a family design that would work on a Semi-truck which they felt needed 500m field of view instead of the taxi 300m. The 'process' for the Pacificas was horrible and very expensive. The Jaguars were hopefully going to be better with involvement of Magna on the original build in Austria and the Kitting plant in Phoenix. Still a bad experience as the vehicles were far from plug n play. More like an RV conversion. Waymo knew what they wanted and couldn't get a custom build in the US. Geely had EV and taxi experience. The plan was great until Biden ban of communication tech on Russian and Chinese vehicles. He also raised the tariffs. The challenge was the vehicle was already done. I presume Waymo for a custom design had to commit to a minimum. Zeekr, to their credit and ingenuity, repurposed the vehicle to make the MIX. If you look online you will find that the RT and the MIX have the exact same height, wheelbase, length and even retained the novel suspension and the steer-by-wire tech (and dynamic B-pillar). The weird look to the ORIGINAL RT design was the double A-pillars for US crash worthiness.
Now one place where the RT was costlier (at least in theory) was the pathways for the sensor pods to the compute. That should have been built into the stampings with cable routing considered. The pods are crazy large especially on the front for the Zeekrs. The pods are designed for natural ventilation, heating, cooling and power for the cleaning systems. All of that wiring and cabling would be a NIGHTMARE. If you look closely at photos of the vehicles at POE you can see the protruding mounts above the vinyl shipping. The cost and time building the Jaguars is all of the cabling, wiring, power routing, etcetera. Supposedly the design was to include all of the cabling from pod to compute so that termination and testing is easy, fast and reliable. I have no knowledge about the compute though. I would guess water-cooled TPUs if that is possible to make it work. It should not be underestimated that EVERY Western LLM is using NVidia for inference except Alphabet. That is why the per token cost is SO MUCH CHEAPER on Gemini. The TPUs are simply a better mousetrap.
The shipping should not be underestimated. A ro-ro ship can transport a vehicle ~$2K China to POE. Gotta move it to Phoenix, do the integration and then get the vehicles to the deployment cities. It all adds up.
Although Waymo and Tesla are first out of the shoot, I predict that Amazon's Zoox will overtake both, once their vehicles start hitting the road.
Hah i used to work for zoox. Before amazon. It would be interesting to see that, but i have trouble believing they can overcome the lead that waymo and tesla have!
How are you factoring in tariffs on Chinese electric vehicles on the price of Zeekrs? I’m pretty sure the current tariff rate on Chinese EVs is 100%. Wouldn’t that double the cost of the car
Its a good question, and I'm not sure how to factor this in. They might get an exemption, and some amount of these vehicles were already purchased. I think this is the biggest risk factor though for my estimate being too *low* -- if the actual price is higher, I am going to blame tariffs.
Fair enough. In that case I do think your estimate is too low. However, Zeekr owner Geely owns European brands and may be able to shift production to other countries to avoid the 100% tariff (may raise msrp price though). Or, they could sell the cars at a loss and with tariffs still have a very cheap car. E.g., 10k msrp plus 100% tariff = $20k which is still cheap. China has a track record of selling at a loss to gain market share. Though Geely is publicly traded so that may impact its ability to sell for a loss assuming investors would penalize them for it. Idk.
My only quibble with the piece is that prices might still fall quite a bit, even if Waymo were a monopoly (assuming the true cost per mile is a lot lower). Waymo is also in competition with other forms of transit / no transit. Assuming that customers are fairly cost sensitive and Waymo can create as many vehicles as they need in the long run, you'd expect the profit maximizing price (margin x volume) to be significantly lower than it is for existing rideshare.
Yeah i could definitely see costs ending up a lot lower than this per vehicle. Right now customers aren't overly price sensitive- i think the competition is still Uber, and the driverless experience appears worth a premium to many riders.
Although Waymo and Tesla are first out of the shoot, I predict that Amazon's Zoox will overtake both, once their vehicles start hitting the road.