[tristanj]

We're looking at different metrics, you're analyzing the average total cost, while I'm analyzing the marginal cost. Waymo has enormous fixed costs like you mentioned, mapping cities and paying engineers are not cheap, which need to be amortized over a massive self-driving fleet. But those are fixed costs which don't increase with fleet size. Waymo currently operates only ~3000 vehicles, which is not enough to amortize those fixed costs into overall profitability.

What matters most are marginal costs (i.e. how much does it cost for Waymo to add 1 more ride). Looking at marginal costs, Waymo takes in more money than it spends on each ride, so projecting outwards when Waymo operates a large enough fleet, Waymo will be profitable.

Uber/Lyft run enormous fleets of ~2 million vehicles in the US, and that's how they are able to maintain profitability. They can spread their engineering and management expenses over millions of rides.

---

Doing my own math, the marginal costs for Waymo are:

Revenue: Each Waymo vehicle brings in ~$50/hour

Expenses: Waymo must pay for

* Assume the cost of a vehicle is $100k

* Amortized depreciation of the car (assume vehicles need to be fully replaced after ~250,000 miles, vehicles average 25 miles / hour, vehicles need to be fully replaced after 10,000 hours, cost is $10/hour)

* Maintenance (Assume the total cost of maintenance is an additional 25% of the vehicle price, vehicle price is $100k and vehicle lasts 10,000 hours, cost is $2.5/hour). This is likely an underestimate, I didn't model the cost of a mechanic, so this could be as high as $5-7/hour.

* Support (assume 1 support agent can support 10 vehicles, Philippine support agent costs $10/hour, so amortized $1/hour per vehicle)

* Cleaning (needed daily, costs $1/hour per vehicle)

* Datacenter compute for vehicle coordination ($0.50/hour per vehicle)

* Electricity (Assume $2/hour)

10 + 2.5 + 1 + 1 + 0.5 + 2 = $17/hour to operate a Waymo.

In conclusion, the marginal costs for Waymo is very profitable.

[ggreer]

Even when just looking at marginal costs, I doubt Waymo is half that of a human driven vehicle. If we assume a robotaxi lasts 200k miles before being retired, then the cost of the vehicle alone ($75k) is 37.5 cents per mile. If a vehicle drives 200 miles a day, that's $5 of electricity (250Wh/mile x 10 cents/kWh), maybe $15 of labor to clean, and the space to park it near downtown ($3/day). That's another 12 cents per mile for a running total of 50 cents per mile. Then factor in maintenance (tires, brakes, suspension, etc) and you're probably close to $1/mile. Then you also need support staff, remote operators (approximately 1 per 50 vehicles, but paid significantly more than Uber drivers), and plenty of compute and storage for the high resolution maps (which must be constantly updated as the environment changes). And none of that includes the R&D costs to improve the vehicles or the self-driving software. Yes many of these costs decrease as fleet size increases, but it'll be a while before it gets below $1/mile. (Nationwide, Uber's rates are $1-2/mile depending on the area.)

There are other considerations as well. For example, available ride shares can scale up/down with demand, while Waymo & competitors will need lots of spare vehicles to satisfy peak demand.

I'm certain autonomous vehicles will eat up the market currently held by Uber/Lyft/Taxis. It's just going to take longer than a lot of people expect.

[tristanj]

Those are some good points. Waymo & self-driving cars only makes sense in markets where the cost of paying a human driver is much higher than the amortized cost of operating a self-driving vehicle. As an extreme example, in Bangladesh a driver will work for ~$0.25/hour. Self-driving cars don't make economical sense there, the additional cost of making a vehicle self-driving never pays off vs having a dedicated driver.

Waymo can expand easily in markets like SF and NYC, where drivers are guaranteed a minimum pay rate of $22+ per hour, but will make less and less economic sense in cheaper labor markets.

[anvuong]

I believe sooner or later Waymo will start selling their tech package to car companies, who in turn will pass the cost on to customers either as a huge mark up or subscription services. I don't believe Waymo can survive on self-driving taxi alone, the hardware cost is too much to go global.