I don't have numbers but I was thinking to spectacular failures like Google Wave or similar software projects. For sure they were also built by large teams and costed a non trivial amount of money, but this kind of research on autonomous vehicles, it's spawning a much larger amount of time and requires non common experts, there are tons of legal issues related, and you don't stop because people are not using it... it is a technological problem so you spend more and more resources trying to crack it. So I've the feeling that it's significantly more costly than other attempts that Google did in the past, but I've no numbers. I guess the Waymo numbers are simpler to get, because now it's a spin-off (but we should add all the money spent before the spin-off), but the cost of things like Wave are near impossible to obtain, I can imagine.