There is no question that carmakers' initial appetite for driverless cars is driven by market potentials for "cars as a service."
Yes, when I quoted Phil Magney saying, "It is my impression that the Toyota deal is as much about safety as it is about automation," I was referring "safety" for ADAS. Or if I borrow Toyota's own words, Toyota's inteintion to use AI as "co-pilot."
Said this before but when CaaS takes off it will be trouble.
CaaS will penetrate fast in Western Europe due to its cost advantage vs ownership, the US is more complicated but it follows and in a short while CaaS becomes affordable in developing where adoption would explode. That puts a lot of pressure on new car sales and ASPs but what makes it much worse is that the used car market collapses and depreciation is greatly accelerated- maybe 50% lost value in 2 years. The 2-3 CaaS leaders will own the bulk of the global market, that's how marketing works and any car maker that is not one of them is gonna be in real trouble. Tesla is well positioned, unclear if they understand that they need their own fleet to be cost competitive, Google is there too, if they want to do it but maybe they stick with the data business, Apple still has the brand and can even be 2-3 years late because of it. Uber is aware that they can't survive if they aren't a CaaS provider and that leaves little room for car makers with a chance to shine while all other will have to be able to survive a drastic decline in sales and ASPs. Not familiar with Toyota's financials so can't quite say if they can survive.
Are you sure it's about safety? Nvidia advertises Xavier for L3-L4 and on page 2 you write: "The goal is to "enhance the capabilities of Toyota vehicles, enabling them to better understand the massive volume of data generated by sensors on the car, and to handle the broad spectrum of autonomous driving situation," according to Nvidia.". Assuming that by safety you mean driver assist kind of features.
You are somewhat missing the point, both here and in the previous article, taking deep learning to school.
The previous article was in fact about pushing CUDA, hard. CUDA will allow Nvidia to sell hardware even when the hardware is not the best.
With Toyota you got Xavier, 30 TOPS DL, custom ARM cores, 512 Volta cores and a 10 TOPS DL accelerator at 30W. The key point being the Tensor cores (and TensroRT) , present both here and in the absurd 815mm2 GV100. Nvidia is not getting stuck on the GPU being a GPU to allow ASICs to beat it.
The timing is just right too, the deep learning market (all of it not just aurto) is starting to be big enough to justify the investment. AMD was getting close but Volta leaves them behind - granted Vega was delayed so much that it practically lost en entire cycle. AMD's execution is still lacking. Intel is a few years away from their first real salvo and that gives Nvidia time to consolidate their position.
As for Toyota, they are behind in electric, behind in automation, they risk going under before 2025.