Agree with the rest of your comments; but just a note here. Tesla does have a LOT more compute onboard than legacy automakers, but per Munro Associates teardowns the cost is a wash or even cheaper due to the level of integration compared to what's typical in the industry.
The legacy automaker dependencies on hundreds of OEMs tends to spread the electronics everywhere and that also means duplication in engineering, etc.
FWIW - VW has invested in Rivian to gain access to their infotainment, so they may be the first to "learn" here.
(The Chinese automakers are more like Tesla than the old car companies in these regards too. Much more vertical integration).
P.S. Support on the car side is definitely an opportunity with Tesla but with 40% as many parts as a combustion vehicle there's a lot less to go wrong or maintain. I've had good experiences personally - they came to my house to do the work at no charge for warranty stuff. No slimey dealer salesmen is a big plus. Tesla consumer solar support is terrible though.
For years, the tide as moved back and fourth between the end-all-be-all single processing center and many small processing centers model in OEM vehicle architecture.
Remember, these ECU's are JUST computing, they have specialized I/O, signal conditioning, and communications layers. Having all that hardware just sitting around in units that DON'T have the feature is very expensive. Additionally, putting all the compute in the instrument panel creates long runs of wires to reach the physical sensors and actuators (and power). All this copper costs money as well (as do all the connectors).
I only point this out because those studies you point to often neglect to mention these points and instead focus only on the ECU costs alone .... which is misleading.
Electric vehicles in general have FAR fewer parts than ICE vehicles. It is inevitable that they will become less expensive, and more robust than ICE vehicles. It is only a matter of time.
I think that EV makers that have decided that all the knowledge that ICE OEM's learned over the last 100 or so years should just be thrown out because EV's are all new ...... will soon find themselves utilizing many of those old "outdated" processes as they re-learn why things were done that way in the first place.
Tesla's Megacasting as an example. Yes, it simplifies manufacturing, but it drastically increases repair costs and by extension, insurance cost.
Speaking of insurance, many people believe that the biggest barrier to full autonomy is a technical one. I believe it is a financial architecture issue. Question: Who is liable when a fully autonomous vehicle gets in an accident? How can an insurance company representing an individual insure a vehicle vs an individual's driving record?
These questions will be much more difficult to handle IMO than the technical issues (which aren't to be considered small by any stretch either).