My personal pet peeve theory is that automotive engineering missed the memo that the hard part of software based systems is integration and remote dependencies.

My personal pet peeve theory is that automotive engineering missed the memo that the hard part of software based systems is integration and remote dependencies. All their engineering and supply structures have evolved from mass producing mechanical marvels which happen to have a few electrical and electronic E/E components, marvels and components which are hard to manufacture but simple to integrate ("plug together", essentially). At the core, their thinking goes "Creating components is expensive. Putting components together is cheap. Also minimize prototype scrap with extensive planning and scrupulous list checking processes. If parts don't work together you didn't plan well enough." And that leads to a culture that is the exact opposite of CI/CD, fail early, fail often, rebase or reintegrate continuously. Thus in the automotive SPICE and six sigma world it takes 2-3 years to replace an ECU with another one with sufficiently similar specs. Also, they prioritize "economies of scale" over flexibility and time to market. For example a "gateway controller", a router and switch, routing and switching between ethernet and CAN, with some spare CPU cycles for centralized functions? 2-3 years. Now let's say you need 3 million of these devices per year (Volkswagen group, Toyota, Stellantis each make ~10 million vehicles y/y), and you'll produce cars with that same device for 5 years? Then cost savings of 10$ per device are equal to 150 Mio bucks. As a supplier you'll happily put 20 people on saving those 10 bucks on the "bill of materials" --- not adding cool features, no, saving some RAM, CPU, ... the boring and non-innovative side of "economies of scale" Now the device is cheaper but maxed out. So your OEM won't ship new functions. the OEM is also later to market. And the OEM can't replace the part as easily because the spare would have to be as damn dirt cheap as the original. Not alone would you have to redo all these micro optimizations --- the savings from the previous design wouldn't materialize against the lower volume any more. And that is, in a nutshell, imnsho, why automotive asks for more of the old parts instead of switching to a new board design with a next-gen CPU... To summarize: automotive deals time to market and flexibility for "economies of scale" and "automotive grade quality processes". Contrast that to an OEM who thinks like a software CEO... the software CEO understands that the secret to high quality is in CI/CD with outstanding test and validation pipelines. You seek to change any part of your system just at marginal cost. And then you can replace some ECU within 6 months.