I just started in a new Automotive company and they are playing to build Zonal Ecu without Autosar. It is something like https://www.nxp.com/applications/automotive/software-defined-vehicle/automotive-zone-controller-:AUTOMOTIVE-ZONE-CONTROLLER
It looks like they never build SW for an Ecu until now. The reason they don't want to use Autosar is is because it is to expensive.
It is even possible to build an Ecu without Autosar now if you are not an OEM? It is possible I would say but the resources that you need are very very high. The route they plan to go is trying to find as many as possible open source libraries.....
To your question: Yes, it is possible to build an ECU without Autosar. Indeed, there are some ECUs - either on the cpu-intensive side (like modern high-end HMIs) or on the lowest end (too small for the Autosar footprint) - which don't use Autosar. But...
Autosar is a proven standard which is used in more than 70% of the automotive projects worldwide. Some OEMs enforce the use of Autosar by the suppliers, while others are more agnostic. But all of them will require some kind of proof that your delivered software is developed with state-of-the-art techniques in terms of functional safety and cybersecurity, along with the corresponding certifications. And there, my friend, is where the price tag of Autosar starts looking cheap.
For Zonal Controllers, the trend is to mix different OS/middleware, each tailored for specific use cases. This is possible thanks to the automotive multicore architectures that are already available. For example 4 performance cores with FreeRTOS for QM workloads + 2 safety cores with Autosar for ASIL-x workloads.
Developing Automotive ECUs is not simple and it requires a certain sales volume to make a positive business case. But trying to cut down costs by changing Autosar for a dozen of open source projects that have been developed mostly for other domains is just naive. Good luck with the development and certification of your Frankenstein.