edit: It's not even accurate to describe them as separate computers, because parts have travelled between them more than once.
And this kind of device, while it may not be interesting for many consumers, will be very interested for industrial, enterprise, or medical users.
To put it another way, if you were running a medical devices company with limited resources, would you bet on this or bluetooth?
I'm hoping that project Ara will become the perfect test bed for sensor module development. Useful after the initial validation stage of experimental stuff on a breadboard but before the stage where you have established enough customers that you can afford to develop a completely integrated and standalone IOT device.
Then they'll be leaving room for a newcomer to do it.
There is a crossover point wherein overhead reduction is "good enough" and when other factors about the phones begin to dominate.
The iPhone doesn't sell because of a non-removable battery, and the huge number of users who use bulky cases to protect fragile phones, and add-on battery cases, which increase weight and spoil the industrial design, tend to support the case that some people are willing to trade off thinness, weight, or cost for other capabilities.
I don't even think this is controversial. If you look at the number of users asking for microSD whenever a device removes it, or the people wanting additional clip-on lenses, there's a hunger to extend the phone platform.
Now, whether that happens over an integrated UniPro bus, an external lighting cable + case, or bluetooth, is an implementation detail of convenience, ergonomics, and bulk. I personally would rather have upgradeable camera modules or additional battery slots than "Morphie" cases and weird protruding clip ons that weren't designed to work with the builtin optics.
Taken from their developer page (https://atap.google.com/ara/):
More Space. New Ideas. The Ara frame contains the CPU, GPU, antennas, sensors, battery and display, freeing up more room for hardware in each module. We are looking to module makers to create technology never before seen on smartphones.
With 3D printers, cheaper cheapsets, ..., tinkerers can go on and try to create their own module which they can integrate into their phone. If the module is popular amongst users, integrate it either as a standard module in the ARA store (a 3rd party could for example propose to manufacture that module, shifting those costs to them, Google would just deliver the core ARA product).
Every piece they make modular adds cost - cost of connectors, weight of the frame to support them, cost of support when the connectors fail.
You can go on about "waste" all you want, but the lowest cost, lowest material devices are rated for single-insertion connections along with components both stacked and heavily integrated.
The reason they could not do displays, RAM, etc is that these components are tightly integrated to the SOC selected, and change regularly. This was true when they announced, even more true now.
Truly Google is the new Microsoft if they devote such time to obvious executive vanity projects, untouched by the taint of reality.