Horace Dediu has an interesting and thought provoking piece at the Christensen Institute which highlights come very interesting thoughts on whether a truly hardware specific modular approach computing is really ready for the consumer market.

A modular hardware (in addition to a modular software) computer may be conformable to buyer behavior but it may not offer opportunities of collaboration to producers; or vice versa.

The demise of Project Ara is not a testament to the demise of modularity. Modularity is abundantly evident throughout the smartphone business ecosystems. Millions of apps act as modules that enable completely different experiences for users of the same device. There are a vast set of accessories available through a myriad of interfaces such as bluetooth, USB and WiFi. The hardware subsystems are provided by a multitude of suppliers, and the products are assembled by various contract manufacturers.

Project Ara was an attempt to offer a particular interface across a boundary inside a device, but it offered no incentives to those who built devices or components. In contrast, apps offer interfaces “on top of” the device and offer great incentives for experimentation on solving pressing user problems. Modularity isn’t a theory that all boundaries should be open. It’s an observation that value chains evolve. How and when they evolve depends on an understanding of buyer behavior and producer incentives. via Christensen Institute

As Ara was cancelled, Google perhaps thought otherwise though one thing to keep in mind is that the ATAP team works in roughly 2 year sprints to try and develop an idea from prototype to product.  It’s possible that Google (and I have no information) may actually be working on something else … Even though considerable resources were likely deployed and some potential partner bridges built … Google has a pretty solid history of working on multiple concurrent efforts aka the Thunderdome Method.  I still think that this model has some legs … though perhaps not today.

While Dediu makes some solid points about the interface challenge as well as access to the stack, a hardware accessory (or module) could certainly trigger a software event to enable a more integrated experience … when you plug your car and phone together for Android Auto you get a unique experience only available in this situation.  Yes I know that’s an all-Google thing … but it’s a great example of how the phone shifts into a new mode.