In computing, you always have a fundamental choice: do you build on someone else’s platform or build your own? This is true at any level of abstraction you wish to inspect: be it operating systems, programing languages, instruction sets or any of the other myriad of turtles down to electrons on silicon.
Even the earliest computers relied on off-the-shelf components such as vacuum tubes, in their own way an abstraction with known properties.
Over time, the economics of standardization started to drive customer demand and decision making in design. Be it in programming languages such as C, hardware interfaces like USB or operating systems like Unix. Vertical integration, where one vendor controls the entire stack from silicon up to user experience, fell out of favor; replaced by companies specializing in different platforms out of which an OEM could build an end user experience. The pendulum swung from vertical integration to open systems.
It is in this context that Apple moving to silicon of their own design for the Mac is so curious. Apple now have a range of handheld and desktop devices that are running
- their own operating system, that includes
- unique programming frameworks (for example Metal over OpenGL) built using
- a unique language, Swift, on dedicated
- hardware of their own design running
- custom silicon
Now, of course, Apple doesn’t make all of this. Many of the components, from displays to storage are built by OEMs like LG, Sony and Samsung. Yet Apple buys at such volume they are able to push R&D at the OEMs to meet their needs. The value is in owning the intellectual property and systems design, not building the components.
It should be noted, as well, that Apple pulls heavily from open source projects as well as giving back, to some degree, with projects like Webkit and Swift. Even then, though, there is a trend for Apple to replace open source packages with their own open source of their own like Clang.
As the initial reviews of the new Macs come in, there are definite product benefits to Apple’s decision: the machines are faster than their Intel equivalents, with better battery life and comparatively cheaper. These are great reasons for Apple to own all their turtles.
Can anyone match them? Nintendo arguably has taken a similar approach with their gaming hardware over the years, yet for every Wii there’s been a Virtual Boy. Further, the scope is far more limited: Nintendo doesn’t have the same need to ensure compatibility across generations nor build differentiated products within the same generation.
Microsoft is attempting the same integration with the Surface line of tablets/laptops but it’s hard to imagine them ever owning their own silicon (and they must thread the needle of selling these products alongside partner/competitors). The closest analogs are not in consumer computing but cloud, with companies like Amazon and Google designing their own data centers and servers running increasingly proprietary walled garden platform as a service offerings like Spanner.
Where does this end? Full vertical integration is only achievable by the the biggest players with the deepest pockets. The economic and differentiating benefits to the approach are unarguable if you can strategically develop your platform over many years. Yet those same benefits ultimately become the seeds of destruction: like Compaq’s clean room implementation of the PC Bios, some scrappy startup will innovate in a way that challenges status quo and drive market demand for standarization. Then the pendulum will be nudged to swing another direction.
Featured image of Foucault’s pendulum by Andreas on Flickr