Microsoft Builds On AI Development At The Edge

9 May 2018
Justin Warren

Microsoft’s annual Build conference is on this week, and the headline announcements are centered on all things AI and IoT/edge computing.
Microsoft is betting on a world where there are three main computing locations: the cloud, in datacenters, and at the edge, with a layer of AI sprinkled over the top of it all. The company wants to do this with two key platforms: Microsoft Azure, and Microsoft 365.

Sound familiar? It should, because Microsoft has done this before. Azure is Windows, and 365 is Office. These were the two cash cows that propelled Microsoft to become the behemoth that it is today. The plan is the same: make the experience of Microsoft products ubiquitous so that they become the de facto standard, and then print money.

The key to this plan is the same as it was for Windows: developers. If developers find the Azure/365 platform compelling, they’ll use it to write their apps, and then customers will want to use those apps, and you end up with a lovely flywheel, because developers want to write apps for the platforms that all their customers are already using. It’s why Blackberry, and Nokia, and Windows Phone all died: there was no there there.

The phone battle is lost, so Microsoft is moving on to the edge where the battle is fresh and new and there’s no one with a lock on a majority market share. Right now it’s the wild west, with the majority of device makers building on some kind of embedded Linux with a few custom utility functions thrown on top, which is why IoT devices are so incredibly robust and secure.

Or, they could use Azure IoT Edge instead and let Microsoft provide all the functions that are expensive to develop on your own, like an app and security update mechanism. When it’s a commodity to have (because any competitor can just run IoT Edge as well) then it becomes a hygiene factor. Why try to roll your own security (because that always goes well) when you can just buy it and get on with building stuff customers are willing to pay for? If a bunch of your competitors are using such a platform, not using a similar platform makes your product less desirable.

Right now, security is terrible everywhere, so having to charge a lot more for a low-priced device is a disadvantage when your competitors can just throw something out the door and cry “YOLO!” all the way to bank. The economic incentives pretty much demand that no one care about cybersecurity for cheap computers, particularly for consumer-focused ones. But if it’s easy to get security and all the other functions like AI modelling that links to the cloud, why wouldn’t you develop with that?

“Windows can be your primary dev box, whether building your cloud back-end, or building experiences for any of your edge devices: PCs, phones, IoT, HoloLens, or even XBox,” said Kevin Gallo, corporate vice president of Windows Developer Platform.

It’s a clever strategy, and it’s worked for Microsoft before. It’s also a reaction to developers moving en masse to Linux for their cloud-native applications, which posed an existential threat to Microsoft: If developers stopped writing apps for Windows/Azure/Office/365, everything else would come crashing down because without apps there are no customers, and with no customers there is no money. It’s why Microsoft is suddenly best friends with Linux and Open Source: they were winning.

It also reinforces Microsoft’s attempt to put Azure everywhere. The other cloud competitors have cloud, sure, but datacenters aren’t going away completely, and now with this third location of edge computing, would you prefer to write apps three times, or once? If I can code for Azure and have it run on my IoT devices or in my datacenter with something like Azure Stack, why wouldn’t I? As a developer, if I can keep my development costs down and make my apps available to run in more places and therefore get a bigger addressable market, why would I choose not to do that?

That’s the question Microsoft’s competitors will have to answer for developers and customers. They offer similar services in the cloud and got there first, so Microsoft has been playing catch-up but it’s made a lot of progress very quickly. Google has cloud and phones, but no datacenter story, no edge story, and no enterprise story. AWS has been focused on the cloud, though it’s pushing Lambda out into the edge as well, which is something to keep a close eye on for the same reason as this Microsoft approach: if I can write Lambda-based programs that run anywhere, why would I target a different platform? But AWS is really focused on enterprises leaving the Windows world behind and moving wholesale to its vision of the world, which for many is a bridge too far.

The idea of ubiquitous computing has been an elusive dream since before Java promised write-once-run-anywhere programs. It’s a really hard thing to pull off, and it’ll take quite a while, but Microsoft is a force to be reckoned with and you underestimate it at your peril.

This article first appeared in here.