Microsoft’s IoT strategy came into sharper focus during the keynote at Microsoft Build Monday. The software company is betting on the cloud, while acknowledging the needs of those at the edge. Microsoft is building tools to link its Azure cloud and various services to the edge devices that are gathering and analyzing data, but the base assumption is that the cloud will be an essential element for analysis and management. And that’s where we’ll see Microsoft try to control the experience and revenue. But to get developers into the cloud it’s glitzing up its offerings at the edge.
So let’s talk about the edge for a moment. This is a trendy and poorly defined term that can encompass sensors that deliver temperature information from a farmer’s field to a massive computer that sits on the floor of a factory floor.
A few years ago I spoke with Sam George, who is in charge of Microsoft’s IoT efforts. He said that Microsoft wasn’t really focused on the edge or the middleware connecting the edge and the cloud. Partners would handle those aspects of the IoT ecosystem. Clearly, the company has adapted that strategy, because it’s too hard to create an IoT device or service by gluing those different pieces together without a lot of development expertise in areas such as embedded programming, application development and various cloud development.
So Microsoft has pushed out from the cloud to offer both actual software and services as well as partnerships to help developers bridge the entire IoT ecosystem. It’s also using this holistic view of what software must tackle to integrate machine learning (from Microsoft) in all parts of the IoT ecosystem. So what does this look like in practice?
Its security platform Azure Sphere runs on tiny microcontrollers that the company expects to be in billions of connected devices. However, the security comes from the hardware directly on the chip, calling the cloud to confirm certificates and also for checking for new vulnerabilities. This combines the edge, a custom OS, and the cloud. From a revenue perspective, Microsoft gets a cut of every Azure Sphere certified chip sold and the costs of the OS updates and cloud aspects are part of the cost of buying the chip.
For another view, on Monday during his keynote, Microsoft CEO Satya Nadella said that Microsoft would open source its entire Azure IoT at the Edge suite. This move should assuage fears from big-name clients who want to control everything about the gateways and sensors running on their factory floors and offices. They can pull down the code, inspect it, tweak it and even resell their newly tweaked software. The idea is that making the IoT at the Edge more available and in highly customized versions for various industries will drive users to use the Azure Cloud services for their data and everything else. The cloud is where Microsoft can control costs and profit margins, so this is where it needs to drive users.
You also see this strategy play out with the machine learning services it plans to offer IoT customers. These services range from image recognition to translation. Microsoft’s use of containers is optimized to move and share those services on both the edge software and in the cloud. This makes it really easy for developers to build services all within the Microsoft ecosystem, reaching out through Microsoft-curated connectors for popular outside APIs. This is less about flexibility and more about consistency across millions of edge devices and the cloud.
Another indication of the primacy of the cloud comes from the partnerships Microsoft announced with chip firm Qualcomm and drone maker DJI. Microsoft will put its IoT Edge and computer vision services on Qualcomm chips designed for connected cameras and other video devices. It also plans to make it easy for DJI drones carrying cameras to link their data and use Microsoft’s computer vision services without friction. These screenless devices were never going to run a full fledge OS or need the type of productivity software Microsoft made its money on.
Instead, the software maker must find services that a variety of new device types can use and make it incredibly easy and compelling to use them. Amazon is trying the same thing, although it’s taking a slightly different tack such as bringing on the creator of the FreeRTOS embedded OS to get developers and embedded systems engineers to design in hooks for AWS into physical products. I also don’t see the emphasis on higher level services around AI models and machine learning that Microsoft is pushing. Amazon has a few, but Microsoft has a few dozen.
What we’re seeing is Microsoft acknowledging that there are going to be tens of billions of connected devices, but its core business will remain in the cloud. It’s strategy so far is to make it easy to get people to build ecosystems entirely built on Azure and to push a variety of AI-based models to them.