A new approach to chips could help future-proof the IoT

It all comes down to the silicon. At the heart of every connected device are chips that serve as a brain by taking in information and relaying information back out, along with dozens of components that provide an internal clock, power management, and more. While AI and software updates can change the functionality of our devices in pretty awesome ways, the underlying hardware will always be a limiting factor.

You can’t send an update if the device doesn’t have enough memory to hold that update in storage for the turnover, for example. Nor can you can teach an old chip new tricks if it doesn’t have the processing power. And over time, those limitations — along with the physical limitations of silicon — show their age, turning a once-cool product into a heap of junk. Just ask Tesla.

But a new way of selling chips could help IoT device makers future proof their long-lived devices by making it easier to upgrade silicon after it is designed into a product.

The industry’s solution to future-proofing has typically focused on building modular boxes that can slot into long-lived devices, such as cars. The idea being that, when the brains of the system wear out, a manufacturer just pops in an updated version. But while many manufacturers — from automotive companies to makers of large kitchen appliances — talked about modularity, I haven’t seen much emphasis on it in product design.

Some of the hesitation on the part of manufacturers stems from the user’s reluctance to buy an expensive product and then shell out another decent chunk of change some five or seven years later for an upgrade. There is also the legitimate concern over how easy it really is to support modularity — first at the design phase and then at the eventual servicing phase — as well as supporting the inventory and supply chain for decades worth of devices.

So most manufacturers future-proof their connected products by putting in as much memory and much computing power as they can afford. They may also add sensors and other components to support feature upgrades. All of which makes sense but is expensive and can raise the cost of the end product — which can lower the product’s chance of success.

But at CES, I saw a solution to this problem on the horizon. Qualcomm announced a connected car cloud and chip package containing what it calls a “Soft SKU:”

Qualcomm Car-to-Cloud also features an innovative Soft SKU capability, allowing for field-upgradeable chips to securely support new functionality as performance requirements increase or new features become available. This allows for mid-cycle performance boosts, regional customization and feature upgrades tailored for specific product segments. In addition, because of Soft SKU capabilities, OEMs can customize a single SKU for multiple tiers and markets,  which can lead to saving costs associated with dedicated investments in multiple SKUs.

There are two things here to focus on. One is that you can buy a single chip and customize it regionally, which is important for global manufacturers. The other is that you can turn on new features using over-the-air updates as needed. This last feature is incredibly compelling for companies trying to make future-proof connected devices. If you need more memory, just turn on more memory and send a licensing payment to the chipmaker.

What’s enabled this sort of business model is the evolution of high-grade security between the device and the cloud, according to Tyson Tuttle, CEO of Silicon Labs. He says Silicon Labs plans to offer similar functionality for some of its chips this year.

For device manufacturers, this is a way to keep the costs of silicon low and still offer features for users over time. On the semiconductor side, Tuttle assures me that Silicon Labs would still be able to make and sell the version of the chip with fewer features at a profit, so if device makers never turned on the upgraded features, Silicon Labs wouldn’t lose money.

While relatively new at the actual chip level, the idea of building in more capability and then turning on more features over time is fairly common, says Kevin Krewell, a principal analyst at Tirias Research. IBM used to turn on more cores for customers in their servers as needed. But when it comes to the internet of things, this capability could be a game changer for manufacturers trying to build a generation of computer-containing products that won’t become obsolete just three years later.