Connecting One Trillion Things

Connecting One Trillion Things

The theme at this year's Arm TechCon was about the road to one trillion smart, connected devices. This staggering number frames Arm's vision and its direction, which CEO Simon Segars blogged about earlier this year: "We think the ArmRegistered ecosystem is the rallying point for tomorrow—and the heart of a future built on the silicon of a trillion secure, connected devices that will lead to world-changing inventions." Earlier this year, Masayoshi Son, chairman and CEO of Softbank, which acquired Arm, pledged that the company will deliver about one trillion chips for the internet of things (IoT) over the next 20 years. From there, it will, of course, take far more than a village—an ecosystem, really—to generate enough ideas and ingenuity to lead to real products.

All of which begs the question: how do we build and connect a trillion things?

For Arm Fellow Rob Aitken, the company's director of technology, the answer comes down to this: "For the trillion things to exist and do useful things for us, they have to work separately, together, automatically, and resiliently." Presenting on the final day of the conference, Aitken outlined what this means to ensure that the IoT works safely and securely:

  • Working separately: If you're designing a humidity sensor, then your product must work well as a humidity sensor. While this point may seem obvious, it serves as a reminder to not lose sight of your original design intention.
  • Working together: Communications between connected things and with the cloud is important to ensure that the things make the best and most efficient use of the data collected. Managing the energy and bandwidth between them is also crucial because there are limits on both. So, these constraints must be factored in. Processing locally to minimize communications can help ensure that things will work together.
  • Working automatically: Machine learning can help bring things to life and enable them to operate autonomously. But the algorithms must be trained with enough data and context to enable, say, a security video camera to distinguish between a person and a non-human. And the more that the device can handle on its own, the more we can reduce latency and communications bandwidth needs.
  • Working resiliently: When safety-critical things like locks, pumps, and power tools fail, there are serious consequences. Resilience needs to be built in at the component, network, and system levels.

How many connected things do you see in this kitchen?

Do We Have Enough Engineers to Design a Trillion Chips?

The guidelines that Aitken highlighted answer some of the "how to build one trillion things" question. Then there's the technology part. How do we, for example, build a trillion batteries? With limitations in rare earth metals used in batteries, noted Aitken, it's clear that energy harvesting will be essential to supply power. What about building a trillion chips, as is Arm's goal for the next two decades? Well, said Aitken, consider a sensor chip, which takes up roughly 2mm x 2mm of silicon. This would require 28 million wafers—roughly 30% of worldwide chip production. In this context, a trillion seems achievable. What about the people needed to design these chips? Currently, the world has roughly 5,000 RF designers. Designing a trillion radios would require 200 million RF designers. As for IC design teams, there are about 20,000 on the planet, and reaching that trillion target would require 50 million. By comparison, there's an abundance of coders—about 20 million in the world, while reaching a trillion chips would need about 50,000. One option, said Aitken, is to create a future where hardware design is like software design, so that coders are well equipped to build. In a Semiconductor Engineering article that Aitken published on the topic just before Arm TechCon kicked off, he noted that if all of the world’s coders designed sensors and each made 50,000 copies of their designs, we'd get to the one trillion. So the model could be one of either a long tail of highly customized designs, or a small number of standardized platforms that designers could use, he noted.

So, there's opportunity to rethink how we do things and also opportunity to innovate. For Aitken, the takeaway is simple: "The solutions we make don't have to be perfect, but, ideally, they have to be invisible to users. A trillion won’t happen if things don't work."