[U2] Update: Winning at the Edge
Greetings from the Edge,
A short-form update is the format this week to augment recent profiles on key topics. I will do this to keep to weekly communication. Three (3) key developments are driving this update.
I’m in Seattle launching products & services with a strategic customer. The long days are competing for my time, but the insights gained help validate recent profile insights on Apple and Tesla that also extend to Amazon.
Much attention today is focused on AI and cloud services. As a knowledge-base scales, so does the ability to differentiate and monetize asset(s) with sustainable revenue models.
Winning will increasingly happen at the Edge.
As mentioned in profile [E3.2], Apple & Tesla have vertically-integrated semi-conductor design into product teams. Apple remains fab-less like many other companies except for ones like Samsung and Intel who also fabricate their own semi-conductor chip designs. Apple contracts with third-party fabs like Taiwan Semiconductor to build their chipsets. But building expertise and scale on these key component designs adds a competitive advantage. Driving more capability into high volume, mobile devices differentiates those products, but also allows key technology to migrate into higher end platforms later to extend those capabilities into more products.
Edge Computing
A growing technology domain is referred to as edge computing. As more processing has consolidated to the cloud to leverage scalable & feature-rich cloud services, latencies to access those services have increased relative to on-device & on-site computing services. For many applications this has no real negative effect where the ambition is simply to scale & leverage services cost-effectively. But for real-time applications the latencies can become unacceptable. Edge computing extends cloud capabilities closer to the end user to decrease latency and to enable real-time or near real-time functionality. Content hosting is an early adopter of edge computing through companies like Cloudflare and Fastly to decrease response time to news and other media.
Industrial IoT is an example where edge computing can play a role if sensing data is paired with a real-time response. Cloud services with an AI engine build insights to model actions from a sensor feature-set. These features can then be pushed into the devices with the sensors to enable real-time capabilities. This can then lead to an iterative 1.) deploy, 2.) cloud activation to scale AI models & insights and 3.) transfer these insights to the edge to enhance real-time capabilities. This is a key reason for the growing adoption of neural processing chips on devices like the iPhone 12 and Tesla EVs. This enables these devices to run the AI models developed in the cloud on the actual devices with the sensors collecting data. One example is Face ID on more recent iPhone models. This feature is possible with facial recognition that requires significant processing on the device itself.
Tesla is at the extreme edge where human drivers using autopilot and full self driving services leverage AI for autonomous navigation until they intervene [E3.1]. These interventions lead to exceptions that are uploaded to Tesla for further analysis to train the cloud AI engine. Over the air updates then take improvements to the AI engine that are implemented on the neural chip in each vehicle. This evolving sensing to operate at the edge in real-time, exceptions at the edge, cloud AI engine enhancements and pushing improvements to vehicles at the edge creates a feedback loop that easily scales with more units the field. Internal combustion engine (ICE) vehicles will not be disrupted by the electric motor and batteries. ICE vehicles will be disrupted and mostly eliminated by edge computing that leverages electric vehicles. EVs win. iPhones win. AWS wins. Cloudflare wins.
Let’s win at the Edge,
Stephen
Nothing in this post is intended to serve as financial advice. Do your own research.