The silicon industry invisible revolution

I’ve always wanted to have a smartwatch that would be at least minimally useful. That minimal usefulness is very well defined - at least a week of battery life, virtual bezel, fitness apps, HR sensor, GPS, NFC and LTE optionally. Unfortunately, I have been waiting for anyone to produce such a device since the debut of the first Moto 360.

Today, I’ve created a map, and I think I finally know why no such watch will appear in the near future - and it is because we are facing a silent revolution in the silicon industry.

Let’s look at a map of the environment:

Multiple user needs drive the consumption of the computing power. This growth must be so extraordinary that the underlying value chain is being scrutinised for efficiency, and new approaches are being investigated. The AI plays a notable role here, as it created demand that prodded multiple players to enter the race of designing and improving chips.

Just look at the not so new news:

The effects of this are quite interesting. Existing players face challenges. They not only risk missing the future market, but they are also fighting for survival:

This, in turn, sets the landscape for wearables, and since they are a market nowhere near as a big as the market for specialised AI chips, nothing new is being created. Even Qualcomm, the provider of Snapdragon Wear 2100, a platform that powers majority of Android Wear smartwatches, seems to be content with such an old and outdated product.

It looks like the market of chip producers got rebooted, and because it is so deep in the value chain, an average consumer will never get a chance to see the bloody fight that is already happening in that space. More than that. Existing limitations in chip design know-how may be lifted if only they will become significant enough (they will), and we witness the pendulum of power moving back to hardware designers.

Only then, after the dust settles, companies will focus on smaller markets again, and I might be able to buy my desired smartwatch.

Edit: Apparently, I knew when to complain. A new mobile chipset is rumoured to be released in September 2018, along with new smartwatches. It looks like IoT and AI markets were not that big…

… and that chipset turns out to be disappointment. The addition of the smaller, always-on computing unit does not change the 2014ish technology (compare it with the Apple progress in this space).

It’s safe to assume quantum computing will create a period of punctuated equilibrium. According to a report by Tractica, the quantum computing market will reach $2.2bn by 2025.

To what extent is investment being diverted away from silicon to quantum computing R&D?

The number of companies involved in quantum computing is growing. Is innovation in silicon being stifled by a brain drain of engineers and scientists drawn by the huge potential of quantum?

I’d be interested to see any Wardley Maps, similar to the one above, that look at both silicon and quantum in terms of IoT and AI needs.

@john.grant, I wrote a master thesis about Quantum Computers, and while I am no expert, I think I still understand significant limitations, and the most important one is that there is no such thing as a programmable quantum computer - they have to have their circuits designed to solve a specific class of problems.

That essentially means that the overlap between QC and traditional computers is currently very minimal. Hardware is entirely different, still a subject to massive experimentation and needs university-class physicists, not everyday engineers. Until we see an efficient-enough approach to building a quantum computer, this will not change. Definitely custom-built hardware.

The second part is more interesting. Since you cannot set arbitrary quantum state, all the quantum algorithms have to be computable as a fixed step mathematical formulas, without the concept of computer read-write memory. That means that a creation each of such an algorithm is a tough challenge. When I was looking into that space ten years ago, there were 2 (two) algorithms, and today there are 18 of them.

Each algorithm solves a class of similar problems, and industries that have to solve those problems on a daily basis will face a revolution. The question is - what issues are specifically being resolved by QC, and how important they are for the business. The number and usefulness of QC algorithms will determine how much that industry will compete with classic, silicon-based approaches.

I am much more pessimistic about the impact of QC than Tractica.