Hi Chapman,
The Raspberry Pi board is the key to saturating the market. At $10k the shuttle PC board is almost exclusively for businesses. For the average customer there's not a significant benefit buying one of these over a Raspberry Pi, so most people will purchase a Pi.
At $49-69 the Raspberry Pi is affordable for most consumers and hobbyists. It is cheap, user-friendly and requires minimal coding experience. Being so cheap these will end up in most schools, Universities, robotics clubs and the like. If Brainchip are to be believed that are at least 3 years ahead of the competition, this in turn means they will probably become the standard for neuromorphic chips.
So in a few years time, a large portion of people coming out of school or university will have had some sort of experience with an Akida embedded Raspberry Pi. They will remember that it is user-friendly, requires minimal coding experience and is cheap. And wherever they end up working they will be able to tell their employer about the benefits. They will be able to demonstrate how easy it is to use and to develop their own solutions that their employers may in turn start thinking about ways to integrate these cheap devices.
If a customer wants to prototype a novel sensing solution, all they need is an Akida Raspberry Pi board and the particular sensor. So there is potential here to prototype new and novel solutions for less than $100. This is a recipe for disruption IMO. For <$200 in hardware you may be able to prototype a new lightweight, small, handheld, novel solution that draws negligible power. So provided someone has access to an appropriate sensor or other required equipment, they could effectively create their own portable breathalyser, intelligent camera, or cybersecurity monitoring device, or sharemarket analyser, etc etc. Importantly you won't need a university degree to do any of this. It's just a matter of pairing the right hardware and obtaining sufficient data.
I see it taking a few years but Akida will likely become the standard for neuromorphic chips. It's worth factoring in a time frame here though. There are 40 million Raspberry Pi's currently, which means they are probably selling 5-10 million per year. So if Brainchip were to start selling them now, I would anticipate 5-15 million being sold by next financial year. It will take some time for hobbyists to hear about Brainchip, plus additional time for the chips to be produced (keeping in mind the current worldwide chip shortage). I don't anticipate schools to start buying them straight away, it will take some time to hear about them and then get funding etc.
Once Brainchip becomes the default neuromorphic chip, it is going to be very difficult for competitors to steal market share. Why? Because Brainchip are setting a very high bar for potential neuromorphic chip competitors. To compete, they need to be able to address the following:
Price:
They would need to sell a chip for <$25 to compete.
Technical capabilities:
They would need to make the chip compatible with all major machine learning languages (eg. Python, Tensorflow at a minimum).
They will also need an extensive model zoo with examples customers can immediately try out themselves.
They will need a MetaTF equivalent simulator.Features:
They would need to implement at a minimum one-shot learning and on-chip convolution, plus additional features to make people have the need to switch. If all they can do is match Brainchip's efforts, no one will want to take the effort to learn to use this new hardware. It will be very difficult to do because Brainchip has patented their capabilities.
Customer service:
They will need to have engineers and potentially neuroscientists or the like trained up to help customers with their solutions. Brainchip are currently helping customers fine tune and develop their solutions as not all companies have expertise here.
Obviously this is just a quick summary, there are a lot of other important features and the like which give Brainchip an unfair advantage.
Pure speculation, DYOR
(20min delay)