It may not be the case that more iterations mean no one will buy it. The way I see it, when they said Akida 1000 was never meant to be a revenue generator but a demo chip, what happened was that they invented a ground breaking, earth shattering chip but there was no application for it. They have EAP's with the tech and money to try it out and when they said "this does not work" or "it would be btter if this would happen instead" type of things and Brainchip put all the bug fixes and enhancements in AK1500 and the more fundamental changes into AK2000.
I believe that cash is not coming in for AK1000 but with AK1500 for the more run-of-the-mill stuff and AK2000 for the more advanced stuff like deep space, star wars and medical applications, revenue can be explosive. The question now is when? Hopefully before some tech giants come up with a better alternative.
I am just a bit frustrated with the moving of the goal post. We saw no big revenue in 2022 when we were told to watch the financials and now it looks like nothing worth mentioning in 2023 as well.
I guess that is the risk of investing in something so new, it may take much longer time than originally expected to identify and build the applications.
I believe the context is that the actual "chip" was not meant to be a revenue generator. They did not have an enormous amount of the chips manufactured, and those they did manufacture were put into the devices sold to EAP partners, evaluators, and developers.
The AKD1000 IP is still a very viable and marketable commercial product. MegaChips and Renesas have licensed it for their products and customers. BrainChip spelled this out long ago on the roadmap shared at the various technology and investor conferences.
NVISO makes its human behavior SDKs available for the AKD1000 platform but also demonstrates substantial performance differences between BrainChip's technology and more well-known competitors.
My point is that it was not a "throwaway" first pass that was never intended to be a product; quite the contrary. Yes, neuromorphic is still quite a new concept in AI hardware and will require adoption. I imagine the first adopters are making neuromorphic sensors, such as those for vision processing. More will follow over time.
In the meantime, it makes no sense for BrainChip to stop working on its second-generation technology influenced by other companies whose interest has now been piqued. I imagine they will be treated to architectural improvements that will provide the functionality required to fulfill their vision while providing enhanced performance for specific applications.
There are plenty of applications that can be fulfilled using 1st generation Akida architecture.
and as a consequence no one will ever buy it.
