2021 BRN Discussion, page-11338

Hi ER
Firstly where do you get your evidence that TSMC did not do the first production run last year. I was in no doubt that they did as part of a shared wafer fabrication?

The issue as I understand it for this second run is that Brainchip have added new functionality to the AKD1000 chip and as a result of the AKD1000 engineering sample being available for experimentation they were also able to work out a way to further improve and reduce energy consumption. So clearly taking a steady as we go approach Brainchip under the watchful eye of Mr. Chip, Anil Mankar, they are having this new iteration of AKD1000 produced in sufficient numbers to confirm that it works and what will be the yield so as to determine price for signing up customers.

What I have read about TSMC is that they have the highest yield rate in the world at 28nm which is greater than 90%. So one could expect that a minimum yield from TSMC will be in this order. Again this was a distinct advantage of 28nm selection as it drops significantly as you go down in size from here and this reduced yield in part accounts for the rapidly increasing price as size is reduced. We know that the additional layers required for AKD1000 has pushed out the normal production time for 28nm beyond the usual 40 days so this adds a potential opportunity for faulty chips but the numbers of layers increase also well beyond what AKIDA requires at 28nm as you go down in size so it should not be a particular issue as TSMC are very experienced at laying down layer after layer on much tinnier chips successfully.

Now if we go back to the former CEO's comments in October, 2020 he made it clear then that the IP customers they were working with were waiting for the green light of silicon in hand before finally signing off.

As we know Renesas jumped not much more than eight weeks after the chips and board were available. So I would say this, the first engineering samples were all about proving to IP customers that this technology really could be made to work and this was a complete success as Renesas signed on the dotted line.

In my opinion this proves that this second generation of AKD1000 is not relevant where IP customers are concerned. In fact if I rely upon the reports of those who listened to the AGM live Brainchip is in discussion with other potential IP customers as I type this and have an expectation of further deals being done in the near term. Again if this is a correct recollection of what was said (and I do not doubt the other posters) then this is further proof that the sale of the IP has nothing to do with this further AKD1000 being successfully produced.

The AKD1000 chip that has gone to TSMC recently is a chip that has had features added at the request of a customer or customers and it would be those who want the actual AKD1000 chip to use in their product or invention. If IP customers want to add features this would be accommodated at their plant by their engineers in consultation with the Brainchip engineering teams assigned to them.

So we go to this question of is the next AKD1000 with the added features likely to work. Personally I have no doubt. Mr. Chip, Anil Mankar knows what he is doing and I am sure that if it was too risky to add these features he would have put his foot down because there is absolutely no reason to be frightened of the competition as there is none at this stage. So I cannot see Anil Mankar recommending the throwing away of scarce resources on a fools errand. I have watched Anil Mankar for a number of years now and the change in his confidence since the first AKD1000 silicon popped out and was tested has been incredible and obvious to all concerned. He does not speak to me as a person who has any doubts whatsoever about the successful production in August of the AKD1000 chip.

Secondly even after Anil Mankar signed off on the practicality of adding these features it has gone off to Socionext who successfully worked on AKD1000 last year and they would have, I am sure, been forthright about whether these additional features were a step too far and in consultation with Anil Mankar decided not to change the design.

So I personally do not have any of concerns you have expressed but I am keen to know the answer to my first question as to why you believe TSMC did not make the earlier version.

My opinion only DYOR
FF

AKIDA Ballista