Hi Iseki,
I would like to see the chip made available. It is certainly easier and faster to sell the chip and have it in customers hands, but the issue is that the many applications for Akida 2e do not need 256 nodes (1024 NPEs). There is whole spectrum of node requirements, so manufacturing and warehousing all the different chips would be uneconomic at this stage. This is a totally new market, so it would be risky to guesstimate the market size for each size of Akida.
I don't see the need for a consortium as we are working with 50-odd disclosed customers (and N undisclosed customers) to develop individual products, so they must see the incentive to implement the revolutionary technology and take advantage of its adaptability, speed and power efficiency. It takes a long time from drawing board to product.
We know that Akida has a sweet spot for lidar and DVS.
Cyber security is also a great application where quasi-real time response and on chip learning are required.
We know that camera and video images can be converted to spike signals for image processing by Akida.
While Akida is sensor agnostic, there is also a need for model libraries against which the sensor signals can be compared.
Maybe in time, when the company has established positive cash flow, they can address the chip market, but then again, that may be best left to a licencee. That said, ARM, feeling they are missing out on the full benefit of their technology, has been talking about going into chip making, so it is something to think about, but not at present for BRN.
