Stop me if you've heard this before:
I have drunk the Kool Aid (even though I thought dropping Akida 1 chips was a mistake).
Akida 1 is an advanced digital SNN processor. There are some less capable SNN processors produced by startups finding their way to market, both analog and digital. The big players are still trailing behind.
The field of AI technology is undergoing an unprecedent rate of change and this makes it extremely difficult to gain traction in the market. There has never been such a rapid and continuous rate of development in the field of electronics as we are witnessing today.
At this stage of the company’s development the company is in a protracted pre-revenue period in which the business model has changed in the last 2 years from:
- chips plus IP (Akida 1)
- Akida 1 IP only,
- “would you like chips with that”,
- Akida 1 IP or Akida 2 with TeNNs, ViT & long skip IP only (possibly as software),
- Akida 2 with TeNNs etc IP or stand alone TeNNS
while researching Akida 2 for LLMs and ongoing TeNNs development - not to mention my hypothesis about software simulation.
All the while, under cover of NDAs, the engineering and marketing efforts of the company are focused on the customers. As mentioned above, the rapid rate of development of the technology has militated against finalizing a deal. Our engineers are too clever for our own good. Hopefully TeNNs will provide a plateau at which EAPS are happy to pause in their quest for the holy grail, and some EAPs will ink an agreement.
I think it is likely that the EAPs have had access to TeNNs since the patent was filed 2.5 years ago, and that they have been testing software simulations. As TeNNs continues to develop, the software can be readily update, which would not be possible with silicon.
Possibly TeNNs uses such low power and low latency that a software simulation provides a significant advantage even in EVs.
We know EAPs such as Valeo, MB and others are using software to process their sensor signals. If they are using Akida simulation software, this would be a way of enabling ongoing upgrading as the capabilities of Akida continue to evolve.
A cloud-based FPGA has been foreshadowed in the capital raise notice. Basically this will be a configurable COTS product, far cheaper than taping out and manufacturing a wafer, and, while no where near as efficient as an NSoC, it will serve for demonstration purposes (something we are copying from Intel's Loihi demonstrator).
The Brainchip university continues to expand its reach.
ESA and other space agencies and participants have endorsed Akida.
We have over a hundred engagements.
Yes - the NDAs are frustrating, but I believe the green shoots are about to erupt into a rainforest.
The extra cash gives a little more comfort. Sorry to see the BRI being expatriated and the research "standardized", but it shows our focus on LLMs, and will reduce some costs.
exoswan.com
