2025 BrainChip Discussion, page-4166

Thank you for providing references.

The reference Grok provides for Loihi takes me to an Intel document which I have read. There was nothing new in this document but I extracted the following as it makes clear what all genuine shareholders know . Intel’s Loihi is a long way behind Brainchip. Despite the heading “Available Loihi 2 Hardware” if you read the fine print it should more accurately read “Unavailable Loihi 2 Hardware”. I will let you read it for yourself but the only Loihi 2 that is available is a limited availability Gulch and then only to a select few with the rest accessing via the Cloud.

Then if you take the time to read about Lava it is designed so it is hardware agnostic and open sourced a fact which I already knew and referred a question about to Anil Mankar when it was first released. My question was whether Lava would run on AKIDA and his answer was yes but we have our own software. Of course my question was a Dorthy Dixor because agnostic means it can run in all hardware but you cannot be too careful.

The premise of your question to Grok “If” leads to a flawed analysis. Grok in accepting that needs to accept that all the issues identified by Mike Davies that need to be resolved with Loihi technology have been resolved. The bottle necking no longer exists and programmability has been solved. In accepting that Grok then proceeds with the analysis and arrives at a series of flawed conclusions.

The reality is Loihi technology is not commercially available and is not on Intel’s product roadmap through to and including 2026 according to the 2024 Annual Report. Mike Davies is on record in an interview with EETimes in 2023 as saying a commercial Loihi 2 was about 5 years away. This takes us to 2028.

So unless the Annual Report and Mike Davies cannot be relied upon then I fail to see the value in your Grok hypothetical.

AKIDA 3.0 including 16 bit is timetabled to be released by 2026. AKD1000 was designed so that up to 64 could be connected together for increased performance but theoretically the design allow ms for up to 1024. In saying this it might seem like I am criticising the use of Grok but with Lava being hardware agnostic and AKD1000 being so scalable logic suggests there are some potential flaws in its logic about Loihi 2 -v- AKD1000 before the more advanced AKIDA offerings are considered.

My opinion only DYOR

Fact Finder