I have a growing concern in my waters that although we have an...

I have a growing concern in my waters that although we have an outstanding management team, and outstanding research staff, there may be technical hurdles that just can be jumped.

I refer to memory size and indirectly a selector that facilitates more compact, and thus larger memories.

For some time the maximum memory that has been produced is 1MByte (often expressed as 8 Mbits to make it sound larger).

I challenged Coby about this limit on the last shareholder conference. He hesitated, before fluffing an answer about embedded memory only needing small sizes and 1MB being plenty. He also suggested that they may produce larger memory arrays in the future but that was a long way off and not a priority.

I place this along side the selector issue. Years ago we were on the verge on a new and better selector. Then when that was due to be unveiled, the language morphed into we are evaluating a number of alternative selectors. That now seems to be an ongoing position, with no key candidate, and is something that will take time as do all things semiconductor.

I believe 1MByte may be the practical limit in the near, and perhaps mid term.

Now this presents a problem. 1MByte won't even hold a decent selfie, it won't support sufficient parameter storage for any self respecting AI model running at the edge, nor do I suspect give much power for in-memory computing. Is this limit the reason that the opportunity for discrete with the current selector (transistor) never got off the ground.

I accept there is a range of products that will find 1MByte embedded sufficient.

Problem is we (or at least I) have always evaluated potential market to include discrete, and every device on the planet. However I think 1MB embedded gives as a much more limited market size, and is not set to dominate the world, A respectable market, but not Intel inside.

I have been a long term holder, and love the talent that the company has assembled. I do not doubt the business cred they have, but fear there may be a technical limit that can't be overcome. Is this the reason all the other ReRams have fallen by the wayside?