While not the same, the folding logic technique that stacks several processing layers on their new cpu's, also focus on data-transmissions and thus saves on heat, quicker transport, and a few other secondary benefits.

It seem like Huawei rethinks communication on all HW scales. ..and that thinking 'kind of' goes for Chinese AI models also. Impressive that Huawei/China can jump in and teach old dogs some new tricks, imho..

Wonder if we will get active-computing ram where data+simple compute mostly stays on ram, active computing screens or other 'crazy' optimisations on traditional designs. At least I'm confident where it will come from..