[xpuente]

I wonder what will happen if someone figures out that fused FP mult add is no longer needed (e.g. just count spikes and add subtract permanence). This could be a big problem for the guys with all their eggs in one basket (like NVIDIA).

[zarzavat]

This is what I believe as well. There's no way that FMA is the optimal way to compute NNs in silicon. It's overly precise.

NVIDIA will be okay though, they have the volume to get the newest nodes first. Only a few can compete with them.