I wonder how Oxide's basic value proposition changes with the very recent growth in rack-scale and multi-rack (datacenter-wide) compute for AI. Surely these other vendors have tech (particularly around network interconnect) that can be repurposed effectively for the more general private cloud workloads Oxide is focusing on.

Conversely, has Oxide discussed in depth how they see GPU compute?

I’m guessing for their current target market of existing enterprise software it doesn’t matter, but the ownership and economics stories are at least as compelling for matmul code.