At Stanford CS153 Frontier Systems, Jensen states word for word that he "would like to be at low MFU all the time" &…
SemiAnalysis Twitter · SemiAnalysis (@SemiAnalysis_) · 2026-05-17
SemiAnalysis reports that Jensen Huang told Stanford CS153 he deliberately targets low Model FLOP Utilization (MFU) as a philosophy of intentional over-provisioning of compute, networking, and memory resources, and hints that xAI's kernel engineers may follow this approach.
Appears in
Extraction
Topics: gpu-infrastructurehardware-utilizationnvidiaai-infrastructure-strategy
Claims
- Jensen Huang stated at Stanford CS153 that he would like to operate at low MFU at all times.
- Huang's rationale is that low MFU reflects deliberate over-provisioning of FLOPs, networking, and memory rather than a failure of utilization.
- SemiAnalysis implies xAI's kernel team may be applying a similar infrastructure philosophy.
Key quotes
would like to be at low MFU all the time
he wants be so smart, he over-provisioned the work like flops, networking, memory, etc.