NVIDIA, Google and Microsoft push 800 VDC datacenter power through OCP — 2MW per row, 80+ vendors building to spec
NVIDIA's 2026-08-11 post makes the case that AI rack density has outgrown AC distribution, and that 800-volt DC cuts conversion stages between grid and GPU so more available power reaches compute. Developed with Google and Microsoft through the Open Compute Project, with 80+ equipment manufacturers building to the specification. The rollout is staged: a hybrid AC-compatible power rack in the second half of 2026, a row power center supporting up to 2 megawatts per row in 2027, and a facility-scale DC power block later in the decade, with NVIDIA DSX reference designs as the system-level blueprint. Notably the post gives no efficiency percentages or copper-reduction figures — it is a roadmap and standardization play, so treat the efficiency claim as unquantified.
↳ Follow the thread