The data centers powering the AI boom are no longer passive electricity consumers — chip giant Nvidia wants them to become active partners in stabilising the power grid. At the OCP Summit 2026, Nvidia VP Ian Buck outlined how AI factories can shift and curtail computing workloads in real time to give stressed power grids breathing room.

Demand-Response Computing: What Ian Buck Actually Argued

Buck's core proposition is deceptively simple: because AI training and inference jobs can often be scheduled or briefly paused, hyperscale facilities can function like giant, intelligent batteries on the grid — absorbing demand when power is plentiful and pulling back when utilities need relief. Nvidia's framing positions this not as a sacrifice in throughput but as a coordination opportunity — one that requires software, telemetry, and cross-industry trust to work.

The pitch matters because the scale is staggering. AI workloads have driven electricity demand at hyperscale facilities to levels that are straining regional grids across the United States, Europe, and increasingly Asia. Data-center operators are facing harder questions from regulators and utilities about whether their growth is compatible with grid stability.

Why the Open Compute Project Is the Right Room for This Conversation

The Open Compute Project (OCP), founded in 2011, was built on exactly the kind of cross-company hardware standardisation that Buck is now calling for on the power side. Its annual summit draws the engineers who actually specify rack designs, power delivery units, and cooling systems at companies like Meta, Microsoft, and Google. Bringing a demand-response pitch to OCP is a signal that Nvidia sees grid-awareness as an infrastructure-layer problem — not a one-vendor software feature.

The emerging practice — sometimes called 'grid-aware scheduling' — involves feeding real-time power-grid telemetry into workload orchestration systems so that jobs are automatically deferred or redistributed across geographies when a grid operator signals stress. It requires hardware vendors, cloud operators, and utilities to agree on data formats and response protocols: precisely the kind of open-standards work OCP excels at.

Buck's Role and Nvidia's Broader Power Strategy

Ian Buck, who leads Nvidia's Hyperscale and HPC business and is one of the architects behind the CUDA software platform, carries significant weight in the data-center engineering community. His presence at OCP signals that Nvidia is prepared to put engineering resources — not just marketing — behind grid-responsiveness. Analysts watching the space will be looking for follow-on specifications from OCP working groups and any announcements about grid-aware features in Nvidia's software stack.

The stakes extend well beyond energy efficiency optics. Governments in the European Union and the United States are actively considering regulations that would require large data centers to participate in demand-response programmes. For AI infrastructure companies, getting ahead of that curve — and shaping the technical standards — is both a business imperative and a licence-to-operate question.

If AI factories can learn to flex with the grid rather than fight it, the entire industry's relationship with public utilities changes. Nvidia is betting that it can write the rulebook for how that happens.