AMD Adds a Model Router to Its Helion OCP Blueprint

AMD has added a Model Router component to its Helion blueprint aimed at the Open Compute Project, according to a Tech Times report.

The headline claim is straightforward: with that design, operators can scale their serving capacity by expanding CPU resources rather than purchasing additional GPUs. In effect, the expansion path shifts from buying more accelerators to buying more general-purpose compute.

Folding model routing into a rack-level blueprint moves the scheduling decision up into the infrastructure orchestration layer. Requests no longer have to default wholesale to a GPU pool; they can be distributed across compute types, which challenges the GPU-centric assumption that has shaped recent deployment planning.

For data centers, that speaks directly to cost structure. GPUs are the most expensive and most supply-constrained component in today's AI infrastructure, so pushing a share of workloads onto CPUs could lower the marginal cost of expansion.

Caveats apply: the candidate source offers title-level information only. Technical specifics, supported model types and measured performance were not disclosed, so this reads as a directional architecture signal rather than a benchmark result.

The questions to watch are adoption inside the OCP community and how much real inference workload “CPU scaling” can actually absorb in production environments.