Skip to content
AI supply chain term

Superchip (GPU-CPU module)

A superchip is a module that tightly fuses one or more GPUs with a CPU over a fast internal link so they act as a single, memory-coherent compute unit.

What it means

A superchip is NVIDIA's term for a package that joins a CPU and one or more GPUs on the same module, connected by a high-speed coherent link (NVLink-C2C) instead of a slower external bus. Examples include Grace Hopper and the GB200. Because the CPU and GPU share fast access to each other's memory, large models can spill beyond GPU memory with far less penalty, and data moves with less overhead. In the AI supply chain it sits at the compute node, blurring the old line between 'host CPU' and 'accelerator.' The design is a lever for running models too big for GPU memory alone, and it deepens NVIDIA's system-level integration — selling a coupled CPU-GPU building block rather than a bare GPU that a customer must pair with someone else's processor.

Why it matters to investors

Superchips push NVIDIA further up the stack, from selling chips toward selling integrated compute nodes, which strengthens its bundle and its share of total system value. It also shrinks the role of third-party server CPUs inside AI racks, a shift worth watching for the CPU vendors exposed to the data center.

See Superchip in the AI value chainIts live model score, why it matters, and every company exposed to it.

Companies on this part of the chain

NVIDIANVDACoreWeave, Inc.CRWVNebius Group N.V.NBISMicrosoftMSFT

Named to show where the term sits in the AI supply chain — research, not advice, and never a recommendation to buy or sell.

Related terms

GPU / AI acceleratorGraphics Processing Unit — the massively-parallel chip that does most AI training and inference.Host CPUA host CPU is the general-purpose server processor that sits beside AI accelerators, feeding them data and running everything the accelerator does not.NVLinkNVLink is NVIDIA's high-speed interconnect that links GPUs directly so many chips can act as one tightly-coupled machine.HBM (High-Bandwidth Memory)High-Bandwidth Memory — DRAM stacked next to an accelerator to feed it data fast; often the real limit on GPU output.

See Superchip in the live AI chain.

THE ENTITY maps every constraint onto one live model — which part is tight now, who owns it, and who gets squeezed when it moves. Plain-English reads you can check.

Explore the features
All glossary terms

THE ENTITY is an educational read on the AI supply chain — research, not investment advice. It explains how the chain works and who sits where, never price targets or buy/sell calls.