Nvidia Says Vera Rubin Represents $40B in Revenue Opportunity per Gigawatt

Nvidia says a gigawatt of AI infrastructure based on Vera Rubin represents about $40 billion in revenue opportunity. It is a company-stated opportunity figure, not delivered revenue, but it is more than twice the roughly $18 billion Nvidia associated with Hopper—and raises the economic stakes of selling systems rather than standalone processors.

Nvidia’s explanation is its integrated AI-factory platform. Management says it spans Vera CPUs, Rubin GPUs, NVLink and InfiniBand connections, Ethernet networking, systems, algorithms and CUDA software. That puts more of the compute system—and the software developers use—inside Nvidia’s offering.

Broadcom is pairing custom accelerators with the networking that links AI clusters. It reported $16.70 billion in AI semiconductor revenue in fiscal Q3 2026, up 221% year over year, and forecast $21.7 billion for fiscal Q4. Custom XPUs—chips co-designed for particular customers—made up 73% of the quarterly AI revenue.

Other ways to reach beyond a single chip

• AMD is packaging EPYC Venice CPUs, MI450 GPUs, Pensando networking and Rackham software into its pre-integrated Helios rack. AMD says Anthropic committed to deploy up to two gigawatts of MI450 capacity in Helios, while Meta plans up to six gigawatts of Instinct capacity.
• Marvell designs custom XPU-related silicon as well as optical and Ethernet components for AI clusters. Management expects its custom-silicon business to more than double year over year in fiscal 2028 and accelerate again in fiscal 2029.
• Intel’s route combines Xeon host CPUs with foundry capacity. Intel says its 18A manufacturing process is in volume production and 18AP is in risk production, as it positions the foundry as a U.S.-based advanced-logic alternative.

These claims are at different stages. Broadcom’s Q3 figure is completed revenue; AMD’s capacity commitments are planned deployments; and Marvell’s outlook is forward-looking management guidance. Still, each approach aims to capture a larger portion of an AI installation, whether through a finished rack, custom silicon and connectivity, or manufacturing capacity.

Nvidia says its fiscal 2028 outlook implies approximately 70% growth, while calling that outlook supply-constrained. Its $40 billion figure depends on supplying a coordinated platform through a strained supply chain; the alternatives depend on customers choosing a different mix of components and suppliers.