NVIDIA Refutes 10% PCB Price Cut Rumors; Focus Shifts to HBM4 Bottlenecks and Rubin Platform Delays

A rumor that NVIDIA demanded a 10% price cut from its PCB suppliers has been widely refuted; the real issue is the tight supply of high-end components, not NVIDIA's pricing power.

The launch schedule for NVIDIA's next-generation AI platform, "Rubin," is likely constrained by technological hurdles, primarily the development and validation of advanced HBM4 memory.

A potential delay of the advanced "Kyber" backplane PCB to 2028 suggests the challenges lie in upstream technology, shifting focus from supplier price cuts to the complex AI hardware roadmap.

NVDA.KS000660.KSMU.KS005930.KSROG.KS

A recent rumor suggested NVIDIA was demanding a blanket 10% price cut from its circuit board suppliers, but this narrative appears to be unfounded.

In fact, the story of a forced price cut doesn't align with the current market reality. The opposite seems to be true: the supply of high-end materials for AI servers is extremely tight. Key components like advanced laminates and special copper foils are in short supply, giving suppliers more pricing power, not less. Furthermore, the price of copper, a primary raw material, has been rising this year. In such an environment, a sweeping 10% price reduction demand would be illogical.

So, what's really happening? The focus should be on the immense technological challenges of building next-generation AI systems. The delay concerns surrounding NVIDIA's upcoming 'Rubin' platform are not about saving money on boards; they are about the fundamental difficulty of integrating brand-new, cutting-edge technologies.

The primary bottleneck is HBM4, the next generation of high-bandwidth memory. First, getting this new memory manufactured and validated to NVIDIA's strict specifications is a major engineering hurdle for memory makers, which could easily cause a delay of a few months. Second, the Rubin platform is not just a new GPU; it's a complex ecosystem of six interconnected chips, including new networking and processing units. Getting all these parts to work together flawlessly is a monumental task that naturally impacts the timeline.

This technology-first explanation is supported by reports that the 'Kyber' backplane, an advanced PCB designed for the future Rubin Ultra system, might see its volume production slip from 2027 to 2028. This points to design and integration complexity, not a pricing dispute. The chatter about price cuts was likely a misinterpretation of a much more nuanced and tech-driven reality in the AI supply chain.

PCB (Printed Circuit Board): The green board that acts as the foundation of most electronics, connecting components like chips and memory with copper pathways.
HBM (High Bandwidth Memory): A type of high-performance memory essential for AI processors, where memory chips are stacked vertically to achieve faster data transfer speeds.
Backplane: A large circuit board that serves as a backbone, connecting multiple smaller circuit boards (like server blades) together within a single chassis or rack.