Executive Summary

NVIDIA’s Rubin platform may look like a routine GPU upgrade, but its impact extends far beyond chip performance. By redesigning the internal architecture of AI servers, with a larger motherboard, liquid cooling as a core feature, and a modular dual-layer PCBA, Rubin fundamentally reshapes the value distribution within the supply chain.

PCB and materials suppliers gain new momentum from increased complexity and higher technical requirements. Liquid cooling vendors and metal processing partners become strategic players. ODMs move up the value chain, taking on full system integration, while traditional OBMs see their hardware differentiation erode.

Behind these engineering choices lies NVIDIA’s broader strategy: to standardize server design and consolidate control over the AI infrastructure ecosystem. By defining how systems are built, cooled, and serviced, NVIDIA increases customer lock-in and gradually transitions from a GPU maker to a rule setter for AI factories.

Introduction

In most early observations, Rubin has been viewed as an extension of the Oberon architecture. It continues the same chassis and cooling design language. Yet from a microarchitectural and system perspective, Rubin is not merely an extension but a new-generation GPU platform optimized for large-scale inference workloads. News coverage has largely focused on specifications such as GDDR7 memory, higher compute power, and power management, making Rubin appear to be only a performance upgrade.

However, when we shift our focus from the chip itself to the internal structure of the server, a different story emerges. The motherboard area inside the Rubin chassis has expanded by roughly 30 to 50 percent compared with the previous generation. The cooling system has grown from two plates to a five-layer liquid cooling structure, and the design now includes a high-speed midplane PCB and a dual-layer modular PCBA (see Table 1). These details rarely appear in public discussions, yet they are quietly reshaping value distribution across the supply chain and redefining who holds influence within the AI infrastructure ecosystem.

Table 1. Overview of Technical Changes

Value Chain Shift: Who Wins and Who Loses?

The technical adjustments in Rubin are not merely engineering choices; they directly reshape how value is distributed across the supply chain.

As motherboard size expands, liquid cooling becomes the primary thermal solution, and modular design is integrated into the chassis, clear winners and losers begin to emerge within the ecosystem.

PCB and substrate manufacturers are entering a quiet boom, while liquid cooling suppliers gain stronger pricing power. ODMs are moving to the center of the value chain, evolving from board assemblers into full system integrators.

By contrast, traditional brand-name server makers are seeing their position erode. In large-scale AI projects, they are no longer the design leaders but have gradually shifted toward sales and service roles.

Rubin may appear to be just another GPU upgrade, but in reality, it marks a redrawing of power across the entire supply chain.

1. Motherboard and PCB

One of the most significant structural differences in Rubin lies in the motherboard and PCB design. Compared with the GB300, the Rubin compute tray has expanded in area by roughly 30 to 50 percent and introduces a high-speed midplane PCB to replace traditional cable connections.

This is not simply a physical expansion but a shift in design philosophy. A larger surface area means more component mounting points, higher layer counts, and stricter demands on material stability and signal integrity.

For PCB and substrate suppliers, this represents a quiet wave of upgrades. Each Rubin compute tray carries a far higher value contribution than its predecessor. In other words, while the chip remains the star, the real driving forces behind this generation’s evolution are the PCB and materials that support it.

2. Cooling System

If the expansion of the motherboard symbolizes rising material demand, the transformation of the cooling system reveals a fundamental shift in design thinking. The number of cold plates in Rubin has increased from two in the previous generation to five. This change is not only about managing higher thermal loads but also signals that NVIDIA has officially made liquid cooling a standard configuration.

When cooling evolves from an optional feature to a core structural element, it becomes more than a condition for hardware operation. It turns into a key factor shaping both server design and the distribution of value across the supply chain. For liquid cooling module manufacturers, copper processing firms, and system integrators, Rubin represents a new wave of demand expansion. For companies still relying on air-cooled designs, it is a warning that their competitive advantage is quickly fading.

3. Modular and Rail Design

Another key change in Rubin is the introduction of modular thinking. Unlike traditional single-layer PCBA designs, Rubin’s compute tray adopts a dual-layer PCBA module connected horizontally through a midplane interface. To ensure ease of assembly and maintenance, internal rails have been added so that modules can slide in and out like drawers.

This approach not only improves serviceability but also suggests that NVIDIA is steering server architecture toward a more service-oriented design philosophy. For ODMs, it brings more testing, assembly validation, and added value. For brand-name manufacturers, however, the differentiation once built on maintenance and after-sales service is increasingly being embedded into the product design itself.

4. ODM

With Rubin, the level of compute tray delivery has expanded from L6 (board-level assembly) to L10 (full system integration). ODMs such as Foxconn, Wistron, and Quanta are no longer only assembling specific boards; they are now responsible for SMT, liquid cooling integration, and system testing, all of which significantly increase their value contribution.

In practice, Rubin has brought ODMs from behind the scenes to the forefront, making them key executors within NVIDIA’s platform ecosystem. For ODMs, this shift presents both opportunity and challenge: more value is being allocated to them, but it also requires greater responsibility and investment.

5. OBM / Brand Manufacturers

In contrast to the rise of ODMs, traditional brand-name server makers now face a difficult position. Rubin’s modular and rail-based architecture builds maintenance and replacement capabilities directly into the product, reducing the added value that traditional brands can provide.

As NVIDIA now defines the design language of servers directly, brand manufacturers are being pushed toward software integration and cloud services, gradually shifting from core designers to peripheral service providers within the ecosystem.

As shown in Table 2, NVIDIA’s Rubin may appear on the surface to be a simple “chip replacement with an added midplane,” but its deeper industrial meaning lies in rewriting the standards of server hardware design. This shift is reshuffling the positions of ODMs, PCB and materials suppliers, and cooling module providers, ultimately pushing traditional own brands toward the margins.

Table 2. Potential Value Chain Shifts under the Rubin Architecture

Why Is NVIDIA Doing This?

Rubin’s structural design has become larger and more complex, not simply because of higher power consumption or performance requirements. Behind it lies a clear strategic choice.

NVIDIA’s goal is not only to launch a faster GPU but also to redefine the design rules of the server itself, gaining deeper control over both the supply chain and the broader value network of the industry.

There are at least three motivations behind this decision:

1. Reducing the Differentiation Space of OBMs and ODMs to Keep Design Control in Its Own Hands

In the past, brand-name server makers retained some degree of control over design and system integration, while NVIDIA served mainly as an upstream GPU supplier.

Rubin’s design fundamentally shifts this balance. Through the introduction of the midplane, high-speed PCB, cooling modules, and rail-based structure, NVIDIA now defines the core specifications of the server itself. ODMs and OBMs may take on more assembly and testing work, but the design language no longer belongs to them.

This change turns NVIDIA from a component provider into the authority that standardizes the entire server design framework.

2. Increasing Switching Costs to Keep Cloud Providers Within NVIDIA’s Server Framework

Another purpose behind Rubin’s complex server design is to raise the barriers for customers to move to other platforms. When cooling systems, PCBs, and modular architectures are all deeply tied to the Rubin platform, cloud providers that want to deploy their own ASICs—such as Google’s TPU or AWS’s Trainium—would need to rebuild an entirely new cooling and system infrastructure. The investment required makes such a transition almost impossible, effectively creating an invisible lock-in.

In this sense, Rubin is not just a hardware product but a standard that makes it difficult for customers to step outside NVIDIA’s ecosystem.

3. Evolving from a GPU Maker into the Standard Setter for AI Factories

The deeper intent behind these design choices is to elevate NVIDIA’s role from a chip supplier to a system architect for AI infrastructure. Rubin is not simply a faster GPU; it is a complete design language that defines how motherboards, cooling systems, modular structures, and maintenance are built and integrated.

This shift signals a broader transformation in the industry, from competing on chip performance to competing on who can set the rules. Once NVIDIA controls those rules, ODMs, OBMs, and even cloud providers must operate within its framework.

Conclusion

On the surface, Rubin appears to be another GPU upgrade with higher performance, greater power consumption, and updated memory. Yet the real turning point lies beyond the chip itself.

As motherboard area expands by 30 to 50 percent, cold plates double in number, and dual-layer PCBAs with modular rails become standard, NVIDIA is quietly rewriting the rules of server design.

These structural changes are driving a new wave of hidden growth for PCB and materials suppliers, elevating liquid cooling into an essential part of the value chain, and moving ODMs from behind-the-scenes manufacturers to central players, while reducing the influence of traditional brand-name server makers.More importantly, they raise the switching costs across the industry, making it difficult for cloud providers to break away from NVIDIA’s specifications even if they develop their own ASICs.

The true meaning of Rubin, therefore, lies not in the chip upgrade itself but in who defines the language of system design. It represents a subtle yet profound shift in power.

When we focus only on performance and chip specifications, we risk overlooking the most important truth: in the competition for AI infrastructure, what ultimately determines the winner is not just compute power but the ability to set the rules.

Note: AI tools were used both to refine clarity and flow in writing, and as part of the research methodology (semantic analysis). All interpretations and perspectives expressed are entirely my own.