Rubin Rack BOM Breakdown: Sharp Increase in PCB and MLCC Content, Leading to a Substantial Boost in ODM Gross Profit Margins
AI summary card
Rubin Rack BOM Breakdown: Sharp Increase in PCB and MLCC Content, Leading to a Substantial Boost in ODM Gross Profit Margins
Morgan Stanley estimates Nvidia’s Rubin rack ASP at approximately $7.8 million, with the most significant increases in component content observed for PCBs (+233%) and MLCCs (+182%). Although ODM gross margins have declined slightly, the absolute value added per unit is expected to rise by 35–40%.
- Rubin’s rack ASP is estimated at approximately $7.8 million, with memory costs accounting for 25–30% of the total, while GPU costs have declined to 51%.
- PCB revenue is expected to surge by 233% to USD 117,000, primarily driven by an increase in layer count and the introduction of new modules.
- MLCC content is expected to surge by 182% to $4,320, as robust demand at the high end is prompting ODMs to ramp up inventory.
- ABF substrate content increased by 82%, while the value of power supplies (+32%) and liquid cooling systems (+12%) also rose in tandem.
- The value added per ODM standalone unit is expected to increase by 35–40%. Although the gross margin has declined from 2.7% to 1.9%, the absolute profit level has risen.
- Some projects have shifted to a consignment model, which helps alleviate ODM working capital pressures.
- Our top pick among ODM players is Wiwynn, followed by Wistron, Quanta, and Hon Hai; we are also bullish on component manufacturers such as Delta and AVC.
Report interpretation
Overview
This research report provides a bottom-up, detailed bill-of-materials (BOM) analysis of Nvidia’s next-generation AI server platform, Rubin (VR200 NVL72). The key takeaway is that the average selling price (ASP) of the Rubin rack is projected to be approximately $7.8 million; this figure would decline to $6.7 million if procured directly by hyperscale cloud providers through SOCAMM. Despite market concerns that standardization could erode ODM (original design manufacturer) value, Morgan Stanley’s analysis indicates that, driven by higher compute density, power density, and assembly complexity, the absolute value added by ODMs on the Rubin rack will increase by 35–40% compared with the Blackwell series. Furthermore, the report quantifies changes in the component mix of downstream key parts, with PCBs and MLCCs showing the most pronounced growth.
Core views
The cost structure of the Rubin rack has undergone a significant shift, driven primarily by rising memory prices and increased memory content. According to our research report, memory now accounts for 25–30% of the VR200 rack’s bill of materials (BOM), up sharply from 5–10% in the GB200 era. This change has reduced the GPU’s share of total costs from roughly 65% to 51%. Such structural adjustments have resulted in an ASP for the entire rack that is notably higher than that of its predecessor. On the downstream component side, PCBs represent the category with the largest increase in content, projected to rise by 233% compared to the GB300, reaching approximately $117,000 per rack. This growth stems mainly from the introduction of new modules—such as ConnectX and midplane PCBs—as well as upgrades to existing PCB layer counts and copper-clad laminate (CCL) grades. For instance, compute boards have been upgraded from 22-layer HDI to 26 layers, and CCL grades have advanced from M7 to M8. MLCC content is expected to grow by 182%, to $4,320 per rack, underscoring robust demand for passive components in high-end AI servers—explaining why ODMs are aggressively building inventories in anticipation of the Rubin’s ramp-up starting in the second half of 2026. ABF substrate content is forecast to increase by 82%, largely due to a doubling of the ASP per chip—from $100 to $200—and a corresponding surge in the number of NVLink and ConnectX chips deployed. Regarding value capture by ODMs, our analysis refutes the market narrative that “standardization will erode ODM margins.” Instead, we project that the ODM value-added per Rubin rack will expand by 35–40%, rising from approximately $108,000 on the GB300 to around $150,000 on the VR200. This improvement reflects heightened complexity in assembling compute boards, switch boards, cooling components, and rack-level integration, along with additional testing requirements for newly added modules. Although the implied gross margin may decline slightly—from 2.7% on the GB300 to 1.9% on the VR200—the report emphasizes that investors should focus on the absolute dollar‑based profit expansion. Furthermore, the growing trend toward consignment models among ODMs such as Foxconn and Quanta is likely to help alleviate the associated increase in working capital, which we view as a positive factor over the longer term. From a technological evolution perspective, power solutions are transitioning from conventional 110 kW power racks to standalone high-voltage DC (HVDC) power racks. The 800 V DC architecture is expected to be widely adopted in the Rubin Ultra platform beginning in the second half of 2027. In terms of liquid cooling, Rubin racks will fully embrace fanless designs, leveraging expanded tray manifolds, quick-disconnect fittings (QD), and advanced cold plate architectures to elevate the thermal management component value to approximately $72,000 per rack.
Analysis framework
The firm employed a quintessential bottom-up Bill of Materials (BOM) decomposition approach. First, based on supply-chain surveys, it determined the unit prices and quantities of key components—such as GPUs, CPUs, memory, PCBs, and passive parts—and aggregated these to arrive at the rack-level average selling price (ASP). Second, by comparing the specification differences between prior‑generation products (GB200/GB300) and the latest generation (VR200)—including PCB layer counts, chip counts, and thermal management solutions—the firm quantified the percentage changes in the component mix. Finally, factoring in the complexity of ODM assembly and testing processes, it estimated the value added and the corresponding margin trajectory at the ODM stage. This methodology cuts through macro‑level narratives, enabling a direct derivation of financial implications from hardware physical attributes and supply-chain pricing dynamics.
Methodology notes
BOM Content Analysis
By disaggregating the quantity and unit price of each component within a single end-product—such as an AI server rack—we can forecast the revenue growth potential of upstream suppliers. This research report, by comparing the bill of materials (BOM) of new versus legacy product generations, precisely identifies opportunities for simultaneous volume and price increases in specific segments, including PCBs and MLCCs.
Absolute Profit vs. Gross Margin
When evaluating manufacturing companies, analysts focus not only on the gross margin (Margin %) but place greater emphasis on the absolute dollar profit per unit (Absolute Dollar Profit). According to the research report, although the Rubin rack’s high total price has compressed the ODM’s gross-margin percentage, its substantial increase in per-unit value-added has actually boosted the ODM’s absolute profit—information that is critical for assessing the company’s true profitability.
Asset mapping & comparison
Structured mapping from thesis to named assets (strengths, weaknesses, peers, risks).
- Wiwynn (6669.TW)ODM’s Top Picks
- Strengths
- Based on the greatest upside potential of the target price
- Comparison
- Outperforms Wistron, Quanta, and Hon Hai
- Wistron (3231.TW)ODM Secondary Selection Stocks
- Strengths
- Management guidance is consistent with the research report’s view that ODM value-added will increase.
- Comparison
- Second only to Wiwynn, and outperforming Quanta.
- Quanta (2382.TW)ODM’s Third Option
- Strengths
- Mention of the transformation of the consignment model
- Comparison
- Second only to Wistron
- Hon Hai (2317.TW)ODM’s Fourth Choice
- Strengths
- Pioneering the mention of the consignment model
- Comparison
- Second only to Quanta
- Delta Electronics (2308.TW)Component Supplier Preferences
- Strengths
- Participating in the development of an HVDC power supply platform
- Comparison
- Listed alongside AVC, Unimicron, and other companies on the preferred watchlist.
- Unimicron (2368.TW) / ZDT (4958.TW)PCB Supplier Beneficiaries
- Strengths
- Directly benefiting from a 233% increase in PCB content.
Key data
- Rubin Rack ASP (ODM Caliber)Approximately US$7.8 millionIf the cloud vendor procures SOCAMM directly, the cost would drop to approximately US$6.7 million.
- Memory as a share of the BOM25-30%A substantial increase from the 5%–10% level during the GB200 era
- PCB content increase+233%From approximately US$35,000 for the GB300 to approximately US$117,000 for the VR200.
- MLCC content is increasing+182%From approximately $1,530 on the GB300 to approximately $4,320 on the VR200.
- ODM value-added growth+35-40%The absolute amount increased from approximately USD 108,000 to approximately USD 150,000.
- ODM Implied Gross Margin~1.9%Down roughly 2.7% from the GB300 level, yet absolute profit has increased.
Impact & implications
The research report argues that the launch of the Rubin platform will create significant structural opportunities for Taiwan’s technology hardware supply chain. For PCB suppliers such as Unimicron and ZDT, as well as MLCC manufacturers, the rapid growth in component content implies that even with overall market volumes remaining stable, they can achieve revenue growth rates exceeding the industry average. As for ODM players, despite pressure from margin compression, the expansion of their absolute profit pool and the potential broader adoption of consignment models are expected to enhance their long-term profitability and cash‑flow profile. Currently, the ODM sector trades at an average 2027 forward P/E ratio of approximately 13x, which remains attractive relative to its 20‑year historical average of 11.5x—particularly given the rising share of AI‑related revenue.
What to watch
- The mass-production ramp-up schedule for the Rubin rack in the second half of 2026.
- The impact of changes in SOCAMM’s procurement model—whether through Nvidia reselling or direct procurement by cloud providers—on ASPs.
- The penetration rate of the ODM consignment model in real-world projects
- Adoption of the 800V DC power architecture on the Rubin Ultra platform