Morgan Stanley: NVIDIA's Rubin Platform Drives 166% Increase in MLCC Value Per Rack, Market to Reach $44.5B by 2031
nashnova research
Morgan Stanley expects the AI-server MLCC market to hit $23.3 billion by 2031 — more than double its prior forecast — signaling that AI investment dollars are spreading from GPUs into tiny capacitors most investors have never thought about.
What are MLCCs, and why do they suddenly matter?
MLCCs — multi-layer ceramic capacitors, tiny components that store and release electrical energy on circuit boards — have long been among the most overlooked parts in electronics.
Morgan Stanley projects the global MLCC market will reach $44.5 billion by 2031, a 20.3% CAGR from 2025. This means → a traditionally "quiet" component market is expanding at near-AI-chip speed.
The AI server and data-center sub-segment grows even faster: $23.3 billion by 2031, a 57.1% CAGR, revised up from the prior $10.8 billion forecast.
How much does Nvidia's Rubin platform change the equation?
Nvidia's next-gen Rubin platform (VR200 NVL72) requires roughly 570,000 MLCCs per rack — about 80% more than the GB300 NVL72's ~320,000.
The dollar jump is even steeper: per-rack MLCC value rises from $4,664 to ~$12,411, an increase of roughly 166%. This means → capacitor cost alone nearly triples inside each rack.
In plain terms = as chips grow more powerful, the tiny capacitors that feed and stabilize their power must multiply — and their price tag multiplies with them.
How extreme is the demand for high-capacitance MLCCs?
In the Rubin platform, MLCCs rated above 47µF rise from 18% of the mix to 31%; some compact high-capacitance units cost 5–10× the price of conventional parts.
Cloud-AI demand for 47µF-plus MLCCs is projected to surge from roughly 4 billion units in 2025 to over 40 billion by 2027 — a 10× jump in two years.
This reflects a broader pattern: AI hardware upgrades are not just about swapping in bigger GPUs — every link in the supply chain is scaling up in lockstep.
Why can't competitors easily muscle in?
High-end AI MLCCs require stacking hundreds — sometimes over 1,000 — dielectric layers while holding defect rates extremely low and passing rigorous customer qualification. The technical barrier is steep.
Murata and Samsung Electro-Mechanics together hold roughly 85% of the high-end AI MLCC market, forming a de facto Japan-Korea duopoly.
This means → unlike the 2017–18 conventional MLCC shortage, new entrants cannot simply build a factory and grab share. In plain terms = last time, capacity solved the problem; this time, the technology threshold is too high, and the bottleneck could become a long-term supply-security issue.
Why is Murata reluctant to lock in long-term contracts?
Murata's management recently said rising production loads on high-end products are making it increasingly hard to meet low-end demand.
Morgan Stanley notes that Murata has been relatively cautious about signing long-term fixed-supply agreements for high-value AI and data-center MLCCs. This means → Murata likely believes prices still have room to rise and locking in now would leave money on the table.
This reflects pricing power concentrating at the top of the supply chain — buyers scramble for capacity while sellers are in no rush to commit.
What are Morgan Stanley's ratings and targets?
The bank maintains Overweight on Murata, Samsung Electro-Mechanics, and Yageo; it upgrades Taiyo Yuden from Underweight to Equal-weight.
Implied upside from September 11 closing prices: Murata ~50% (target ¥11,000, cut from ¥12,500), Samsung E-M ~87%, Yageo ~93%, Taiyo Yuden ~14%.
Morgan Stanley sees the recent pullback in MLCC-concept stocks as restoring value in select names, but the key test is whether Murata and Samsung E-M can convert product-mix upgrades and higher ASPs into real earnings leverage.
市场有风险，内容仅供研究参考，不构成投资建议。