Morgan Stanley's latest research argues that the AI infrastructure buildout is rapidly shifting its center of gravity from downstream GPU makers and foundries to the critical upstream materials that make the entire stack possible. This transformation is not only redrawing the supply chain maps for printed circuit boards (PCBs) and multi-layer ceramic capacitors (MLCCs), but it is also ushering in a high-conviction, long-duration supercycle for raw materials.

The investment bank's metrics show that AI will contribute nearly all of the industry's net growth over the next five years. Morgan Stanley estimates the global PCB market will expand from roughly $58 billion in 2025 to about $135 billion by 2030—a compound annual growth rate (CAGR) of 18%. Within that, AI and data center demand is projected to surge more than sixfold, from around $13 billion to $86 billion, vaulting its share of industry revenue from 20%–25% to a dominant 64%. In other words, by 2030, essentially the entire net increase in PCB demand will come from AI, while other end markets remain flat.

The copper-clad laminate (CCL) trajectory is even steeper. Morgan Stanley forecasts the global CCL market will grow from roughly $19 billion to $47 billion (a CAGR of about 20%), with AI and data center CCL demand skyrocketing over sevenfold from roughly $4 billion to about $30 billion (a CAGR of approximately 47%). That single segment is expected to account for about 90% of the industry's net growth.

As the Nvidia platform migrates from Blackwell's M8 grade to Rubin's M8.5, and onward to the anticipated M9/M10 of Feynman, the material content and value per unit continues to climb. Manufacturing complexity is also transforming: the end-to-end lead time for an advanced AI PCB is roughly 100 days—2.2 times longer than the 46 days required for a standard multilayer board. An Nvidia-grade computing board requires 22 layers and five lamination passes, whereas a conventional 12–16 layer board needs only one.

Three materials are in the tightest supply. Topping the list is high-end fiberglass fabric. Constraints from Toyota Industries' limited weaving machine output (just 300 units shipped monthly by end-2027), tight supply of premium fiberglass yarn, and formidable technical barriers have already doubled Chinese fabric prices this year, with inventories at extreme lows. Morgan Stanley expects a supply deficit of roughly 40% in 2026, worsening further in 2027 before easing to around 30% by 2028. Japanese fabric prices have climbed from 3,274 yen per kilogram in 2023 to 4,589 yen in 2025, a 40% increase. In response, Nittobo has raised its cumulative capital expenditure for fiscal 2025–2028 from 80 billion yen to roughly 120 billion yen.

The second-tightest material is HVLP4+ ultra-low-profile copper foil. According to CBC Metals, HVLP foil demand for AI servers will jump 260% year-over-year in 2026 to 24,000 tons, and then rise another 108% to about 50,000 tons in 2027. Chinese copper foil utilization rates already exceed 93%. Morgan Stanley's model shows a supply deficit of roughly 31% in 2027 (monthly demand of 3,270 tons against capacity of just 2,495 tons), with an approximate 20% shortfall still expected in 2028. Supply is highly concentrated—Mitsui Mining & Smelting alone controls 41% of global capacity in 2026, making it the primary beneficiary of both volume and price gains.

Third on the list is optical fiber, now at seven-year price highs. Hyperscale AI data centers consume 5–10 times more fiber than traditional facilities, with some estimates showing 5–36 times more per rack. The AI-driven data center fiber demand CAGR is expected to exceed 20% from 2025–2030, with data center-related demand surging 69% year-over-year in 2026. Corning has committed to increasing its US fiber capacity by 50% and expanding optical connectivity capacity tenfold—yet Lumen alone has already locked up 10% of its global capacity for 2025–2026.

The recent pullback, in Morgan Stanley's view, offers a second entry point. While AI materials stocks have risen 91% over the past 12 months, they still underperformed the PCB/CCL complex, which gained 147% in the same period. Moreover, the group has corrected roughly 36% on average from its May–June 2026 52-week highs. At current levels, the sector trades at about 26 times 2026 expected earnings versus roughly 35 times for PCB/CCL, leaving room for the valuation gap to narrow. Morgan Stanley stresses that "the next phase is about content, not volume"—its 2030 CCL market forecast of approximately $47 billion is significantly above the consensus range of $30–35 billion, reflecting an underappreciated increase in material value per unit. The capital expenditure backdrop remains robust, with Morgan Stanley projecting cash cloud capex of around $1.2 trillion in 2026 (+104% year-over-year) and $1.6 trillion in 2027 (+38%).

Morgan Stanley's preferred picks in the space are Nittobo, Grace Fabric, Mitsui Mining & Smelting, Co-Tech, Corning, Furukawa Electric, Mitsubishi Gas Chemical, and Denka. Resin systems (PPE/OPE, BMI, cyanate ester) are structural beneficiaries but carry lower scarcity premiums. In MLCCs, Murata's April–June book-to-bill ratio hit 1.47—the highest since 2004—with utilization at about 95%. Longer-dated optionality lies in synthetic diamond (the Rubin platform's 2,300W power draw exceeds traditional copper cooling limits; diamond-copper composites achieve 600–1000 W/m·K thermal conductivity, while high-purity CVD diamond reaches 2000–2400), the tungsten-to-molybdenum transition in 3D NAND, and rare earths plus scandium oxide for SOFC-based data center power solutions.

Key risks include mid-term capacity expansion in optical fiber that could normalize supply-demand dynamics and pressure prices; the Toyota Industries bottleneck remaining unresolved in the near term, constraining fiberglass fabric expansion; and MLCC dielectric powder price increases that face difficult negotiations downstream, potentially squeezing component margins.

Morgan Stanley's core thesis is that downstream AI hardware valuations already price in most of the narrative. The most significant upside surprise lies in upstream materials—especially premium fiberglass fabric, HVLP copper foil, and optical fiber—where supply tightness is expected to persist through at least 2028. That is where the greatest expectation gap of this supercycle resides.