From "Chasing the Light" to "Moving Upstream of the Light"! Morgan Stanley Asserts: Under the Computing Power Tsunami, Fiberglass Fabric and Copper Foil Set Off a Materials Supercycle

Morgan Stanley has released its latest research report, pointing out that the frenzied expansion of global AI infrastructure is spreading comprehensively from downstream GPUs and wafer foundries to upstream key basic materials. This trend is not only reshaping the supply chain systems of printed circuit boards (PCBs) and multilayer ceramic capacitors (MLCCs), but is also giving rise to a "supercycle" in upstream materials with a long duration and high certainty. AI accounts for almost all of the industry's net growth Morgan Stanley estimates that the global PCB market will expand at an 18% CAGR from about $58 billion in 2025 to about $135 billion in 2030; among this, AI/data center-related demand will grow more than sixfold, from about $13 billion to $86 billion, with its share of industry revenue jumping from 20%25% to 64%. In other words, by 2030, AI and data centers will contribute almost all of the industry's net incremental growth, while other end demand remains basically flat. The slope for copper-clad laminate (CCL) is even steeper: the global market will grow from about $19 billion to $47 billion (CAGR of about 20%), while AI/data center CCL demand will surge more than sevenfold from about $4 billion to about $30 billion (CAGR of about 47%), contributing about 90% of the industry's net growth. Nvidia's platform migration from Blackwell's M8 grade to Rubin's M8.5 and then to the expected M9/M10 for Feynman continues to push up material usage and value per unit. Manufacturing complexity is also undergoing a qualitative change: the end-to-end lead time for an advanced AI PCB is about 100 days, 2.2 times that of an ordinary multilayer board (46 days); Nvidia-grade compute boards require 22 layers and five lamination cycles, while conventional 1216 layer boards require only one lamination cycle. The three most scarce materials Ranked first: high-end fiberglass fabric. Constrained by three factorslimited Toyota loom supply (monthly shipments of only 300 units by the end of 2027), tight high-end fiberglass yarn, and technical barriersChinese fiberglass fabric prices have already doubled during the year, with inventories at extremely low levels. Morgan Stanley expects a supply gap of about 40% in 2026, worsening further in 2027, and easing only to about 30% in 2028. The average price of Japanese fiberglass fabric has risen from 3,274 yen/kg in 2023 to 4,589 yen/kg in 2025, a gain of 40%. Nittobo has sharply raised its cumulative capital expenditure for fiscal 20252028 from 80 billion yen to about 120 billion yen. Ranked second: HVLP4+ ultra-low-profile copper foil. According to CBC Metals, demand for HVLP copper foil for AI servers will grow 260% year over year in 2026 to 24,000 tons, and rise another 108% in 2027 to about 50,000 tons; China's copper foil industry operating rate has already exceeded 93%. Morgan Stanley's model shows a supply-demand gap of about 31% in 2027 (monthly demand of 3,270 tons versus capacity of only 2,495 tons), and still about 20% in 2028. Supply is highly concentrated, with Mitsui Mining alone accounting for 41% of global capacity in 2026, making it the biggest beneficiary in both volume and price. Ranked third: optical fiber, with prices hitting a seven-year high. Fiber usage in hyperscale AI data centers is 510 times that of traditional data centers, with some estimates showing usage per rack is 536 times higher; AI-driven data center fiber demand is expected to have a CAGR of more than 20% from 20252030, and data center-related fiber demand will grow 69% year over year in 2026. Corning has committed to increasing U.S. fiber capacity by 50% and expanding optical connectivity capacity tenfold, while Lumen alone has locked up 10% of its global capacity for 20252026. Valuation: the pullback offers a second chance The AI materials sector has risen 91% over the past 12 months, but still underperformed the PCB/CCL sector's 147% over the same period; moreover, it has retraced an average of about 36% from its 52-week high in MayJune 2026. The sector currently trades at about 26x expected 2026 earnings, versus about 35x for PCB/CCL, leaving room for the valuation gap to narrow. Morgan Stanley emphasizes that "the next phase is content, not volume": its 2030 CCL market forecast of about $47 billion is significantly above the market consensus of $3035 billion, reflecting that the increase in material value per unit is still underestimated. The capex backdrop remains strongMorgan Stanley expects cash cloud capex of about $1.2 trillion in 2026 (+104% year over year) and about $1.6 trillion in 2027 (+38%). Portfolio and long tail Morgan Stanley's top picks are: Nittobo, Grace Fabric Technology (Grace Fabric), Mitsui Mining, Co-Tech, Corning, Furukawa Electric, Mitsubishi Gas Chemical and Denka. Resins (PPE/OPE, BMI, cyanate ester) are structural beneficiaries but less scarce; in MLCCs, Murata's AprilJune order-to-bill ratio reached 1.47, the highest since 2004, with a utilization rate of about 95%. Longer-term options lie in: synthetic diamond (Rubin's power consumption reaches 2,300W, traditional copper cooling is no longer sufficient, diamond-copper composite materials can achieve thermal conductivity of 6001,000 W/mK, and high-purity CVD diamond even reaches 2,0002,400), material substitution for tungsten and molybdenum (3D NAND shifting from tungsten to molybdenum), and rare earths and scandium oxide (SOFC data center power supply solutions). Risk warning Aggressive mid-term capacity expansion in optical fiber may normalize supply and demand and suppress prices; the Toyota loom bottleneck has no short-term solution, constraining fiberglass fabric capacity expansion; MLCC dielectric powder price increases still require difficult negotiations with downstream customers and may squeeze component profits. Conclusion Morgan Stanley's core judgment is that valuations for downstream AI hardware have already priced in most of the story, while supply shortages in upstream materialsespecially high-end fiberglass fabric, HVLP copper foil and optical fiberwill persist at least until 2028. That is where the biggest expectation gap in this supercycle lies.