On October 29, 2026, market research firm TrendForce announced its forecast that the shortage of high-capacitance MLCCs (multilayer ceramic capacitors) for AI servers will continue into the first half of 2027. AI use is expanding beyond training to inference and AI agents, which is increasing demand for CPUs. According to TrendForce, demand for high-capacitance MLCCs is also swelling around CPU power circuits and memory. At Murata Manufacturing, MLCC orders alone reached 1.47 times sales in the April–June 2026 quarter, and NVIDIA began volume shipments of its next-generation AI platform, Vera Rubin, in August. Meanwhile, Murata's planned increase of more than 20% in MLCC production capacity will be carried out over two years, through fiscal 2027. The prolonged shortage can be explained without assuming that "capacity won't grow until 2028": the timing gap between the sudden surge in demand and the capacity expansion is enough.
AI inference growth boosts CPUs, and MLCC demand for power circuits and memory surges
What TrendForce highlighted in this announcement is demand growth that starts with CPUs rather than GPUs. According to the firm, as AI inference and AI agent workloads increase, demand for CPUs is rising, and so is the number of power modules and memory devices the CPUs require. As a result, demand for high-capacitance MLCCs is expanding not only on CPU motherboards but also on memory modules, putting pressure on the supply of standard high-capacitance products. TrendForce had covered the order-to-shipment ratios of Japanese and Korean manufacturers in July and Samsung Electro-Mechanics' (SEMCO) price increases in August, but this time it stresses that demand growth has reached the areas around CPUs and memory.
The expansion in CPU demand can also be confirmed in semiconductor makers' earnings. In her earnings call on August 26, NVIDIA CFO Colette Kress said the spread of agentic AI is accelerating demand for data center CPUs, and revealed that the next-generation Vera CPU has entered full-scale production. She also gave a preliminary outlook that CPU revenue in fiscal 2028 (roughly February 2027 to January 2028) would more than double. AMD CEO Lisa Su, in the August 4 earnings release, likewise indicated that demand for EPYC would accelerate in the second half and that the Helios launch would begin alongside broader Instinct deployment.
The reason MLCC demand rises on the memory side lies in the change to power circuit architecture with DDR5. According to Intel's technical documentation for FPGAs, in DDR5 RDIMMs (server memory modules), the power management functions that were traditionally handled on the motherboard have been moved to a PMIC (power management IC) on each memory module. This PMIC integrates four step-down switching regulators and three LDOs (low-dropout linear regulators). In other words, voltages previously generated on the motherboard are now generated on each memory module. When the power circuitry moves to the module, the capacitors and other components it requires are also mounted on the module. This change in power circuit configuration is likely behind TrendForce's point about rising demand for high-capacitance MLCCs for memory modules.
However, growth in CPUs does not necessarily translate directly into growth in DDR5 RDIMM demand. For example, NVIDIA's Vera CPU uses an 88-core Arm-based design and carries up to 1.5TB of LPDDR5 or LPDDR5X memory in SOCAMM form, so it does not use DDR5 RDIMMs. MLCC demand growth via DDR5 RDIMMs applies mainly to x86 servers built on AMD EPYC or Intel Xeon. NVIDIA's CPU ramp supports the picture of expanding data center CPU demand, but TrendForce's announcement does not make clear how MLCC usage around Vera CPU memory will specifically increase.
On the GPU side, too, procurement of high-capacitance MLCCs is becoming more active. According to TrendForce, AMD has begun procuring specialized high-capacitance MLCCs for its MI450 platform. In its August earnings release, AMD described a partnership with Anthropic to deploy up to 2 gigawatts of MI450-series GPUs using Helios, a rack-scale system built around the MI450 series. Simultaneous growth in demand for high-capacitance MLCCs across multiple uses—CPUs, memory and GPUs—is making the shortage more severe.
Demand surged in summer 2026, while Murata's capacity expansion continues through fiscal 2027
In Murata Manufacturing's fiscal 2026 first-quarter (April–June) results announced on July 31, net sales rose 20.7% year on year to ¥502.3 billion, while orders reached a record ¥673.9 billion. The order backlog grew to ¥617.8 billion, and data center-related sales rose 81.1% year on year to ¥69.7 billion. In the earnings briefing, Executive Vice President Masanori Minamide said orders were very strong, led by data centers, and disclosed that the book-to-bill ratio for MLCCs alone had reached 1.47. The company also expects full-year data center-related sales of ¥370.6 billion, up 109.7% from the previous fiscal year.
The book-to-bill ratio is the ratio of orders to sales. A value above 1 means orders are coming in at a higher value than what has been shipped and booked as sales, and the difference adds to the order backlog. On a companywide basis, Murata's ratio was 1.34 (orders of ¥673.9 billion divided by sales of ¥502.3 billion), with orders exceeding sales by about ¥170 billion. The MLCC-only figure of 1.47 means that for every 100 of sales, 147 of orders came in. Orders for MLCCs are piling up faster than the company average, and reducing the backlog requires raising shipment volume. That, in turn, makes expanding production capacity indispensable.
The demand growth that became apparent in April–June also showed up in finished-product shipments in August. In her August 26 earnings call, NVIDIA's Kress said that the company began volume shipments of Vera Rubin earlier that month (original: "We commenced production shipments of Vera Rubin earlier this month"). She said Vera Rubin would be the fastest ramp in the company's history and expected it to account for about 20% of data center revenue in the third quarter, which covers late July to late October 2026. According to the earnings release, Vera Rubin racks are already operating at CoreWeave, Google Cloud, Microsoft Azure, Oracle Cloud Infrastructure and Nebius. TrendForce forecasts that Vera Rubin shipments will increase month by month from October–December 2026, and Kress has also indicated that NVIDIA's supply constraints will continue at least through the end of fiscal 2028.
By contrast, expanding MLCC makers' production capacity takes time. According to Taiwan's Economic Daily News (via UDN), citing an interview with Nikkei, Murata plans to invest an additional roughly ¥80 billion in MLCC production equipment over the two years through fiscal 2027, increasing capacity by more than 20%. However, this rate of increase covers MLCCs as a whole, and the rate of capacity growth for high-capacitance products specifically for AI servers has not been disclosed.
Capacity expansion across the industry has not kept pace with demand either. South Korea's ETNews reported in August that lead times for high-end MLCCs had stretched to 5–10 months or more, and that new capacity originally scheduled to come online in October–December 2026 had slipped into 2027. Data from major electronic component distributor DigiKey also shows lead times of about 40 weeks for some of SEMCO's high-capacitance MLCCs.
The timeline below organizes the growth in demand and the capacity expansion moves. NVIDIA's fiscal quarters and Murata's fiscal years have been converted to calendar months for easier comparison.
| Period (calendar) | Event | Entity / Source |
|---|---|---|
| April–June 2026 | Murata's MLCC-only book-to-bill ratio reaches 1.47 | Murata Manufacturing (EVP Minamide's earnings briefing) |
| End of June 2026 | Book-to-bill ratios reach 1.30 for Murata, 1.31 for SEMCO and 1.25 for Taiyo Yuden, the highest since the pandemic. The industry overall is 1.04 | TrendForce tally (compiler and definition differ from Murata's disclosure) |
| August 2026 | Volume shipments of Vera Rubin begin | NVIDIA (CFO Kress) |
| August 2026 | Lead times for high-end MLCCs stretch to 5–10 months or more | ETNews (cited by TrendForce News) |
| October–December 2026 | SEMCO raises MLCC prices for OEMs/ODMs | TrendForce estimate (announced August 27) |
| November 2026 | Price negotiations for January–March 2027 begin | TrendForce |
| January–June 2027 | Period when ODMs expect the high-end MLCC shortage to be most severe | TrendForce survey |
| Full year 2027 | SEMCO supplies MLCCs for AI servers under a ₩1.07 trillion contract | Korea Herald |
| March 2028 | Fiscal 2027, the period covered by Murata's 20%+ capacity expansion plan, ends | Murata (reprint of Nikkei interview) |
| March 2028 | Final order deadline for some general-purpose products Murata is discontinuing | Korea Herald (industry sources) |
ODMs (original design manufacturers) surveyed by TrendForce expect the high-end MLCC shortage to be most severe in January–June 2027. While volume shipments of Vera Rubin began in August 2026, Murata's plan to raise capacity by more than 20% continues through the end of fiscal 2027 (March 2028). In other words, the peak of the shortage forecast by ODMs will arrive while Murata is still in the middle of expanding capacity.
On the demand side, moves are also under way to secure the supply needed in the meantime through long-term contracts. According to the Korea Herald, SEMCO signed a contract worth a total of ₩1.07 trillion (about $774 million) with global customers to supply MLCCs for AI servers throughout 2027. It is the largest MLCC contract in the company's history, and it reportedly has long-term supply agreements with more than 10 counterparties. The move suggests customers are trying to lock in future requirements early amid the expected shortage.
Will Murata's expansion really not materialize until 2028?
In addition to the timing gap between demand and supply, TrendForce's announcement contains a statement about Murata's production capacity that raises questions. The firm said that although Murata is bringing forward its expansion plans, new capacity is unlikely to come online before 2028, so supply will not improve significantly for the time being. As its basis, it cited Murata executives acknowledging that existing capacity cannot fully meet market demands. Read on its own, this explanation could be taken to mean that Murata's capacity will not increase at all during 2027.
However, the plan described by Murata separates capacity increases at existing sites from the construction of future new plants. According to what Minamide reportedly told Nikkei, the company will spend about ¥80 billion on capital investment over the two years through fiscal 2027 and raise production capacity by more than 20%. Meanwhile, what it is considering to meet demand from 2028 onward is a large new factory whose construction has not yet been decided. According to the same reprinted article, the buildings needed through 2028 have already been prepared ahead of schedule, and the new factory is being considered at existing domestic and overseas sites. Minamide has said that demand for AI-related MLCCs will continue to expand at least through 2028.
Whether the two accounts conflict depends on the scope of what TrendForce means by "new production capacity." If it refers to the capacity of a newly built large factory, it is compatible with Murata's explanation. This reading is that capacity increases in stages through fiscal 2027 by installing equipment in existing buildings, while the company also considers building a large new factory for demand from 2028 onward. If, on the other hand, it refers to overall production capacity including equipment additions in existing buildings, it conflicts with Murata's plan to increase capacity by more than 20% by fiscal 2027. TrendForce's announcement alone does not make clear which it means.
Still, the outlook that the shortage will continue into the first half of 2027 holds under either interpretation. If capacity at existing facilities is increased in stages through fiscal 2027, January–June 2027, when the shortage is expected to be most severe, would still fall in the middle of the expansion. If the start-up of new capacity is delayed further, supply constraints would last even longer. Therefore, there is no need to assume that capacity won't grow until 2028; the time lag before production increases catch up with rising demand is enough to explain why the shortage continues.
Murata is also reviewing how it allocates its limited production capacity. According to the Korea Herald, citing industry sources, Murata has notified customers that it will discontinue certain part numbers across nine product series. The targets are general-purpose MLCCs for consumer and industrial equipment and some automotive products, with a final order deadline of March 2028 and final shipments one year later. Murata reportedly explained that it is revising its product mix to redirect capacity to other products. It is unclear how much capacity this will free up, but it appears the company is trying to shift existing lines to products with strong demand, such as those for AI, while waiting for new equipment to come online.
Murata also raised its fiscal 2026 capital expenditure plan by ¥5 billion to ¥255 billion. However, this is companywide capex and not an amount devoted solely to MLCCs. The second-quarter results to be announced later may provide new information on the scale and timing of capacity expansion.
What's behind the 24.5% rise in import prices, and what it means for MLCC materials
TrendForce also cited rising raw material and logistics costs as factors adding to the burden on the supply side. According to the firm, Japan's import price index rose 24.5% year on year in September, and procurement costs for nickel, copper, silver and other materials used in MLCC electrodes have also increased. However, checking the Bank of Japan's corporate goods price index (September 2026 preliminary), released October 13, the 24.5% figure is the yen-based overall average for all import price items and does not represent the price increase for MLCC materials alone.
| Series (September 2026 preliminary) | Year-on-year, yen basis | Year-on-year, contract-currency basis |
|---|---|---|
| Import prices, overall average | Up 24.5% | Up 17.9% |
| Metals and related products | Up 32.4% | Up 25.2% |
The 24.5% rise in import prices cited by TrendForce is an overall average of all items converted into yen; on a contract-currency basis, the increase is 17.9%. Meanwhile, "metals and related products," which include MLCC electrode materials, rose 32.4% on a yen basis and 25.2% on a contract-currency basis. Energy-related items did much to push up the overall average, with "petroleum, coal and natural gas" up 43.6% on a yen basis. "Electric and electronic products" also rose 29.5%.
The difference between the yen basis and the contract-currency basis lies in whether currency movements are reflected. On a contract-currency basis, for a dollar-denominated transaction, for example, the index measures the change in the dollar transaction price. On a yen basis, that price is converted into yen, so if the yen weakens, the yen-converted import price rises even if the dollar price is unchanged. The yen-based rate of increase in import prices therefore includes the effect of exchange rate movements in addition to price increases overseas.
Calculating from the change rates of the overall-average indices on a yen basis and a contract-currency basis, 1.245 ÷ 1.179 − 1 comes to roughly 5.6%. This is a rough estimate of the yen-converted price effect compared with a year earlier, suggesting that currency movements account for about this much (editorial calculation; the Bank of Japan does not directly show the contribution of exchange rates). Meanwhile, "metals and related products" rose 25.2% even on a contract-currency basis, showing that overseas prices have risen sharply even excluding the effect of the weaker yen.
However, "metals and related products" is a classification that bundles multiple items, and the price change for nickel alone cannot be confirmed from the Bank of Japan's preliminary release. Also, in the September month-on-month change on a contract-currency basis, the main contributors to the rise in this category were copper ore, silver bullion and molybdenum ore. The import price index is not the actual procurement price paid by MLCC makers, and the September figures are preliminary and may be revised.
Rising raw material costs are also affecting manufacturers' earnings outlooks. Murata revised its fiscal 2026 full-year forecast upward to net sales of ¥2.11 trillion and operating profit of ¥430 billion, but the increase in operating profit was only ¥50 billion. EVP Minamide cited soaring material costs, including precious metals, and higher fixed costs as the reason. The Bank of Japan's statistics and Murata's earnings briefing differ in scope and method of aggregation, but both confirm that cost pressure is intensifying. With orders far exceeding shipments and raw material costs also rising, Japanese MLCC makers have reasons to seek price increases on both the supply-demand and cost fronts.
Will major Japanese MLCC makers also raise prices in November negotiations?
SEMCO has led the way on MLCC price increases. On August 27, TrendForce indicated that SEMCO would lead price increases for OEMs/ODMs in October–December 2026. The estimated increases reach an average of 25–30% for X5R products for consumer devices and an average of 10–20% for X6S high-end products for AI servers. However, actual prices vary depending on the terms negotiated with each customer. At that time, TrendForce analyzed that the three Japanese makers—Murata Manufacturing, Taiyo Yuden and Kyocera—were holding OEM/ODM prices steady while watching market trends.
Since then, the differences in supply and demand by application in the MLCC market have become sharper. According to TrendForce, major notebook PC ODMs expect orders in October–December 2026 to decline an average of 10–15% from a year earlier, and smartphone makers are also reducing shipments of entry-level and mid-range models. Against this backdrop of weak demand, spot prices for low- to mid-capacitance MLCCs for consumer devices fell about 5% in early October before stabilizing for the time being. Price momentum for high-end MLCCs has also begun to slow, but supply constraints for AI remain strong. TrendForce says the shortage has also spread to MLCCs for optical communication modules.
TrendForce forecasts that the high-end MLCC shortage will continue into the first half of 2027 and that manufacturers will retain the upper hand in price negotiations. Negotiations for January–March 2027 are scheduled to begin in November 2026, and the firm said it is increasingly likely that Japanese makers such as Murata Manufacturing will follow SEMCO in raising prices. However, because demand is weak for low- to mid-capacitance products for consumer devices, price increases will not necessarily advance uniformly across all MLCCs.
As a concrete production increase plan that could ease the shortage, Murata has indicated a policy of raising production capacity by more than 20% over the two years through fiscal 2027. SEMCO's full-year 2027 long-term supply contract, meanwhile, does not increase production capacity itself; it reflects customers moving early to secure future supply volumes. While Murata's production increase proceeds in stages through fiscal 2027, negotiations over supply volumes and prices in the meantime will get fully under way from November.
There are three main points to watch. First, whether Japanese makers raise OEM/ODM MLCC prices in the November negotiations. Second, how Murata's second-quarter results update order trends and its MLCC capital investment plans. Third, whether NVIDIA's third-quarter results show Vera Rubin reaching the projected share of about 20% of data center revenue. These developments will show how long demand growth continues and how far manufacturers can raise supply capacity. What determines when the shortage of high-capacitance MLCCs eases is the point at which makers such as Murata secure enough production capacity relative to growth in demand.