MLCC Shortage 2026: Why Capacitor Prices Keep Rising

8th Sep 2026

In late August we published a guide to what capacitors and MLCCs are and why AI servers had turned a part that costs a fraction of a cent into something the supply chain fights over. That piece ended with order books at their strongest since the pandemic and lead times on high-end parts as long as ten months.

A lot has happened since. On August 27, TrendForce, the Taiwan-based firm that tracks the passive component market, said the MLCC industry has “officially entered an upcycle.” Samsung Electro-Mechanics started raising prices for the manufacturers it supplies directly, not just distributors. Taiyo Yuden’s second increase of the year took effect September 1. And on September 4, CNBC ran a piece on the capacitor bottleneck as an AI trade in its own right.

This article covers what has happened to MLCC prices this year, why the big suppliers are raising them, what the Goldman Sachs call actually says, and what could end the cycle.

What has happened to MLCC prices in 2026?

Prices are up across the board. The increases started with the scarcest AI-grade parts and spread to ordinary commodity parts over the summer.

The turn came in late February. According to 36Kr, the industry had spent four years, from 2022 into early 2026, working off excess inventory, and distributors were carrying almost no stock. When AI server orders jumped and EV demand picked up, distributors decided prices had bottomed and started buying again. Spot prices in Shenzhen’s Huaqiangbei electronics market began to climb. By late June, TrendForce News reported that quotes on popular parts were changing as often as every 30 minutes. Common parts were up 15 to 20 percent since the spring, and high-capacitance parts for AI servers were up 50 to 60 percent. By July, TrendForce estimated that distributors who had stocked up early were selling at prices 20 to 25 percent higher on average, and some spot quotes had reached two to three times their earlier levels.

Then the manufacturers moved. The sequence below is compiled from TrendForce, DigiTimes, and supplier notices as reported in the trade press.

• April 1. Murata raised prices 15 to 35 percent on AI server and high-end automotive MLCCs. It was the company’s first broad increase in about three years. Bloomberg had reported in February that Murata was considering it.
• May 1. Taiyo Yuden made its first adjustment of the year, covering selected MLCC series plus inductors, RF devices, and aluminum electrolytic capacitors.
• June 1. Walsin Technology raised prices on chip resistors and some MLCCs, citing raw material costs.
• July 1. Yageo raised capacitor prices broadly. Industry sources put the increase at roughly 50 percent on some lines. Walsin raised consumer MLCC prices 5 to 15 percent the same day.
• August 1. Samsung Electro-Mechanics told customers it was raising prices 30 percent on all MLCCs sold through its distribution business. The company said supply chain pressure had gone beyond what it could absorb.
• September 1. Taiyo Yuden’s second increase took effect. It covered fewer product categories than the May round but more of the MLCC lineup. The company cited higher costs for barium titanate powder, nickel powder, and rare earth additives.
• Fourth quarter. Samsung Electro-Mechanics is now extending increases to the OEM and ODM customers it supplies directly. TrendForce estimates increases of 25 to 30 percent on consumer-grade X5R parts and 10 to 20 percent on the high-end X6S parts used in AI servers. Taiwanese and Chinese suppliers are expected to raise prices 10 to 20 percent on mid- to high-capacitance consumer parts.

The trade data say the same thing. Japan’s Ministry of Finance figures, cited in market coverage, showed MLCC export unit prices up 16 percent year over year in April and export value up 28 percent. June export value came in around 72 billion yen, up 31.5 percent from a year earlier, according to figures reported by Investing.com.

Why does TrendForce say an upcycle has started?

It comes down to who is paying more. Through the spring and summer, most of the announced increases hit distributors and the spot market, which serve smaller customers and rush orders. The big OEMs and ODMs, the companies that build servers and phones under long-term contracts, were mostly protected. That changed in late August. In its August 27 note, TrendForce said Samsung Electro-Mechanics’ fourth-quarter increase broke with the company’s practice of limiting price changes to distributors, and called it confirmation that the industry has entered a new pricing upcycle. DigiTimes reported that ABF substrates, the material AI chips are packaged on, are getting more expensive at the same time.

TrendForce’s explanation of the pricing is worth understanding. The bigger increase is on consumer-grade X5R parts, and TrendForce reads it as a way to manage capacity: raise prices on commodity orders to slow them down, then move the freed-up capacity to the higher-margin X6S and X7R grades that AI servers use. Japanese and Korean suppliers have been converting X5R lines to those grades for months. That has left less capacity for the everyday 1 to 22 microfarad parts that go into phones, PCs, and industrial equipment, and channel inventory for those parts has dropped below 30 days. The overflow orders are going to Taiwanese and Chinese suppliers, some of which are now running near 90 percent utilization. Demand for optical modules in AI data centers is adding to the squeeze.

The order data back this up. By late June, book-to-bill ratios at Murata, Samsung Electro-Mechanics, and Taiyo Yuden were 1.30, 1.31, and 1.25, the highest since the pandemic. June shipments from the three hit a five-year monthly high, around 140 billion, 98 billion, and 40 billion pieces. Even so, demand outran supply. TrendForce’s August 18 update put lead times on high-end parts at five to ten months and said the tightness may last into 2027. The firm expects the supply-demand gap to widen again in the fourth quarter.

One thing TrendForce flags: Murata, Taiyo Yuden, and Kyocera have not yet raised their direct-customer prices and are waiting to see how the quarter develops. Whether the Japanese majors follow Samsung Electro-Mechanics is, in TrendForce’s view, the key signal for whether the uptrend spreads across the whole industry.

What did Goldman Sachs say about MLCCs?

Goldman’s line was that MLCCs are “the next memory.” In notes starting in late May, Goldman Sachs analysts argued that capacitors were set to follow the path memory chips took: AI demand soaks up high-end capacity, high-end production crowds out general-purpose parts, and shortages spread to the whole category. The main points from the reports, as covered in the financial press:

Cost. Goldman put MLCCs at number three on an AI server’s bill of materials, behind only the GPU and memory. We noted this in the explainer.

Market size. The bank estimated the total MLCC market at about $15 billion, with AI servers making up about $1.3 billion in fiscal 2025. It expects that segment to grow around 80 percent a year and reach roughly $5.8 billion by fiscal 2030, more than a fourfold increase. In units, Goldman’s reported estimates have AI server MLCC demand at 72.6 billion pieces in 2026, up 87 percent, and 136.7 billion in 2027, up another 88 percent. By 2030 it expects demand to be about 4.3 times the 2025 level.

Supply. Industry capacity has historically grown around 10 percent a year. Suppliers build much of their own production equipment, and a new high-end line usually takes more than a year to set up.

Pricing. Goldman reportedly estimated that high-end, high-capacitance MLCC prices could rise 100 to 150 percent over the following year, with standard parts having 30 to 50 percent of catch-up room. The bank raised its 2026 industry price forecast from roughly flat to a range of zero to 5 percent, and said actual increases could be much larger. The profit sensitivity is high: by Goldman’s math, a 5 percent increase in average selling prices would lift Murata’s fiscal 2027 operating profit by about 13 percent and Taiyo Yuden’s by about 37 percent.

Duration. Goldman’s May 27 note, written after a meeting with Murata’s president, is the one that changed expectations. Murata had previously expected AI infrastructure investment to peak around 2028. According to coverage of the meeting, management now sees the buildout running to around 2030, partly because power shortages are slowing it down, with edge AI uses like software-defined vehicles and humanoid robots as a second wave of demand behind data centers. Murata also explained that it does not usually raise prices on an existing AI server part. It sets new prices when each platform generation changes the bill of materials. Goldman took that as a price reset that works like a price increase. Across its notes, and as summarized by the South China Morning Post, Goldman’s conclusion is that this could be the largest and longest cycle in MLCC history, partly because MLCC prices started rising later than memory and optical components and have further to run.

Other banks have made similar points. A Morgan Stanley teardown of NVIDIA’s Vera Rubin rack, reported by 36Kr, estimated MLCC content per VR200 NVL72 rack at about $4,300, up 182 percent from roughly $1,500 in the GB300 generation, with the biggest jump on the compute boards. TrendForce’s August 28 AI server outlook expects VR200 racks to make up the largest share of NVIDIA’s rack shipments in 2027, with total NVL72 output value above $710 billion that year. If that plays out, the platform that uses the most capacitors per rack becomes the main one.

What is driving the increases?

Three things at once: more capacitors per server, capacity that cannot grow fast, and higher raw material costs.

More capacitors per server. The counts from the explainer still apply: about 6,500 MLCCs on a GB200 compute board, around 12,000 on Rubin, and roughly 440,000 in a full NVL72 rack. AMD swapped every aluminum electrolytic and tantalum capacitor on its MI450 platform for MLCCs during validation. The newer problem is concentration. TrendForce notes that as cloud providers ramp their own accelerators, like Google’s TPU and Amazon’s Trainium, alongside GPU platforms, demand is piling into a small number of premium specs: small case sizes, high capacitance, high temperature ratings. That is harder to fix than a general shortage, because the missing parts are specific ones.

Capacity that cannot grow fast. TrendForce estimates it takes 18 to 24 months to add high-end MLCC capacity. Murata’s new Izumo plant is not expected to reach full output until 2027. Yields on the newest ultra-high-capacitance parts, such as 47 microfarad 0402 and 100 microfarad 0603 X6S components, are still a problem across the industry. Murata has said AI-related orders are running at about twice its capacity, with utilization at 90 to 95 percent. Its baseline plan adds about 10 percent capacity in fiscal 2027 and another 20 to 25 percent in fiscal 2028, and it is trying to pull some of that forward. At Samsung Electro-Mechanics, analysts estimate that two AI server supply contracts signed in June and July, worth a combined 749 billion won (about $515 million) and running through 2027, could take up around 10 percent of the company’s projected 2027 ceramic capacity on a commodity-equivalent basis.

Higher input costs. The materials that go into an MLCC cost more than they did a year ago. Taiyo Yuden’s September notice cited barium titanate powder, nickel powder, and rare earth additives. 36Kr reports that cuts to Indonesian nickel ore exports pushed up the price of the fine nickel powder used for internal electrodes, and that higher oil prices raised the cost of the PET release film used in tape casting. Murata has reportedly pointed to silver prices as well. Other capacitor types are seeing the same thing: Nichicon and Nippon Chemi-Con have raised aluminum electrolytic prices by a reported 10 to 15 percent, citing aluminum foil, chemicals, and electricity, and tantalum capacitors are in a similar upcycle.

How is this different from 2018?

The 2018 shortage was the worst in the industry’s history until now, so it is the obvious comparison. That one was driven mostly by channel hoarding, supply disruptions including earthquakes in Japan, and a round of price increases by Taiwanese makers. Underlying demand had not really changed, so the rally ended quickly and prices fell back through 2019.

A few things are different now. The demand driver is structural. An AI server uses ten to fifteen times the MLCCs of a conventional server, and each new platform pushes the count higher. The way parts are bought is changing too. TrendForce reports that customers are moving from quarterly price quotes to long-term agreements that lock in volume and price, and that Samsung Electro-Mechanics had more than ten AI MLCC long-term customers by late July, including global hyperscalers and major fabless chip designers. And the increases are already showing up in earnings, not just forecasts. Samsung Electro-Mechanics reported second-quarter operating profit up 106.7 percent year over year on revenue up 24.2 percent, with its component division up 29 percent on AI server and automotive MLCCs. Murata reported record fiscal first-quarter revenue of 502.3 billion yen, up 20.7 percent, and operating profit up 59.8 percent. It raised full-year guidance to 2.11 trillion yen in revenue and 430 billion yen in operating profit, and said it expects data center-related sales to more than double.

What could end the cycle?

There is a case on the other side, and TrendForce makes part of it. Its August note is cautious on everything outside AI. It calls the outlook for traditional computing and automotive conservative and warns that slower growth in China and the US could mean a weak second-half peak season for consumer electronics. The big increases on consumer-grade parts are meant to reduce demand, and eventually they will. Spot prices in Shenzhen have swung both ways, with quotes expiring daily. New capacity is on the way: Murata’s Izumo plant in 2027, the fiscal 2028 expansions at the Japanese majors, and new lines at Taiwanese and Chinese suppliers, several of which are going after the mid-range parts the leaders are leaving behind. Silicon capacitors, which we covered in the explainer, are gaining ground at the chip package level and could take some MLCC demand in the most advanced designs, though Murata argues MLCCs still win on capacitance and ease of use.

The biggest risk is the obvious one. All of the forecasts above assume hyperscalers keep spending on AI infrastructure at close to the current pace through the end of the decade. If they do not, this cycle ends like every component cycle before it, with new capacity arriving as demand cools. Research firm and bank forecasts are estimates, not guarantees, and this cycle has already surprised the people closest to it.

What to watch

• Q4 pricing from the Japanese majors. Whether Murata, Taiyo Yuden, and Kyocera follow Samsung Electro-Mechanics on direct OEM and ODM prices is TrendForce’s key signal for the next stage.
• The Q4 supply gap. TrendForce expects it to widen. Book-to-bill ratios, lead times, and channel inventory days will tell.
• The Rubin ramp. Volume production of VR200 racks in 2027 is the biggest scheduled step-up in MLCC content per system. Delays would push demand out; ODM inventory builds ahead of the ramp would pull it in.
• Long-term agreements and capacity announcements. More multi-year contracts would show customers paying for supply visibility. Bigger or faster capacity additions would show suppliers believe the demand is durable.
• Japan’s export statistics. The Ministry of Finance publishes monthly MLCC export values and unit prices, a public read on pricing that does not depend on supplier disclosures.
• Consumer devices. As capacity moves to AI grades, the cost and availability of ordinary capacitors in phones, PCs, and appliances is where the cycle shows up for everyone else.

The bottom line

When we wrote the explainer, this was a supply squeeze Wall Street had started to notice. A few weeks later, it is a declared upcycle, with direct price increases reaching the biggest customers and the leading industry researcher expecting the gap between supply and demand to widen. The reasons are simple enough. AI servers need more of a specific kind of capacitor that only a few companies can make, those companies cannot add capacity quickly, and their raw materials cost more. Whether this turns into the longest cycle in the industry’s history, as Goldman suggests, depends on how long the AI buildout runs and how fast capacity catches up. Either way, capacitors are no longer a rounding error in the electronics economy.

Frequently asked questions

Why are MLCC prices rising in 2026? AI servers use ten to fifteen times the multilayer ceramic capacitors of conventional servers, and demand has outrun the industry’s ability to add high-end capacity, which takes 18 to 24 months. Suppliers have shifted production toward AI-grade parts, which has tightened supply of ordinary consumer-grade parts too. Raw materials such as barium titanate, nickel powder, and rare earth additives have also become more expensive.

Which companies have raised MLCC prices in 2026? Murata (April), Taiyo Yuden (May and September), Walsin Technology (June and July), Yageo (July), and Samsung Electro-Mechanics (August, with a further increase for direct OEM and ODM customers in the fourth quarter) have all announced increases, according to TrendForce and other industry reporting. Taiwanese and Chinese suppliers are expected to follow in the fourth quarter.

How long is the MLCC shortage expected to last? TrendForce has said tightness in high-end MLCCs may last into 2027, with the supply-demand gap expected to widen in the fourth quarter of 2026. Goldman Sachs, after meeting with Murata’s management, has suggested the AI-driven demand cycle could run to around 2030. These are third-party estimates and are subject to change.

What does it mean that MLCCs are “the next memory”? The phrase comes from Goldman Sachs research comparing MLCCs to memory chips, where AI demand for high-bandwidth memory absorbed leading-edge capacity and caused shortages and price increases across conventional DRAM. Goldman argues the same pattern is playing out in capacitors as AI-grade production crowds out general-purpose parts. For background on what an MLCC is, see our explainer.

Sources

• TrendForce, “SEMCO Leads 4Q26 OEM Price Hikes, Signaling Start of MLCC Industry Upcycle,” August 27, 2026
• TrendForce, “NVIDIA’s AI Rack Transition Accelerates; NVL72 Output Value to Exceed US$710 Billion in 2027,” August 28, 2026
• TrendForce, “AI Demand Pushes Japanese and Korean MLCC Suppliers to Record Monthly Shipments,” July 28, 2026
• TrendForce, “CSP In-House ASIC Boom Drives MLCC Specification Concentration,” June 17, 2026
• TrendForce News, “MLCC Lead Times Diverge as High-End Products Stretch to 5-10 Months,” August 18, 2026
• TrendForce News, “Across-the-Board 30% Price Hike: MLCC Price Surge Intensifies,” August 4, 2026
• TrendForce News, “Samsung Electro-Mechanics Lifts MLCC Prices 30% Starting Aug. 1; Taiyo Yuden Reportedly Eyes Sept. 1 Hike,” July 29, 2026
• TrendForce News, “MLCC Turns From Electronic Rice Into a Gold-Like Rally,” citing 36Kr, June 25, 2026
• The Korea Herald, “AI Demand Drives Broader Samsung Electro-Mechanics MLCC Price Hikes,” August 2026
• DigiTimes, “Samsung Electro-Mechanics Eyes 30% MLCC Price Hikes, ABF Up Too,” September 2, 2026
• The Elec, “Samsung Electro-Mechanics to Raise MLCC Prices by 30% Starting in August,” July 30, 2026
• CNBC, “The Capacitor Supply Chain Bottleneck Is Becoming a Bigger AI Trade in the Market,” September 4, 2026
• South China Morning Post, “How Massive AI-Driven Demand for MLCCs Triggered a Supply Bottleneck,” June 15, 2026
• Yahoo Finance, “Wall Street’s Next AI Trade Is Smaller Than a Grain of Rice,” July 9, 2026
• Futu News, “Goldman Sachs Has Identified ‘the Next Memory’: MLCC,” June 30, 2026
• BigGo Finance, “Goldman Sachs Flags AI’s Next Bottleneck: MLCCs Poised to Replicate the Memory Chip Price Surge,” June 2, 2026
• BigGo Finance, “MLCC Sector Surges 9% in a Single Day: Goldman Sachs Says It Could Be the Next Memory Chip,” June 2, 2026
• BigGo Finance, “Goldman Sachs: AI-Driven MLCC Supercycle Could Extend to 2030,” May 27, 2026
• BigGo Finance, “Wall Street Giants Unanimously Call It: AI Supply Chain’s Scarcest Puzzle Piece Emerges,” June 1, 2026
• BigGo Finance, “AI Boom Triggers MLCC Supply Crunch; Samsung Electro-Mechanics to Hike Prices 30% From August,” July 29, 2026
• BigGo Finance, “Murata Manufacturing FY2026 Q1 Earnings Call,” August 4, 2026
• 36Kr, “Epic Price Surge of MLCC Is on the Horizon,” 2026
• 36Kr, “MLCC Demand Skyrockets: Could It Be ‘The Next Memory Semiconductor’?” June 5, 2026
• Investing.com, “Morgan Stanley Raises MLCC Forecast on Japan Output Data,” August 2026
• European Passive Components Institute, “TrendForce: CSP AI ASICs and Looming High-End MLCC Shortage,” June 18, 2026
• Aetrix, “Taiyo Yuden MLCC Price Increase 2026: Supply Risks,” 2026
• Bloomberg, “Murata Explores Raising Prices of Key AI Server Component,” February 2026 (as cited in industry coverage)
• Goldman Sachs, Morgan Stanley, and Citi analyst estimates as reported in financial media, May through August 2026