Sakai Chemical AI Server Materials Turn Makeup Powder Expertise Into a Supply Chain Advantage

Sakai Chemical is considering a double-digit capacity increase after some annual orders for its AI server materials reportedly doubled. The unexpected bottleneck is not a processor, memory chip, or networking component. It is an ultrafine ceramic powder whose manufacturing lineage stretches back more than a century.

The Osaka-based company originally developed powder-processing expertise for safer white makeup materials. Traditional face paint worn by geisha and kabuki actors had relied on lead compounds that could cause chronic poisoning. Sakai Chemical supplied high-purity zinc oxide as an alternative, then applied its knowledge to sunscreen, plastics, tires, and electronic materials.

That progression now places the company upstream from Murata Manufacturing and Taiyo Yuden, two major producers of multilayer ceramic capacitors. These tiny components, commonly called MLCCs, regulate electrical current and suppress interference inside electronic circuits. AI servers require many high-capacitance components that operate reliably under demanding power and temperature conditions.

The Sakai Chemical AI server materials story is therefore an industrial reversal. A capability developed around cosmetics has become relevant to the densest computing infrastructure. Yet the real contest is not cosmetics versus computing. It is precision materials production versus the growing pressure to expand capacity quickly.

Sakai Chemical AI Server Materials Face a Capacity Test

Sakai Chemical has moved from benefiting quietly behind the AI boom to considering a meaningful expansion of its powder production.

President Toshiyuki Yagura said the company was calculating how much additional capacity it needed, according to a reported capacity plan. A double-digit percentage increase was likely, he said, with an investment decision expected by March.

The urgency comes from repeated revisions by customers. Yagura said advanced MLCC manufacturers, including Murata and Taiyo Yuden, had increased their order forecasts at successive meetings. Some Sakai Chemical products received twice their previous annual order volume after sales began accelerating during the second half of last year.

That does not mean every extra order corresponds directly to one AI accelerator. Barium titanate serves MLCCs across smartphones, vehicles, industrial equipment, networks, and servers. Demand can move between these markets, while customers also adjust inventories and procurement schedules.

However, Sakai Chemical’s recent disclosures provide evidence that AI infrastructure is changing the mix. During the latest reported quarter, electronic-materials sales increased 17.2% from the previous year. Segment profit rose 22.1%, reaching ¥581 million.

The wider company delivered a more complicated result. Consolidated revenue increased 6.6% to ¥21.47 billion, but operating profit fell 10.5% to ¥1.71 billion. Weak Japanese construction demand, changes in sunscreen preferences, and the wind-down of pigment-grade titanium dioxide weighed on other operations.

This divergence matters. Electronic materials have overtaken plastic additives as Sakai Chemical’s largest profit generator, according to the Bloomberg account. The company is becoming more dependent on a technically demanding growth business even while its broader portfolio faces uneven demand.

Its shares had gained roughly 15% during 2026 when the report appeared. That increase reflects investor interest, but it does not settle how much of the anticipated AI demand will become sustainable earnings. Capacity spending, customer negotiations, raw-material costs, and manufacturing yields will shape that conversion.

Sakai Chemical has not announced the final size, location, cost, or commissioning date of the proposed expansion. Until management approves the investment, the double-digit figure remains an informed direction rather than a completed production increase.

The immediate change is still concrete. Customers are asking for more material, some orders have doubled, and management is preparing a capacity decision. Those signals move Sakai Chemical from the edge of the AI supply-chain discussion toward a more visible position.

Why AI Servers Need More Ceramic Capacitors

AI accelerators concentrate extraordinary computing and power requirements onto limited board space, increasing the need for compact, high-capacitance components.

An MLCC consists of alternating ceramic dielectric layers and metal electrodes. The dielectric is an insulating material that allows a capacitor to store electrical energy. Stacking many thin layers raises capacitance without requiring one large component.

Servers place these capacitors around processors, accelerators, memory, networking equipment, and power-delivery circuits. They help smooth voltage, reduce electrical noise, and supply short bursts of energy when a chip’s demand changes quickly.

These functions become harder as accelerator power rises. AI chips switch enormous numbers of transistors while processing models, creating rapid changes in current demand. A board must maintain stable voltage despite those changes, heat, and restricted physical space.

Murata estimates that an AI server baseboard now uses between 15,000 and 25,000 capacitors. Its earlier estimate had been between 10,000 and 20,000. The company expects server capacitor demand to grow at an average annual rate of 30% between fiscal 2025 and fiscal 2030, according to its AI capacitor outlook.

The exact count varies with server architecture. A baseboard built around several accelerators will not match a general-purpose CPU server. Changes in voltage regulation, packaging, networking, and cooling also affect component selection.

The direction is nevertheless clear. Higher instantaneous power demand requires more local energy storage and tighter voltage control. Component makers must fit that capability into boards where every square millimeter competes with processors, memory, connectors, and cooling hardware.

Samsung Electro-Mechanics says AI servers can use 10 to 15 times as many MLCCs as conventional servers. Its AI component strategy emphasizes components that combine small dimensions with high capacitance.

The company also announced a one-year AI server MLCC supply contract covering 2027. It described the agreement as its largest MLCC supply contract and said it had long-term arrangements with more than 10 customers. That disclosure shows the demand surge is reaching established component manufacturers, not only upstream powder suppliers.

Samsung Electro-Mechanics cited a projection that the AI server MLCC market would grow from $1.4 billion in 2025 to $5.8 billion in 2030. Such market estimates should be treated as forecasts, not guaranteed sales. Still, disclosed supply agreements give manufacturers a stronger planning signal than broad projections alone.

Sakai Chemical occupies an earlier position in this chain. It does not sell a finished capacitor to a cloud operator. It supplies dielectric powder that capacitor manufacturers process into extremely thin ceramic layers.

That distance can make the company almost invisible to people tracking GPUs and cloud spending. It also creates leverage. A shortage of suitable powder can constrain output even when downstream manufacturers possess enough assembly equipment.

The potential bottleneck is measured in microscopic consistency. Each particle needs the required size, purity, shape, and crystal properties. A server customer ultimately cares about capacitor reliability, but that reliability begins well before the component reaches a circuit board.

The Advantage Lies in Uniform Barium Titanate

Sakai Chemical’s central advantage is its ability to produce small, highly uniform barium titanate particles through hydrothermal synthesis.

Barium titanate is a ceramic compound with strong dielectric properties, making it a core material for high-capacitance MLCCs. Producing it is not simply a matter of mixing barium and titanium compounds.

Conventional solid-state production combines raw materials, heats them, and mills the resulting mass into powder. That route can be effective, and major capacitor manufacturers often produce some dielectric materials internally. However, grinding can produce particles with uneven sizes and shapes.

Sakai Chemical uses hydrothermal synthesis for its high-end material. The process causes compounds to react inside a liquid solution under controlled temperature and pressure. Manufacturers can adjust multiple variables to influence particle growth and crystal formation.

Uniformity matters because an MLCC contains many extremely thin dielectric layers. Large particles, aggregated clusters, impurities, or gaps can interfere with layer formation. They can also create paths for electrode material to penetrate areas where it does not belong.

Sakai Chemical says its method supports thinner layers, greater layer counts, and structures with fewer gaps. Those characteristics can increase capacitance while preserving a small component footprint. They can also support reliability when a capacitor operates under sustained voltage and heat.

The company’s integrated materials report identifies powder processing as a core technology spanning foundation materials, sunscreen ingredients, and electronic dielectrics. That common foundation explains the striking transition from face makeup to servers.

Both applications require control over very small particles. Cosmetic powders must spread predictably, interact safely with skin, and deliver specific optical or textural effects. Electronic powders must form dense, repeatable layers with tightly controlled electrical behavior.

The products are not interchangeable, and the performance requirements differ greatly. The connection lies in accumulated manufacturing knowledge. Particle formation, purification, classification, dispersion, quality measurement, and scale-up all require experience that is difficult to acquire from a laboratory recipe alone.

Yagura argued that MLCC manufacturers depend on Sakai Chemical for materials used in advanced AI server capacitors, even though some produce other materials internally. That is a company claim, and Sakai Chemical has not disclosed customer-level shipment volumes or independently audited market shares.

Still, customer behavior offers partial support. Repeatedly increased forecasts and doubled orders suggest that buyers see limits in immediately replacing these materials. A producer would not normally push an upstream supplier toward expansion if equivalent capacity were readily available elsewhere.

Consistency is especially important during mass production. A supplier can produce one excellent laboratory sample without delivering the same result across thousands of batches. Customers qualify materials around repeatable performance, then design manufacturing controls around those characteristics.

Changing an upstream powder can therefore involve more than signing a new purchase order. A capacitor maker may need to adjust slurry preparation, layer formation, firing profiles, electrode interactions, and inspection standards. It must then demonstrate that the finished component still meets customer requirements.

This creates switching friction but not permanent protection. Competitors can improve their processes, customers can expand internal production, and component designs can change. Sakai Chemical must keep refining particle size, purity, and production efficiency while protecting manufacturing knowledge.

Its advantage also depends on raw-material access. Barium feedstocks are essential to dielectric production. Sakai Chemical has said it is diversifying sourcing and developing an international procurement network, including a subsidiary in India.

A reliable powder process cannot compensate for inconsistent inputs or interrupted supply. The company’s value therefore comes from an entire system: raw-material procurement, hydrothermal synthesis, powder processing, analytical control, customer qualification, and dependable delivery.

The cosmetics history demonstrates that this system has evolved across applications. It does not prove that Sakai Chemical owns an unassailable AI monopoly. It explains why a century of work on particles can matter when server designers need more electrical performance from less board space.

Expansion Puts Precision Against Speed

The main risk is that Sakai Chemical must add capacity without weakening the consistency that made customers depend on it.

Manufacturing expansion creates several pressures at once. New reactors and processing equipment must reproduce existing particle characteristics. Operators need training, inspection systems need calibration, and suppliers must provide adequate raw materials.

A double-digit capacity increase sounds modest compared with semiconductor megaprojects. Yet powder quality can change when equipment geometry, fluid movement, temperature distribution, pressure, or batch timing changes. Scaling a chemical process is not equivalent to copying a software installation.

The company also faces timing risk. Building too little capacity could leave customers short and encourage them to qualify alternatives. Building too much could burden returns if AI infrastructure spending slows, server designs change, or customers reduce inventories.

Yagura expects the boom to continue through at least 2027, although he said growth had begun to moderate. That distinction is important. Continued high demand does not require every quarter to repeat the steep increases seen during the initial acceleration.

Competitors are already responding downstream. Samsung Electro-Mechanics disclosed plans to strengthen its AI data-center MLCC and package-substrate operations in Busan. Its investment strategy covers development and manufacturing through 2040.

Murata is advancing smaller, higher-capacitance components while improving thermal performance. Taiyo Yuden also serves high-end capacitor markets. These companies compete for finished-component orders, but they can simultaneously pressure upstream suppliers for more output, better specifications, and stable terms.

Vertical integration presents another risk. MLCC manufacturers already make some dielectric materials themselves. If powder scarcity persists, they have a stronger incentive to invest in internal processes or support an alternative supplier.

Sakai Chemical’s restraint on pricing adds a commercial tension. Yagura said the company did not plan increases beyond passing through higher input costs. That approach can protect customer relationships during a shortage, but it can also limit returns from scarce capacity.

Management appears to treat reliability and long-term trust as more valuable than maximizing immediate margins. That choice is defensible in a qualified industrial supply chain. Customers may remember which vendors maintained predictable terms when demand exceeded supply.

Investors can still question whether conservative pricing compensates for expansion risk. New equipment consumes capital before it generates qualified volume. If customers capture most of the scarcity value, Sakai Chemical could bear the manufacturing challenge without receiving a proportional increase in profit.

The reported 22.1% increase in electronic-materials profit provides a favorable signal, but one quarter cannot establish a long-term margin trend. Readers should separate segment momentum from consolidated performance, which included weaker results elsewhere.

Chinese materials manufacturers add competitive pressure. Yagura acknowledged their increasing presence in semiconductors, electronic materials, and MLCCs. He argued that Japanese manufacturers retain an advantage in delivering consistent quality, rather than only producing a superior laboratory sample.

That judgment comes from an interested executive. Chinese suppliers have the capital, domestic demand, and technical incentives to improve. Even if they cannot immediately replace qualified Japanese materials in the most demanding applications, progress in mid-range products can free resources for higher-end development.

Geopolitics complicates sourcing decisions further. Server and component companies increasingly consider the country of origin, trade restrictions, logistics, and concentration risk alongside technical performance. A customer may accept higher qualification costs to reduce dependence on one supplier or region.

Sakai Chemical must also avoid letting the AI narrative obscure its wider business. The company still sells materials for cosmetics, vehicles, plastics, construction, and other industries. Weakness in those markets can offset gains from electronic materials, as the latest consolidated figures demonstrated.

The core pressure is therefore operational, not promotional. Sakai Chemical must decide how much capacity to add, qualify it with demanding customers, preserve batch consistency, and generate an acceptable return. A shortage makes that task urgent, but urgency makes mistakes more expensive.

Three Signals Will Show Whether the Advantage Lasts

The next capacity decision, customer demand, and production performance will determine whether Sakai Chemical becomes durable AI infrastructure or a temporary shortage beneficiary.

The first signal is the investment decision expected by March. Management should disclose the planned percentage increase, production site, capital commitment, and expected operating date. Those details would turn a broad expansion possibility into an executable manufacturing plan.

A larger-than-expected commitment would indicate that customer forecasts remain strong enough to justify long-lived equipment. A delay or smaller expansion would suggest either softer demand, difficulty scaling the process, or caution about returns.

The second signal is the trajectory of electronic-materials sales and profit. Continued double-digit growth would support the argument that server demand is converting into commercial results. Investors should also watch margins, because sales can rise without creating comparable value after raw-material, labor, and depreciation costs.

Customer announcements will offer another view. Murata has already raised its estimate of capacitor content per AI server baseboard. Samsung Electro-Mechanics has disclosed long-term agreements and a large 2027 supply contract.

Additional customer commitments would strengthen Sakai Chemical’s demand case, especially if they include compact, high-capacitance MLCCs. Inventory reductions, postponed data-center projects, or lower component forecasts would weaken it.

The third signal is manufacturing quality after expansion. Sakai Chemical should maintain yields, delivery performance, and customer qualifications while increasing output. Public disclosures rarely reveal every relevant quality measure, so recalls, delays, write-downs, or unusually high startup costs become useful warning signs.

A successful expansion would demonstrate that the company can transfer its process controls to more equipment without losing uniformity. That matters more than a one-time increase in orders. Repeatable scale is what turns specialized knowledge into lasting supply-chain leverage.

The server architecture itself deserves attention. Greater accelerator density and higher power demand currently support more capacitors per system. Advances in voltage delivery, packaging, silicon capacitors, or other component technologies could alter the number and mix of MLCCs.

Such changes rarely eliminate mature components overnight. They can shift which capacitance, voltage, temperature, and size classes grow fastest. Sakai Chemical must keep its powder development aligned with those requirements rather than treating all MLCC demand as interchangeable.

The broader lesson extends beyond capacitors. AI infrastructure depends on layers of materials, power components, thermal systems, optical connections, and manufacturing equipment. Many of the most consequential suppliers sit several steps away from a recognizable AI product.

That distance makes them easy to overlook. It also means demand can arrive through several customers simultaneously, creating shortages before cloud providers or developers notice the underlying constraint.

For developers and enterprise AI buyers, the immediate effect is indirect. They will not select barium titanate powder when deploying a model. However, material availability influences how quickly server vendors can deliver systems and how reliably those systems operate.

Teams evaluating AI infrastructure should therefore track component and material capacity alongside GPU roadmaps. A searchable knowledge base can help engineering groups connect supplier disclosures, qualification records, architecture changes, and delivery risks.

Sakai Chemical’s history makes the story memorable, but history alone does not make it an AI linchpin. The stronger case rests on present evidence: doubled orders for some products, rising electronic-materials profit, repeated customer forecast increases, and a pending capacity decision.

The next few months should answer a practical question. Can Sakai Chemical expand its AI server materials while preserving the particle consistency that customers value?

If the company approves new capacity, maintains quality, and converts demand into sustained segment earnings, its century-old powder expertise will have found a durable role in AI infrastructure. If those signals weaken, the current attention will look more like a shortage cycle than a structural transformation.