Fiberglass in the sights of AI: scarcity, prices, and hidden power

The race for the Artificial Intelligence This is causing brutal pressure on the technology supply chain, and one of the most sensitive points has a name and surname: high-precision fiberglass used in chips and circuitsWhat was just a few years ago a relatively unremarkable material has become a strategic resource capable of influencing launches, multi-million dollar investments, and even the industrial policy of several countries.

Behind this special fiber, very different from that found in insulation or surfboards, there are a handful of Japanese and Chinese companies, but above all, one undisputed protagonist: Nitto Boseki (Nittobo), the almost exclusive supplier of so-called T-glass or glass clothThe demand unleashed by AI chips from NVIDIA, GoogleAmazon and other giants are stretching their capacity to the limit, opening the door to price increases, delays, and a new global bottleneck similar to the one experienced with RAM.

From two Japanese weavers to an invisible pillar of the chip era

A little over a century ago, two Japanese textile companies, Fukushima Boseki Co., Ltd. and Katakura Seishi Iwashiro Bosekisho, decided to join forces to create what we know today as Nitto Boseki Co. Ltd., or simply Nittobo, the discreet giant that weaves the "fabric" on which modern chips relyWhat began as a textile business gradually shifted towards advanced materials, especially those related to glass.

Over the decades, the Japanese firm moved away from traditional textiles and into the manufacture of technical fibers. In 1938, almost simultaneously with Owens Corning Fiber Glass in the United States, Nittobo became one of the first companies in the world to produce fiberglass on an industrial scaleopening up a huge field of industrial and electronic applications that, with ThereThey would become critical.

The big technological leap came in 1969, when the company first developed an extremely fine and uniform glass fabric, known as glass cloth. This material began to be used as base of the printed circuit boards (PCBs) that populate any electronic deviceThe key was to achieve a stable, heat-resistant fabric with very specific physical properties to withstand increasingly demanding manufacturing processes.

Back in 1984, Nittobo launched its famous T-glass, an evolution of that fiberglass fabric that took performance a step further. It wasn't your typical construction or leisure fiberglass, but a ultrathin fabric, with a very low coefficient of thermal expansion, designed to serve as a substrate for advanced chipsThis property is essential: when a chip heats up and cools down rapidly, any difference in expansion between layers can generate stresses, failures, and even breakages.

Thanks to its dimensional stability and extremely high quality, Nittobo's T-glass gradually made its way into the products of major technology companies. Apple was one of the first to bet on this glass as a base material for key components of its iPhoneand over time other giants in the mobile and semiconductor sectors have followed suit.

What makes T-glass fiberglass so special

Although at first glance it may seem like a simple rigid plastic, the reality is that T-glass or glass cloth used in the manufacture of chip substrates is a material of tremendous sophistication. Each filament of this fiber must be thinner than a human hair, perfectly round, and free of bubbles.Any microscopic imperfection can result in a serious failure in the final chip.

The main function of this glass fabric is to serve as a "skeleton" for the substrates where chips are mounted and for many circuit boards. Therefore, in addition to its strength and purity, it stands out for an extremely low coefficient of thermal expansion and a rigidity that ensures stability even at very high temperaturesThis prevents the structure from deforming when the chip is operating at maximum performance.

Another key feature is its ability to to facilitate very high-speed data transmissionAn unstable or micro-defective substrate can cause interference, electrical noise, or signal delays, which is unacceptable in high-performance processors, data center chips, or GPUs designed for AI, where every nanosecond counts.

This type of high-quality fiberglass is not an interchangeable commodity. Leading chip manufacturers, from those producing mobile processors to those developing AI solutions, rely almost entirely on it. a type of T-glass that only a few suppliers can manufacture to the necessary standardsThat is why the concentration of supply is so worrying for the entire sector.

In this context, the Japanese company Nittobo has become the dominant player in the production of the most advanced fiberglass used in chip substratesOther firms, such as China's Grace Fabric Technology or Japan's Unitika, are trying to compete, but today they are still far behind in both scale and consistency of quality, which makes large chip manufacturers very reluctant to change suppliers.

The AI ​​explosion: everyone wants the same fiberglass

The rise of artificial intelligence has completely changed the priorities of the technology sector. From training massive models to real-time inference, AI needs colossal amounts of computing power and memoryThis translates into an unprecedented demand for GPUs, specialized CPUs, accelerators, and, by extension, all the materials involved in their manufacture.

Companies like NVIDIA, Google, and Amazon Web Services have seen a surge in orders for AI-specific chips. Each new generation of GPUs for data centers comes with... large contracts for the supply of substrates and advanced componentsAmong these, T-glass is absolutely central. The result is fierce competition to secure as much of Nittobo's production capacity as possible.

Before the AI ​​boom, mobile market giants like Apple and Qualcomm rarely had trouble sourcing this fiberglass. Their consumption was high, but Nittobo's production could easily meet demand. However, The arrival of superchips for AI radically changed the balance, leaving many actors competing for a limited resource.

RAM and NAND flash memory have been the first serious warning: the strong demand for training AI models has led to sharp price increases and lack of availability for other segmentsWith T-glass, something very similar is happening: the bulk of the capacity is being diverted towards the highest-paying customers who require the most volume, that is, the large firms focused on AI.

The consequences are felt throughout the entire supply chain. Consumer electronics manufacturers, from smartphones to portable and other devices, fear being displaced in the bidding for high-quality fiberglass. When key materials are sold to the highest bidder, the margins and bargaining power of companies with more competitively priced products suffer., and their sales forecasts may be affected.

A global bottleneck with a 2026 date

Semiconductor industry experts warn that The shortage of high-quality fiberglass could become one of the biggest bottlenecks in the technology industry by 2026.The situation is reminiscent of the post-pandemic chip crisis, but in this case the focus is on a specific material with very few suppliers.

Nittobo acknowledges that, at least in the short term, the situation is practically at a standstill. According to the Asian business press, a company executive even stated that “If we don’t have additional capacity, we don’t have it, no matter how much pressure is put on Nittobo.”It's a polite way of saying that, quite simply, factories are operating at their limit and you can't get more output out of thin air.

The Japanese company has plans to increase its capacity, but that increase won't happen immediately. Expansion projects and new facilities require substantial investments, permits, and, above all, time to set up extremely complex production linesEverything suggests that the significant increase in supply will only become clear in the second half of 2027.

This timeline coincides with other market tensions. Firms like SK Hynix have warned that The RAM shortage could last until at least 2028The convergence of two bottlenecks—high-precision memory and fiberglass—can lead to a prolonged scenario of high prices, launch delays, and extreme customer prioritization.

The shortage of this material doesn't just affect consumer hardware. Sectors such as telecommunications, banking, and energy are increasingly dependent on it. data center infrastructures and cloud services that also require advanced chips and large amounts of fiberSeveral analysts point out that, if the situation persists, hardware and cloud service costs could rise significantly from 2026 onwards, impacting the digital economy as a whole.

Apple, Qualcomm and glass diplomacy

One of the most striking consequences of this silent crisis is the direct involvement of governments and top executives to ensure supply. Apple, which has used Nittobo's T-glass components in several generations of iPhones, has sent executives to Japan to negotiate face-to-face. with the supplier and with representatives of the Japanese government.

Apple's concerns aren't limited to current models. The company has particularly sensitive products on the horizon, such as its highly anticipated foldable iPhone and future devices with even more intensive use of AI and advanced connectivity. Any restrictions on access to high-quality fiberglass could disrupt their launch schedules., something that in a company of this size means billions at stake.

Qualcomm, for its part, has also found itself caught in this bottleneck. Known for its Snapdragon processors for mobile phones and other connected devices, the American company has tried to diversify its supply. It has sounded out Unitika, a smaller Japanese fiberglass manufacturer., in the search for an alternative to Nittobo, but so far has not found a solution capable of absorbing the volume it needs.

The problem for both companies—and for many other chip manufacturers—is that The quality required of T-glass for advanced substrates is so high that a manufacturing defect can ruin entire batches of components.It's not simply a matter of "trying" another provider: the performance risks, reliability and reputation are enormous.

Analysts like Chiu Shih-fang, from the Taiwan Institute of Economic Research, emphasize that The stability of T-glass is crucial for the final quality of the substratesIf even a customer the size of Apple encounters limitations, smaller manufacturers risk being relegated to the sidelines, with much more restricted access or prohibitive prices.

Nittobo, maximum caution to avoid repeating history

The obvious question is why Nittobo doesn't simply increase its capacity to take advantage of this strong demand. The answer has to do with the recent memory of other markets, such as DRAM and NAND memory, which suffered a cycle of overproduction, price collapse and heavy losses for manufacturers in 2022.

Following those events, many Japanese companies in the materials and electronics sectors have become particularly conservative. Nittobo doesn't want to fall into the temptation of over-engineering its factories and then being left with full warehouses if demand normalizes or if competitors emerge with alternative technologies. The message is clear: they will grow, but slowly and cautiously..

Nittobo's CEO, Hiroyuki Tada, has publicly admitted that "it is inevitable that we will lose some market share" and that there is a limit to the level of risk that a relatively small company can acceptIn other words, the company prefers to give up part of the pie rather than embark on an uncontrolled expansion that could backfire.

This stance, understandable from a business perspective, nevertheless fuels market tension. Large buyers of fiber for AI, with extremely high margins (NVIDIA, for example), can afford to pay more and sign long-term contracts, securing the bulk of current capacity. Consumer electronics manufacturers, with their tighter margins, are left in a much more precarious position. and they must look for alternatives, even if they are still immature.

Meanwhile, China and Taiwan are emerging as potential sources of new suppliers. Apple and Qualcomm, among others, are already working on this. diversification strategies towards Chinese and Taiwanese companies capable of producing advanced fiberglassBut the necessary leap in quality is significant, and the industry knows that a mistake at this point cannot be fixed with a simple software patch.

Corning: the other major player in glass in the AI ​​era

Although Nittobo is getting media attention for the T-glass bottleneck, it's not the only glass giant benefiting from the rise of AI. The American company Corning, with almost 175 years of history and headquartered in a small town in upstate New York, has established itself as a key player on multiple fronts related to glass, ceramics and optical communications.

Corning has been achieving technological milestones for over a century. It manufactured Edison's glass envelopes in 1879, launched the durable Pyrex glass in 1915, developed a large mirror for the Hale telescope in 1948, and by 1962, it was producing the first hardened safety windshield for automotive applicationsdesigned to break into small fragments and reduce damage in accidents.

In 2007, a name crossed their path that would change everything: Steve Jobs. The Apple co-founder was looking for a particularly resistant glass for the screen of the first iPhone, fearing that breakage would ruin the user experience. Corning responded brilliantly, and from that collaboration, the iPhone was born. Gorilla Glass, the family of hardened glass that today equips millions of mobile phones, smartwatches and tablets worldwide.

But Corning's contribution to the digital age goes far beyond screens. In 1970, the company developed a low-loss optical fiber capable of transmitting data over long distanceslaying the foundation for modern communications networks. Their cabling solutions, such as Edge and Edge8, are widely used in data centers, the true physical heart of the cloud and AI.

The arrival of artificial intelligence has further increased the importance of this business. A state-of-the-art data center dedicated to AI can require up to 180.000 kilometers of fiber optic cable insidea distance equivalent to circling the globe more than four times. In a high-density, high-speed market, Corning's technological expertise has allowed it to position itself as a preferred supplier for many large technology companies.

Corning, AI and the future of data centers

In the race for AI, Corning isn't just selling cabling and fiber. The company is involved in the development of new optical interconnection solutions for next-generation data center architectureswhere photonic integration is becoming increasingly relevant to reduce consumption and latency.

One example is its participation in Broadcom's CPO (Co-Packaged Optics) program. In addition to classic optical connectors, Corning has developed New FAU (Fiber Array Units) products with higher density and tighter bend radiiOptimized for high-capacity multiplexing systems and data center transceivers, these components, while appearing passive and discreet at first glance, conceal a level of precision and industrial processing that reflects their long tradition in advanced glassmaking.

Among the technical advantages of these solutions are increases of around 35% in fiber density for the same width, reductions of 50% in the height of certain modules thanks to tighter bend radii, and excellent core alignment tolerance for designs with very high fiber countsAll of this contributes to building more compact, efficient data centers capable of handling the massive flows of information typical of AI.

The market context supports this strategy. Microsoft, Google, and other cloud giants have announced investment plans of tens of billions of dollars in data centers, many of them motivated directly by the needs of the Generative AI and associated services. Microsoft, for example, has tied part of its investment to its alliance with OpenAIwhile Google has stated capital spending expectations heavily focused on data infrastructure.

Corning, as a supplier of essential materials and components, benefits from this long-term investment cycle. Unlike low-cost competitors that compete on price and volume, the company has preferred to focus on technological differentiation and to become a key partner of major equipment manufacturers and data center operatorsthus securing its position in the AI ​​value chain.

In addition to optical communications and Gorilla Glass, Corning maintains other strategic lines of business in environmental technologies, specialty materials, and life sciences. It supplies Components for solar energy, space and military applications, including ceramic radomes for missiles, designed not to alter radar signals and to withstand extreme thermal changes, further reinforcing its profile as a critical supplier in very diverse fields.

Economic impact: rising prices and risk of a “technological freeze”

The combination of explosive demand for AI and a limited supply of key materials and components is generating a “technological corral” on a global scaleLarge companies compete to reserve capacity several years in advance, and suppliers with differentiated technology—such as Nittobo in T-glass or Corning in advanced optical fiber—gain considerable bargaining power.

For countries like Spain, this situation is not a distant issue. Telecommunications operators, banks, electricity companies, and large corporations depend on it. data centers, fiber optic networks and cloud services whose cost can skyrocket If hardware prices continue to rise due to material shortages, delays in advanced digitization projects or applied AI deployments could lead to a loss of competitiveness.

Some experts predict that, if the current tension continues, Hardware and cloud service prices will increase significantly from 2026 onwards.This would affect both large companies and SMEs that outsource their technological infrastructure, and ultimately the end user, who would see these costs reflected in more expensive products and services.

Faced with this scenario, supply chain actors are seeking to diversify: more suppliers, more producing regions, and, in some cases, new materials or architectures that reduce dependence on certain critical componentsHowever, the speed at which AI is advancing and the technical complexity of the materials involved make it very difficult to balance the scales in the short term.

Within this complex landscape, veteran companies like Corning have demonstrated remarkable resilience. With nearly 60.000 employees and five major business divisions, the company has managed significant growth in sales of optical communications and a stock market performance far superior to indices such as the S&P 500 or the EuroStoxx 50, supported precisely by the wave of investment in AI and data centers.

The story of Nittobo and Corning illustrates the extent to which The artificial intelligence revolution rests, to a large extent, on seemingly modest materials such as fiberglass and technical glass.The ability to produce them with the required purity, stability, and precision has become a decisive factor that can accelerate or slow down the adoption of AI worldwide, influencing prices, timelines, and the very strategy of large technology companies.