AI drives need for glass and crystalline materials

Bring artificial intelligence (AI) up in conversation and you may be met with a mix of excitement and curiosity. Its future is still unfolding, but AI is already making our lives easier from the facial recognition in your security doorbell to customized social media experiences, and smarter consumer electronic devices.

AI can help in ways yet to be discovered. The technology can help doctors diagnose and treat disease, help manufacturers quickly spot and correct defects, and even enable self-driving cars.

But it requires a crucial component: glass.

Corning invents, makes, and sells revolutionary materials that push AI technology forward—namely its internet-enabling optical fiber as well as specialized glass and crystalline materials that are central to nearly every step of the semiconductor chip manufacturing process.

An exponential desire for computing

With large language models that power systems like ChatGPT, Claude.ai, and Microsoft Copilot increasing in popularity—these AI text-generating websites saw 4.7 billion monthly visits in January 2025—the strain is on data centers to keep up with demand.

Data centers, a.k.a. the brains of the internet, essentially relay all our data and compute our every request. AI is now turbocharging data usage, and data centers are expanding to meet the demand.

For example, GPT-3 launched in 2020 with 175 billion parameters. GPT-4, introduced in 2023, is estimated to have more than a trillion parameters—an increase of 10x within only three years. The technology can’t continue to keep pace.

Corning has a solution for this conundrum—and it all comes down to the graphics processing unit (GPU), which are specialized processors essential for gaming, video editing, machine learning, and other applications that require high-quality graphics and effects.

Glass makes the world smarter

GPUs are the star of the show when it comes to accelerating AI. Regular computing relies on central processing units (CPUs)—a basic unit of electrical circuitry that performs computing instructions stored in its memory, like arithmetic and input/output operations. CPUs are the core components of computers and have been since the 1950s.

But AI requires more GPUs. The power of GPUs lies within their ability to process many pieces of data simultaneously with more powerful microchips—all with tinier features to pack more chips in smaller spaces.

It's a story Corning knows well. For the last 50 years, we’ve produced materials to help industry leaders mass produce the most powerful semiconductor chips, such as Corning’s ULE Glass (ultra-low expansion glass) for extreme ultraviolet (EUV) lithography.

During the early 2000s, the lithography development of 13.5-nm tools required a major design shift in scanner projection optics from refractive to reflective. Produced primarily by flame hydrolysis, the thermal stability of ULE glass—invented during the 1960s for demanding astronomy applications—made it an ideal material for EUV lithography in semiconductor chip manufacturing. Its stability is ideal for applications in which high-precision control is essential, such as photomask substrates, machine reference blocks, scanner optics, metrology, and other processes where any shift in material dimensions due to thermal changes can compromise performance and accuracy.

As EUV lithography becomes the industry standard for advanced node critical layers, we recently unveiled Corning EXTREME ULE Glass, a next-generation material to help chipmakers meet evolving specifications for next-generation products. More precise photomasks have a critical role to play in enabling mass production of today’s most advanced and cost-efficient microchips for GPUs. This process requires a glass with extreme thermal stability and uniformity to ensure consistent manufacturing performance. At the same time, it must also increase photomask flatness and CTE uniformity—and enable the deposition of advanced EUV reflective coatings.

For chip layers made by deep-UV (DUV) lithography, printing chip circuits at this scale demands an extremely reliable laser as well as durable supportive optics. Due to high fluency, excellent lifetime stability, and high damage threshold during the DUV lithography process, crystalline materials like calcium fluoride (CaF₂) have been widely considered the materials of choice for excimer laser-based microchip fabrication. Corning has a long history of growing CaF₂ ingots, and has invested in considerable R&D efforts to improve the durability of this material as the industry demands higher UV power, longer pulse length, and higher pulse repetition rates to print more chips.

Within the GPU, you’ll find high-bandwidth memory that stores massive data sets for the GPU’s computational tasks. Corning supplies glass carrier wafers to support silicon memory wafers during manufacturing. Once the memory wafer is thinned to the appropriate size, say 40 μm, the glass wafer is detached. Our glass carriers also help support the interposer wafer, which connects the GPU to the memory.

Glass is expected to become part of the actual substrates for GPUs, a.k.a. a glass core. We believe glass will help increase the mechanical and electrical reliability of GPUs as they get larger. The exceptional dielectric properties of glass, combined with the enablement of smaller line/spacing interconnects, will be crucial to increase GPU data bandwidth while reducing power consumption.

Optical fiber carries high-speed data

Clusters of GPUs must be networked together to provide computing scale and enable ever-growing large model AI training. Optical fiber, the revolutionary glass material invented by Corning more than 50 years ago, carries high-speed data from GPU to GPU. And Corning’s fiber is famous for its ability to fit into highly dense server environments, which enables more GPUs and more processing.

Corning experts estimate that data centers running AI large language models will require 5x more optical connectivity than today. This means AI data centers need more glass optical fiber to transfer data in the form of light signals.

Where installers once spent hours connecting fiber networks within the data center, our plug-and-play solutions shrink that time. Our EDGE distribution system consolidates dozens of cords into a single assembly to simplify the deployment of cabling. This not only saves time and space, but has the sustainability benefit of using less material.

Whether it’s enabling AI-focused data centers with fiber, helping pattern chips needed for advanced computing, or serving as a support for vital GPU components during manufacturing, glass, in its many forms, is core to our future with AI.

And not even AI can predict what glass will enable next.