Why the Semiconductor Industry is Replacing Plastic with Glass to Build the Next Generation of AI Chips

The artificial intelligence revolution is colliding with a physical wall, and it has nothing to do with the silicon itself. The bottleneck lies in the foundation—the small, plastic-like boards that connect microscopic chips to the rest of a computer. For decades, the semiconductor industry has relied on organic resin substrates to serve as this critical bridge. But as AI accelerators grow to unprecedented sizes and consume hundreds of amps of power, these traditional plastic foundations are literally buckling under the heat.[7]

To keep Moore’s Law alive and enable the next generation of computing, the industry is turning to a material that is older than the Roman Empire: glass. Glass core substrates represent one of the most significant material shifts in semiconductor packaging in decades. By replacing flexible organic materials with rigid, ultra-flat glass, engineers can pack more transistors into a single space, transmit data faster, and prevent the warping that plagues modern chip manufacturing.[2][3]

To understand why this shift is necessary, one must look at how modern chips are built. A semiconductor package is not just a piece of silicon; it requires a substrate to translate the nanometer-scale wiring of the chip into the millimeter-scale wiring of a motherboard. As companies design massive AI processors by stitching together multiple smaller chiplets, the substrate must grow larger to accommodate them.[4][7]

The problem is thermal dynamics. Organic substrates expand and contract at a different rate than the silicon chips attached to them. Under the extreme temperatures of a data center, this mismatch causes the package to warp—sometimes bending enough to break the microscopic connections. Glass, however, possesses a coefficient of thermal expansion that closely matches silicon. This physical harmony reduces package warpage by up to 50 percent, allowing manufacturers to build massive, multi-chip packages without fear of structural failure.[2]

Beyond structural integrity, glass offers a pristine canvas for microscopic engineering. Organic resins are inherently uneven at the microscopic level, which limits how finely machines can print circuit lines. Glass is exceptionally flat and smooth. This ultra-flatness allows lithography machines to draw redistribution layers with circuit lines thinner than 2 micrometers—a density that organic cores struggle to achieve.[1][2]

The electrical properties of glass are equally transformative. In high-frequency applications, signals traveling through plastic suffer from loss, bleeding energy into the surrounding material. Glass has a dielectric constant significantly lower than silicon, meaning electrical signals can travel faster, with orders of magnitude less transmission loss. This results in cleaner, faster data transfer, which is the lifeblood of AI training workloads.[2][3]

The true magic of the technology, however, lies in Through-Glass Vias (TGVs). Because glass is uniform, manufacturers can use highly focused lasers to blast microscopic elevator shafts directly through the substrate. These TGVs allow vertical electrical connections at a density up to ten times higher than what is possible with plastic, creating ultra-fast data highways between the top and bottom of the chip package.[1][7]

Looking further ahead, glass offers a capability that opaque plastic never could: transparency. Researchers are already exploring ways to embed optical waveguides directly into the glass substrate. In the future, this could allow chips to transmit data using pulses of light rather than electrical currents—a concept known as co-packaged optics—drastically reducing power consumption and heat generation.[2][3]

Recognizing these immense benefits, a global race has erupted to bring glass substrates to mass production. Intel was the early pioneer, investing heavily in a billion-dollar research line in Arizona and declaring glass the key to achieving one trillion transistors in a single package by 2030. The company successfully demonstrated early samples combining its advanced packaging with glass cores, proving the concept was viable.[1][4]

However, the economics of pioneering new manufacturing techniques are daunting. Recent industry reports indicate that under its new leadership, Intel is pivoting its strategy. Rather than building the glass substrates entirely in-house, the company is reportedly looking to outsource the manufacturing to specialized vendors. This leaner approach allows the company to reduce development costs and accelerate adoption by tapping into a maturing external supply chain.[5]

Meanwhile, South Korean tech giants are aggressively moving to fill the manufacturing void. Samsung has formed a cross-departmental alliance, combining the semiconductor expertise of Samsung Electronics, the substrate knowledge of Samsung Electro-Mechanics, and the glass processing mastery of Samsung Display. This unique, vertically integrated ecosystem aims to commercialize glass interposers and substrates as early as 2026 or 2027.[6]

Other specialized players are also ramping up. SKC’s subsidiary, Absolics, has established pilot lines and is diversifying its supply chain to prepare for mass production. Japanese firms, leveraging decades of experience in display glass and precision optics, are similarly adapting their legacy technologies to serve the semiconductor market. Analysts project the global market for glass in semiconductors could reach $4.4 billion by 2036.[1][3]

Despite the momentum, the transition is not without significant hurdles. The most obvious challenge is the material's inherent brittleness. Handling ultra-thin glass panels—often less than 100 micrometers thick—in high-speed robotic assembly lines requires entirely new equipment to prevent micro-cracks and catastrophic shattering.[7]

Yield rates also remain a closely guarded industry concern. Creating millions of flawless Through-Glass Vias without compromising the structural integrity of the panel is a metallurgical and optical balancing act. A single microscopic fracture can render an entire substrate useless, making the current cost of glass packaging substantially higher than traditional methods.[6]

Because of these costs, glass will not replace organic substrates in everyday consumer laptops or smartphones anytime soon. Instead, the two technologies will coexist. Glass will debut at the absolute high end of the market—powering the massive, power-hungry AI accelerators and data center servers where its thermal and electrical advantages justify the premium.[3][7]

As the artificial intelligence era demands exponentially more computing power, the industry can no longer rely solely on shrinking transistors. The future of computing requires a stronger, faster, and cooler foundation. By turning to glass, engineers are ensuring that the physical infrastructure of the digital age can keep pace with its software ambitions.[7]