The semiconductor industry is currently obsessed with the "More than Moore" era. As shrinking transistors on silicon reaches the physical limits of atomic engineering, the focus has shifted from shrinking the chip itself to how chips are packaged and connected. At the heart of this evolution lies the "glass-core substrate"—a revolutionary technology intended to replace the organic materials that have served as the backbone of chip packaging for decades. Promised by Intel in late 2023 with a massive $1 billion investment, the technology was hailed as a panacea for the data-intensive demands of artificial intelligence, high-performance computing (HPC), and next-generation graphics. Yet, nearly a year later, the industry remains in a state of purgatory. Despite billions in capital expenditure and a flurry of joint ventures, not a single commercial product featuring a glass-core substrate has hit the market. As timelines slip and initial targets evaporate, the semiconductor industry is forced to confront the harsh reality of moving a nascent material from the lab to the assembly line. The Core Problem: Why Shift to Glass? To understand the delay, one must first understand why the industry is so desperate to ditch organic substrates. Modern chips, particularly those used for AI, are becoming massive. These "chiplets" require high-speed communication, low power consumption, and extreme signal integrity. Traditional organic substrates—essentially resin-impregnated fiberglass—are struggling to keep up. They are prone to warping under the heat generated by powerful AI accelerators. As chips get larger and the connections (interconnects) between them become finer, organic materials fail to provide the necessary flatness and mechanical stability. Glass, by contrast, offers superior thermal stability, lower electrical loss, and the ability to handle significantly higher density of interconnects. It is, in theory, the perfect platform for the next decade of computing. However, theory is not manufacturing. A Chronology of Slipping Timelines The race to mass-produce glass substrates began with immense optimism, but it has been marred by a recurring theme: optimistic roadmaps followed by quiet delays. The Initial Catalyst: Intel’s September 2023 Announcement Intel grabbed headlines when it showcased its glass substrate technology, framing it as the key to scaling the next generation of chip architectures. The company announced a $1 billion investment to build a manufacturing ecosystem in the United States, aiming to lead the industry away from traditional organic packaging. The Absolics Setback SKC, the chemical affiliate of the South Korean conglomerate SK Group, established its subsidiary, Absolics, to build a dedicated plant in Covington, Georgia. The original roadmap was ambitious: mass production by the first half of 2024. As of late July 2024, that deadline has come and gone. According to the company’s July 27 earnings call, samples from the Georgia plant are only now undergoing "package-level reliability evaluation" in Taiwan. The Samsung-Sumitomo Pivot Recognizing the complexity of the task, other industry titans have moved to consolidate resources. On July 2, Samsung Electro-Mechanics announced a 482.1 billion won ($310 million) joint venture with Sumitomo Chemical’s Dongwoo Fine-Chem. Unlike the initial sprint promised by early entrants, this partnership has adopted a more cautious, long-term horizon, targeting the first production in the second half of 2027. Intel’s Retrenchment Perhaps most telling is the shift in Intel’s own strategy. Having kicked off the race, the company has pivoted toward licensing its intellectual property and showcasing "demo vehicles" rather than immediate product integration. Intel’s internal deployment targets have been pushed back, with current projections pointing toward 2030—a lifetime away in the fast-moving world of semiconductor manufacturing. Supporting Data: The Manufacturing Gap The primary reason for these delays is the lack of a mature "ecosystem." Replacing organic substrates with glass is not a simple material swap; it requires an entirely new supply chain. Material Handling: Glass is brittle. Scaling up to large-panel manufacturing while maintaining high yield rates requires robots and handling processes that do not yet exist at the scale of organic production. Via Formation: Creating the microscopic "vias" (the vertical connections that pass through the substrate) in glass requires lasers and etching processes that are significantly more expensive and slower than those used for organic resins. Reliability Testing: As noted in the recent SKC earnings report, the bottleneck is currently at the reliability stage. Glass must survive the "thermal cycling" of a chip turning on and off millions of times without cracking or losing connection to the silicon die. These technical hurdles explain why reports of AMD adopting glass substrates for CPUs in 2025-2026 have failed to materialize. The transition requires the entire packaging ecosystem—the OSATs (Outsourced Semiconductor Assembly and Test companies), the equipment manufacturers, and the chemical suppliers—to synchronize their efforts. Official Responses and Industry Sentiment While public statements remain optimistic, the language has shifted from "imminent revolution" to "incremental development." Industry analysts suggest that the delay is not a sign of the technology failing, but rather an underestimation of the "brittleness" of the supply chain. During the recent quarterly earnings cycles, major players have avoided discussing specific launch dates for glass-based chips. Instead, the focus has shifted to "qualification milestones." The consensus among hardware engineers is that the industry is in a "proof of concept" phase. The technology works in a vacuum, but it has yet to prove it can be produced at the millions-of-units-per-month scale required for companies like Intel, NVIDIA, or AMD. The "qualification" process in Taiwan, mentioned by SKC, is the industry’s way of saying: We have the recipe, but we are still learning how to cook this in a commercial kitchen without burning it. Implications: The Long-Term Outlook The delay of glass substrates has profound implications for the roadmap of computing. 1. The Survival of Organic Substrates Because glass is lagging, the industry is forced to push organic packaging to its absolute limits. We are seeing a surge in "Advanced Packaging" techniques like CoWoS (Chip-on-Wafer-on-Substrate) and EMIB (Embedded Multi-die Interconnect Bridge) that use organic substrates but compensate for their weaknesses through clever design. Organic substrates will likely stay relevant for several years longer than predicted. 2. A Shift in Market Power The delay benefits those who have already mastered high-end organic packaging. TSMC, for instance, has a massive lead in advanced packaging. By keeping the industry on organic substrates for longer, they maintain their competitive edge against rivals who were hoping to use the transition to glass as a "reset button" for the semiconductor hierarchy. 3. The 2030 Horizon If the current trend holds, we are unlikely to see widespread commercial use of glass-core substrates until the end of the decade. The shift to 2030 for Intel suggests that the industry is preparing for a "generational" transition rather than an incremental upgrade. Conclusion The "Glass Age" of semiconductors is coming, but it is arriving on a timeline dictated by physics and manufacturing reality rather than marketing roadmaps. While the initial promises of 2024 mass production were clearly overly optimistic, the investment continues. The industry is currently in the "valley of death" for new hardware technologies: the phase where the R&D is mostly complete, but the manufacturing scale-up is proving to be a logistical nightmare. For the consumer and the enterprise, this means that the next few years of hardware performance gains will rely on clever architectural tweaks and existing packaging technologies. The true jump—the one enabled by the atomic precision and thermal resilience of glass—is still waiting in the wings. When it finally arrives, it will likely be quieter and less flashy than the initial hype suggested, arriving not as a sudden revolution, but as a hard-won, incremental victory for materials science. Post navigation The Wafer-Scale Revolution: How Cerebras Systems is Redefining AI Supercomputing
