Intel has targeted South Korea to support its next big venture in advanced packaging: glass substrates. Industry sources indicate that the company is exploring supply agreements with Samsung, which has positioned itself as one of the leading candidates to provide this new “base” upon which next-generation CPU and GPU will be assembled. The move aligns with a strategic need to diversify suppliers and accelerate entry into a technology that the entire ecosystem considers critical in the era of AI.
This idea isn’t coming out of nowhere. Intel has been in talks for months with glass and substrate manufacturers to evaluate capabilities and timelines. In the Korean landscape, Samsung has an advantage for two reasons: firstly, the conglomerate has heavily invested in this area; secondly, it benefits from internal synergies — Samsung Display has decades of experience working with large-format glass, engraving, and treatments that are now being repurposed for packaging.
Why Glass Is at the Heart of the Next Packaging Leap
The semiconductor industry has been seeking alternatives to organic substrates and traditional silicon interposers for years. Glass has emerged as the leading candidate for three technical reasons:
- Dimensional stability: Glass undergoes less deformation; its coefficient of thermal expansion is lower and more uniform than organic materials, alleviating stresses and deformations as interconnection density increases.
- Signal integrity: superior flatness, rigidity, and reduced dielectric loss help maintain high-speed signals with less noise, essential for chiplets requiring massive bandwidth.
- Process scalability: technologies like TGV (Through-Glass Via) open the door to high-density vias and new interconnection topologies.
In short: the entity that masters glass will be able to package more and better. At a time when GPU and AI accelerators are pushing the limits of HBM, interposers, and substrates, this advantage could be decisive.
Two Competing Families (and Why Intel Will Need Both)
Within the realm of “glass substrates”, today there are two products serving different purposes:
- Glass core substrates (GCS): replace the organic core of the substrate with glass. This is the “heart” of the package. It’s more difficult to manufacture and has slightly less maturity.
- Glass interposers: function as “bridges” in 2.5D and 3D configurations, connecting dies and the rest of the package with high-bandwidth routes. They are expected to reach the market sooner than GCS.
Like its peers, Intel is designing high-performance CPUs and AI accelerators that require both: glass interposers for chiplet architectures with HBM, and glass core for long-term stability and integrity in large packages. This isn’t trivial: each family demands different materials, equipment, and process parameters, and the cost/performance balance is still under development.
South Korea Takes the Lead: Samsung and SKC Absolics Set the Pace
The current market snapshot is well summarized by a Seoul seminar: while Taiwan and Japan dominate many areas of the substrates market, Korean companies lead in glass for packaging. Today, the most prominent names include:
- SKC Absolics: a pioneer with its plant in Covington (Georgia, USA), inaugurated after an initial $600 million investment in 2022. They produce samples of GCS and are preparing a second, six-times larger factory (72,000 m²) after completing financing to scale to volume. Absolics is validating quality with customers and in discussions with AMD.
- Samsung Electro-Mechanics (SEMCO) and Samsung Electronics: they have established a pilot line in Sejong capable of producing 500×500 mm sheets with TGV. The local Chemtronics plans to supply these sheets by year’s end and build its own processing line. SEMCO is working on GCS and interposers simultaneously, and — according to sources — is in discussions with Intel about supply; meanwhile, Samsung Electronics pushes glass interposers to meet urgent demand.
The Korean advantage isn’t just capital: Samsung can leverage its experience and supply chain from Samsung Display (engraving, transport, logistics). In a more fragile substrate than silicon, handling and yield are as critical as design.
The U.S. Isn’t Falling Behind… Backed by Korean Capital
The Absolics plant in Georgia is already producing GCS samples and aims to launch its second factory in the medium term. The expansion plan is straightforward: validation with customers, stepped-up CAPEX, and a ramp-up schedule that, industry estimates suggest, will not reach volume until 2028 (with testing and pre-series happening in 2027 if all goes well). This timeline aligns with other players: build, equip, optimize processes, and finally, mass production.
What Benefits Would Intel Gain from a Partial Partnership with Samsung?
For Intel, an agreement with Samsung would provide three advantages:
- Time: Access to pilot lines and an ongoing ramp-up (Sejong), enabling it to speed up useful samples for validation.
- Diversification: Avoid dependence on a single supplier — TSMC for silicon interposers now, and potentially glass in the future — and increase resilience against bottlenecks.
- Cross-knowledge: Collaborate with a manufacturer that controls everything from TGV and glass processes to logistics and equipment (including know-how from Samsung Display).
That said, industry players don’t commit for life: AMD is talking to Absolics; Intel is in discussions with Samsung; NVIDIA, Tesla, and Apple are interested in glass substrates, and designers are avoiding reliance on a single supplier as standardization efforts continue.
A Realistic Timeline: 2027 as a “Bridge,” 2028 as the Launch Pad
One recurring message from manufacturers is that timelines aren’t being rushed. Building, equipping, optimally processing, and raising yields to acceptable levels take years. Several industry players (including Samsung Electro-Mechanics) estimate that the launch of GCS products will occur in the 2027–2028 window, with a higher likelihood of volume from 2028. For interposers, there’s more evidence of an earlier arrival, since they are less complex than GCS.
Practically, the path for Intel and others will involve working with samples and pilot lots from 2025–2027, finalizing designs, materials, and interconnection routes, and preparing for mass production once suppliers give the go.
Not All Advantages: What Remains Unresolved
- Standards and Design (build-up): Achieving compatibility and repeatability among suppliers requires standardizing parameters (thicknesses, vias, metallization, ground plans, CTE).
- Glass processing: TGV and large-panel handling hit mechanical limits; glass is brittle, and initial performance may be the Achilles’ heel of costs.
- Logistics: Moving sheets with thousands of vias without microcracks demands materials, tools, and procedures currently mastered by glass manufacturers and etching and laser suppliers (many from the display industry).
- Cost/Benefit: until yields improve, GCS can be more expensive than organic substrates; the sweet spot will arrive when the best signals, density, and stability offset the costs.
A Multi-Front Race: Who Arrives First and With What?
The competition is twofold: a technological race to see which family — glass core or interposer — reaches volume first; and a capacity race to see which provider scales earlier and better. Absolics has started showing samples in the US; Samsung is accelerating with Sejong; other players in Japan, Taiwan, and the US are gearing up. Meanwhile, high-volume customers (hyperscalers and fabless) are testing all options.
For Intel, the strategy with Samsung is pragmatic: secure supply and extra buffer. For Samsung, locking in an anchor customer validates its bet and accelerates standardization. In a market that’s expected to “explode” — in the best sense — driven by AI demand, it’s a race with a quorum: there will be room for several, but those who get there first will hold advantages.
Implications for the “Post-Nanometer Era”
After a decade focused on nodes and transistors, the competitive edge now shifts toward how dies, HBM, and chiplets are packaged. Glass substrates are the next link with the potential to shift the paradigm: better signal integrity, thermal stability, and density. Those who arrive early with volume and standardization will have a push over their competitors. That’s why Intel, TSMC, AMD, NVIDIA, Tesla, and Apple are already testing and reserving capacity.
Conclusion
The potential supply agreement for glass substrates between Intel and Samsung fits the 2025 landscape: packaging is the new battleground, with glass poised to be its cornerstone. Intel gains time, diversifies, and partners with a powerful and know-how-rich ally; Samsung accelerates its entry with a demanding and visible customer. In the short term, the focus is on validation and pilot projects; in the medium term, the launch is set for 2027–2028. The race today isn’t about nanometers, but about laminations, vias, and standards. And it’s already underway.
Frequently Asked Questions (FAQs)
What is the difference between a “glass core substrate” and a “glass interposer” in chip packaging?
The glass core substrate (GCS) replaces the organic core of the substrate with glass, improving stability, CTE, and signal integrity of the whole package. The glass interposer acts as a “bridge” in 2.5D/3D, connecting dies (CPU/GPU/HBM) with high-bandwidth routes; it’s generally less complex than GCS and might reach volume earlier.
Why do CPU/GPU manufacturers prefer glass substrates over organic or silicon interposers?
Glass offers lower thermal expansion, better flatness, and superior dielectric behavior, leading to cleaner signals, higher interconnection density, and more stable packages. Compared to silicon, glass can be more economical at large sizes and enables TGV (vias through glass), with better scalability in large formats.
When will volume production of glass substrates for AI chips begin?
Most realistic estimates place volume starting from 2028, with pilot and pre-series runs occurring in 2027. The ramp-up—reaching full capacity after initial deployment—typically takes months of adjustment to equipment and process variables until achieving industrial yields.
Which companies lead the glass substrate ecosystem and who are the major designers talking to?
In Korea, SKC Absolics (Covington, GA plant) and Samsung (SEMCO and Samsung Electronics) are at the forefront: Absolics produces GCS samples and engages with AMD; Samsung operates a pilot line in Sejong for TGV and discusses supply with Intel. Meanwhile, other regions like Taiwan, Japan, and the US are preparing capacities, with clients diversifying to avoid sole dependencies.
Sources: thelec.kr and elchapuzasinformatico