The semiconductor industry's growth is no longer just about smaller transistors, but a fundamental shift to new materials and 3D structures driven by relentless AI demand, according to a new report from Nomura.
A new analysis from Nomura Securities argues the semiconductor industry is abandoning its decades-long reliance on Moore’s Law in favor of a complex combination of 3D transistor architectures, backside power delivery, and a host of new materials including glass and photonic silicon. The shift, driven by the insatiable computing demand from artificial intelligence, is set to reorder the sector’s value chain, with 2027 emerging as a key inflection point for mass adoption.
"The semiconductor industry's growth logic has shifted from transistor density improvement to a combination of 3D transistors, backside power delivery, and diverse new material innovations," Donnie Teng's Asia-Pacific technology team at Nomura said in the report. The analysis suggests that the core challenge of AI is not just a matter of producing more GPUs, but of rewriting the underlying manufacturing process to overcome the physical limits of silicon.
The report identifies a clear timeline for the transition, with technologies like Gate-All-Around (GAA) transistors and SoIC hybrid bonding starting to ramp in 2026. The major shift occurs in 2027, when backside power delivery, wafer-bonded NAND, and glass core substrates are projected to begin mass production. This transition will dramatically increase the value of materials, with the price of specialized metal-oxide photoresist for high-NA EUV machines expected to be two to eight times higher than current materials.
For investors, this marks a systemic re-evaluation of the entire supply chain. According to a recent UBS report on AI investing, the central challenge is determining where value will be sustainably created. The Nomura analysis provides a roadmap, suggesting that the biggest winners may not be the most visible names, but the "picks-and-shovels" providers of new materials, specialty chemicals, and advanced packaging solutions that enable the next wave of AI chips.
From Silicon to Glass and Light
The end of simple scaling is forcing chip designers to look for performance gains through new materials and advanced packaging. The Nomura report highlights glass core substrates as a key technology to watch, offering superior thermal and electrical properties compared to traditional organic substrates for the massive, high-power chips used in AI. Broadcom is expected to be a first-mover, potentially using glass substrates for its switching ASICs in 2027, with Intel also heavily invested in the technology.
At the same time, the data bottleneck inside AI data centers is pushing the industry from electrical to optical communication. This elevates the importance of compound semiconductors like Indium Phosphide (InP), a core material for high-speed lasers that Nomura expects to remain in tight supply through 2027. An alternative route is Photonic SOI (silicon-on-insulator), a technology dominated by Soitec that allows for the integration of optical components directly onto silicon and is seen as essential for future co-packaged optics.
The 2027 Inflection Point
The convergence of these new technologies around 2027 is expected to create a significant supply-demand crunch for the foundational material of the entire industry: 12-inch silicon wafers. Nomura forecasts that new manufacturing techniques like backside power delivery, which requires bonding two wafers together, will nearly double wafer consumption per finished chip. Combined with fab expansions from TSMC, Samsung, and Intel, this could push overall annual demand growth for 12-inch wafers to nearly 10 percent, outstripping the pace of new supply and restoring pricing power to wafer suppliers like Shin-Etsu and SUMCO.
TSMC’s aggressive expansion and regional procurement strategy acts as a major catalyst, creating opportunities for local suppliers of CMP consumables, specialty gases, and packaging materials to get designed into the world’s most advanced manufacturing flows. This dynamic supports the view from other market observers, like Analog Devices, which in its recent earnings call cited robust industrial and AI demand as key drivers, illustrating how value is captured across the enabling layer of the AI economy.
This article is for informational purposes only and does not constitute investment advice.
