Analysts: AI infrastructure build‑out is shifting demand to mature nodes, boosting GlobalFoundries
- Analysts this week argued GlobalFoundries is gaining from AI spending moving beyond flagship graphics processors into older-node chips for power, radio, and optical links. - GlobalFoundries has already committed $16 billion to New York and Vermont, then added silicon photonics and gallium nitride deals tied to AI data centers. - Morgan Stanley says more than 80% of nearly $3 trillion in AI infrastructure spending still lies ahead. (morganstanley.com)
Analysts say the next leg of artificial-intelligence chip demand is landing at GlobalFoundries, not just at the companies making the biggest training processors. (finance.yahoo.com) (investorplace.com) The argument is simple: an AI server is not only a graphics processor. It also needs chips that manage power, move signals, connect optics, and handle radio-frequency and mixed-signal jobs that are often built on older manufacturing nodes. (finance.yahoo.com) (investorplace.com) InvestorPlace said Sunday that mature-node chips in the 28-nanometer to 90-nanometer range are becoming bottlenecks as the AI build-out spreads through the semiconductor supply chain. Yahoo Finance published a similar view five days earlier, pointing to tighter supply-demand and higher mature-node wafer pricing signals from peers. (investorplace.com) (finance.yahoo.com) GlobalFoundries has spent the past year aligning its factories and partnerships with that thesis. On June 4, 2025, it announced a $16 billion plan to expand manufacturing and advanced packaging in New York and Vermont. (gf.com) In that announcement, Chief Executive Tim Breen said the “AI revolution” was driving durable demand for the company’s silicon photonics, gallium nitride power technology, and low-power FDX platform. GlobalFoundries also named Apple, SpaceX, Advanced Micro Devices, Qualcomm, NXP, and General Motors as partners supporting more U.S.-made chip production. (gf.com) Silicon photonics is one piece of that bet. Instead of pushing data through copper wires, it uses light to move information faster and with less power inside and between data centers. (gf.com) On November 17, 2025, GlobalFoundries bought Singapore-based Advanced Micro Foundry and said the deal made it the largest pure-play silicon photonics foundry by revenue. The company said the acquisition would expand production capacity for optical links used in AI data centers and telecom networks. (gf.com) Power is the other piece. On November 20, 2025, GlobalFoundries and Navitas Semiconductor announced a long-term partnership to develop and manufacture gallium nitride chips in Burlington, Vermont for AI data centers and other high-power systems, with development starting in early 2026. (gf.com) That matters because AI spending is no longer confined to a few accelerator chips. Morgan Stanley said in March that nearly $3 trillion of AI-related infrastructure investment is expected by 2028, with more than 80% of that spending still ahead. (morganstanley.com) The bullish case for GlobalFoundries is that more of that money now has to flow into the “supporting cast” of semiconductors: power conversion, optical interconnects, communications chips, and other specialty parts built on mature nodes. The risk, as Yahoo Finance noted, is that pricing pressure in smart mobile and contract resets could still weigh on margins if demand softens. (finance.yahoo.com) So the story around AI chips is widening. The biggest processors still dominate headlines, but the next constraint may be the older, cheaper chips and optical parts that keep the whole system running. (investorplace.com) (gf.com)
