What's happening
Nvidia has reserved the majority of Taiwan Semiconductor Manufacturing Company's most advanced packaging capacity, specifically the CoWoS-L technology used for AI chips including Blackwell GPUs. TSMC's CoWoS technology is experiencing 80% compound annual growth, with packaging solutions head Paul Rousseau stating the numbers "are growing very substantially." The technology addresses a critical bottleneck by bringing high-bandwidth memory directly beside compute chips, as Rousseau explained: "You just can't get enough memory inside your compute chip to fully utilize it."
TSMC is scrambling to expand capacity, building two new packaging facilities in Arizona in 2026 alongside ramping two new sites in Taiwan. Currently, TSMC sends 100% of chips from its Arizona fab to Taiwan for advanced packaging, but the new U.S. facilities will provide domestic packaging capabilities. Intel is also expanding its packaging services, having secured a $5 billion investment from Nvidia in 2025 and $8.9 billion from the U.S. government, with customers including Amazon, Cisco, SpaceX, Tesla, and xAI since 2022.
Why it matters for markets
Nvidia's control over TSMC's advanced packaging capacity represents a strategic chokepoint in AI chip production that could significantly impact competitor timelines and market positioning. With Nvidia trading at $182.08 and holding a $4.43 trillion market cap, this packaging dominance reinforces its competitive moat in the AI semiconductor market. TSMC shares have surged 5.96% to $365.90, reflecting investor recognition of the company's critical role in the AI supply chain through its $1.90 trillion market capitalization.
The packaging bottleneck emerges as a new constraint beyond traditional semiconductor manufacturing, with Georgetown University's John VerWey warning it "can emerge as a bottleneck very quickly if people are not making the CapEx investments proactively." ASE's advanced packaging business is expected to double to $3.2 billion in 2026, indicating the scale of investment required to meet demand. This development follows Nvidia's recent Rubin chip production scale-back due to memory shortages, demonstrating how supply chain constraints continue to shape the AI hardware landscape.
The capacity reservation could delay competitors' AI chip launches and limit their ability to scale production, potentially widening Nvidia's market lead. With TSMC's P/E ratio at 35.3 compared to Nvidia's 37.2, both companies are commanding premium valuations based on their AI market positioning, but supply constraints could affect relative performance as the market develops.
Sectors and assets to watch
Semiconductor companies dependent on TSMC's advanced packaging face potential supply constraints, with AMD, Qualcomm, and Broadcom likely competing for remaining CoWoS capacity. Intel's foundry services division could benefit as companies seek alternative packaging solutions, with the company already serving major clients and receiving significant government backing. Assembly and test services providers like ASE Group stand to gain from the capacity shortage, with ASE's packaging business projected to reach $3.2 billion in 2026.
Cloud computing giants including Amazon, Microsoft, and Google may face delays in custom AI chip development if they cannot secure adequate packaging capacity for their silicon designs. Memory manufacturers like SK Hynix and Micron could see increased demand for high-bandwidth memory as more AI chips utilize advanced packaging technologies that integrate memory and compute components.
What to watch next
Monitor TSMC's quarterly earnings for updates on packaging capacity expansion timelines and capital expenditure commitments, particularly regarding the Arizona facilities scheduled for 2026. Track Intel's foundry services revenue growth and customer wins as companies potentially diversify away from TSMC for packaging needs. Watch for announcements from other semiconductor companies regarding packaging partnerships or delays in AI chip launches that could signal supply constraints impacting production schedules.