The semiconductor sector saw a series of developments during the week ending October 9, 2026, spanning manufacturing and advanced packaging to data center infrastructure and quantum computing. While no single announcement represents a comprehensive market transformation, taken together, these developments illustrate the scale of pressure that AI expansion is placing on the entire chip supply chain.
Advanced Packaging Becomes a New Area of Expansion
TSMC and GlobalFoundries signed a $2 billion manufacturing agreement under which GlobalFoundries will produce the silicon interposers used in TSMC's CoWoS advanced packaging system at its facility in Malta, New York. High-volume production is expected to begin during the first half of 2028.
In a related development, Applied Materials and Intel expanded their research collaboration to develop advanced transistors, interconnects, and packaging technologies, while Rapidus launched the Rapidus CORE framework to support customers from chip design through manufacturing as part of its plans for a 2nm foundry, with 17 design service partners in the first phase.
Xanadu and GlobalFoundries also formed a multiyear partnership to transfer low-loss silicon nitride photonics technologies and superconducting single-photon detectors to GlobalFoundries' 300mm production line in New York.
AI Pressure Extends to Power and Memory
Data from the SIA and WSTS indicated that global semiconductor sales reached $159.7 billion in August, an 8% increase from July, while sales since the beginning of the year exceeded $1 trillion. McKinsey expects inference workloads to account for approximately 60% of AI computing by 2030, driving demand for specialized accelerators, HBM, interconnected chips, advanced packaging, and silicon photonics.
Meanwhile, Brookings estimated that AI expansion in the United States could require $10.3 trillion through 2032 to fund data centers, power, networks, and specialized chips, warning that financing complexity could conceal increasing financial risks.
Wolfspeed received a conditional U.S. government funding commitment of up to $1.5 billion over 30 years to expand production of silicon carbide materials and wide-bandgap power devices. Infineon also announced a 27kW three-phase power supply solution for AI data centers, while Microchip and Navitas collaborated on a reference design that converts 800V to 6V for developing 800VDC data center racks.
Future Manufacturing Technologies, but a Long Timeline
Researchers from ASML and Carl Zeiss presented a proposed architecture for Hyper-NA EUV technology with a numerical aperture of at least 0.75, which could enable the manufacture of features as small as 5nm, more than one-third smaller than those enabled by current High-NA systems. However, the technology may require approximately a decade before it is ready, and there is no announced commitment to introduce it into commercial production.
Canon said that its nanoimprint lithography technology had entered its final phase before large-scale production. Meanwhile, the Technology & Statecraft Center considers DUV immersion systems to remain a major bottleneck for AI chip manufacturing in China, estimating that Chinese fabs had accumulated more than 330 systems in this category by the first quarter of 2026.
Security and Supply Chains Under Scrutiny
The U.S. Department of Justice accused the owner of Earthmade Computer of smuggling export-controlled servers worth more than $300 million to China, allegedly using misleading information about the end user and destination and routing the shipments through Malaysia and Singapore. After modeling more than 10 million scenarios to detect the diversion of AI chips, CSET researchers concluded that verifying location through communication signals may be the lowest-cost option, with its effectiveness potentially improved by combining it with targeted physical inspections.
Why Does This Matter?
These developments show that the challenge is no longer limited to access to smaller manufacturing nodes. The actual expansion of AI also requires packaging capable of integrating multiple chips, high-bandwidth memory, higher-voltage power, optical interconnects, and mechanisms for verifying chip movements across borders. Many of the most ambitious technologies, such as Hyper-NA EUV and quantum computing, remain in the research or early scale-up stages; therefore, a research announcement should not be equated with available commercial production capacity.