TSMC has surpassed Intel in the number of US patent applications for silicon photonics (SiPh), a crucial optoelectronic integration technology essential for artificial intelligence (AI) and data center applications. A survey by Nikkei BP, using Patentfield data covering applications from January 2015 to July 2025, found that Intel's filings in the sector have declined, while TSMC's have steadily increased.
In 2023, Intel and TSMC filed 43 and 46 US patents, respectively, in SiPh, showing near parity. In 2024, TSMC submitted approximately 50 patents, nearly double Intel's 26 applications, demonstrating its rapid progress in next-generation low-power AI computing technologies.
TSMC plans to start mass production of chips using co-packaged optics (CPO) technology in 2026, whereas Intel remains in the research and validation phase. In related technologies such as complementary field-effect transistors (CFET) and back-side power delivery networks (BSPDN), TSMC and Samsung Electronics continued to increase filings, while Intel's pace slowed. The decline in Intel's patent applications reflects ongoing challenges in its wafer manufacturing and logic IC divisions, despite support from the US government and Japan's SoftBank Group, according to Nikkei BP.
Other major semiconductor companies are also advancing silicon photonics to boost data transmission speeds and energy efficiency. Nvidia has integrated SiPh into its Quantum InfiniBand and Spectrum Ethernet switches to enhance AI data center performance, supporting configurations up to 1.6Tbps per port. Nvidia's Quantum 3450-LD InfiniBand switch is expected in late 2025, with Spectrum SN6810 and SN6800 Ethernet switches following in 2026, according to a TechRadar report.
AMD acquired Enosemi, a Silicon Valley startup specializing in photonic ICs, to strengthen its AI and SiPh capabilities and compete with Nvidia, according to Tom's Hardware. GlobalFoundries has increased its overall investment to US$16 billion, allocating US$3 billion for R&D in emerging semiconductor technologies, including silicon photonics for data centers and quantum computing applications, Reuters reported.
Article edited by Jack Wu
