NVIDIA Pushes Into Scale-Up CPO as TSMC Targets 10,000 PIC Chips a Month in 2027
Co-packaged optics, or CPO, integrates optical engines with computing chips to reduce power consumption and latency in high-speed data transmission. NVIDIA is introducing CPO into its GPU scale-up interconnect architecture, driving AI data centers to shift from copper wiring to optical interconnects. TSMC’s photonic integrated circuit, or PIC, capacity has consequently become critical to supply-chain expansion.
Morgan Stanley expects NVIDIA’s optical-engine shipments to grow significantly in 2026, with CPO penetration in AI data centers rising each year. To meet demand for AI computing power, TSMC plans to increase monthly PIC chip capacity to 10,000 units in the first quarter of 2027, accelerating its expansion into scale-up optical interconnects.
All Coverage
2 original reportsThe Backstory
The history behind this eventTSMC Targets CPO Mass Production as Supply Bottlenecks Shift to Lasers and Fiber
Surging artificial intelligence workloads are straining the bandwidth and power efficiency of conventional electrical links between chips. Silicon photonics and co-packaged optics, or CPO, place optical components closer to processors, reducing transmission distances and energy use. The technology is increasingly viewed as a critical interconnect architecture for next-generation AI data centers as computing capacity expands.
TSMC Vice President Hsu Kuo-chin said silicon photonics and CPO are expected to enter mass production in the second half of 2026 and become a mainstream architecture before 2027, pushing back against doubts over large-scale manufacturing. Taiwan’s foundry sector can already produce optical engines, he said, while lasers, optical fiber, connectors and product testing remain the main supply-chain constraints.
Subscribe to Mark Radar Weekly
Every Friday, the week's strongest signals in your inbox. Unsubscribe anytime.
If you search news on Google, you can set Mark Radar as a preferred source—our coverage will show up more often in your results. Set as preferred source on Google →