AI news
1. TSMC and Amkor Commence Mass Production of Silicon--Photonics Interconnects for AI Clusters
Date: June 14, 2026
TSMC, in partnership with Amkor Technology, has announced the commencement of mass production for the industry’s first Silicon--Photonics (SiPh) Interconnects. This technology represents a critical pivot in hardware architecture, addressing the "memory wall" and energy bottlenecks currently hindering the scaling of massive AI training clusters.
| Feature | Impact / Specification |
|---|---|
| Data Transfer | Light--based (Optical) vs. Traditional Copper |
| Latency Reduction | 30% decrease in inter--chip communication |
| Energy Efficiency | 40% reduction in interconnect power draw |
| Deployment | Integration into 2nm--class AI accelerators |
Key Facts & Strategic Impact:
- Scaling Frontier Models: The shift to optical interconnects allows for the seamless scaling of clusters beyond 100,000 GPUs without the exponential heat generation and signal degradation associated with electrical signaling.
- Supply Chain Shift: Amkor’s advanced packaging facilities in Arizona and Taiwan are the primary sites for this integration, signaling a strengthening of the US--Taiwan semiconductor corridor for high--end AI compute.
- Market Forecast: Industry analysts expect Silicon--Photonics to represent 25% of the total AI hardware market value by 2028 as data centers reach the physical limits of copper.
Sources: TSMC Press Center, Amkor Investor Relations