According to www.digitimes.com, export controls imposed by the United States are intensifying pressure on the optical communications supply chain serving artificial intelligence data centers — with direct implications for Taiwanese compound semiconductor wafer supplier Visual Photonics Epitaxy Co. (VPEC).
VPEC’s Strategic Pivot Amid Regulatory Uncertainty
VPEC, headquartered in Taipei, confirmed that demand for its indium phosphide (InP) and gallium arsenide (GaAs) epitaxial wafers — critical substrates for high-speed optical transceivers used in AI cluster interconnects — is accelerating rapidly. The company stated it expects 2026 to be a pivotal year for scaling production capacity as hyperscaler-driven demand surges. According to the report, VPEC plans to allocate over 70% of its new capital expenditure toward optical component manufacturing lines this fiscal year — up from 45% in 2024.
Geopolitical Constraints on Materials Flow
The U.S. Department of Commerce’s Bureau of Industry and Security (BIS) expanded its Entity List restrictions in May 2026, adding three Chinese optical module integrators and two Taiwanese test-and-assembly subcontractors linked to AI infrastructure deployments. These controls specifically target equipment capable of supporting >800 Gbps optical transmission — a threshold now standard in NVIDIA’s GB200 Grace Blackwell systems and Meta’s next-generation AI racks. As a result, VPEC’s export licenses for shipments to mainland China-based customers have faced extended review timelines, averaging 112 days per application in Q1 2026, versus 22 days in Q1 2023.
Industry-Wide Ripple Effects
This regulatory tightening coincides with broader industry shifts. Huawei’s investment in InP chip startup LightMatter Technologies — announced in April 2026 — reflects parallel efforts to localize high-bandwidth optical IC design. Meanwhile, NVIDIA has publicly signaled its long-term transition to co-packaged optics (CPO), but cited persistent shortages of InP wafers and silicon photonics (SiPh) packaging substrates as key bottlenecks. According to the source, global InP wafer production capacity remains concentrated in just five fabrication facilities — two in Taiwan, one in Japan, one in Germany, and one in the U.S. — limiting rapid geographic diversification.
Supply Chain Implications for Practitioners
For supply chain professionals managing AI hardware procurement, these developments necessitate dual-sourcing strategies for optical interconnect components and tighter validation of end-use compliance documentation. A 2025 survey by the Semiconductor Industry Association found that 68% of AI infrastructure OEMs now require full traceability from wafer substrate through final optical module assembly — a requirement that adds 14–19 business days to average lead times. VPEC’s decision to prioritize SiPh-compatible substrates in its 2026 roadmap aligns with this trend, though the company acknowledged in internal briefings that yield rates for 300-mm InP-on-silicon wafers remain below 62% — well under the 85% benchmark required for cost-effective volume deployment.
Source: www.digitimes.com
Compiled from international media by the SCI.AI editorial team.
