China’s export restrictions on indium phosphide, a key material for photonic chips used in AI data centers, are causing significant disruptions in the global supply chain, complicating efforts by major U.S. and international tech firms to scale AI infrastructure.
- China controls 70% of global indium production and restricts exports.
- InP export limits have raised wafer prices by 250%, causing chip shortages.
- US companies invest in domestic capacity but face multi-year build times.
What happened
In early 2025, China imposed export restrictions on indium phosphide (InP), a critical material used in high-speed optical chips for AI data centers. This move followed urgent appeals from U.S.-backed chipmakers like Coherent, who highlighted delays in obtaining export licenses. China's control of the indium market has allowed it to effectively tighten supply, disrupting the global photonics supply chain.
These restrictions sparked broad concerns during U.S.-China trade talks and prompted top executives to raise the issue in diplomatic contexts. The immediate consequence has been a backlog of export permits and a sharp surge in prices, with the average cost for a 6-inch InP wafer jumping 250% to $5,000. Production bottlenecks now threaten to slow the rapid expansion of AI data center infrastructure reliant on photonics technology.
Why it matters
Indium phosphide is indispensable for manufacturing photonic chips that enable data centers to use light signals instead of traditional copper wiring, significantly improving speed and energy efficiency. With AI workloads expanding exponentially, the demand for InP has surged, making it a critical choke point in the AI hardware supply chain. China's dominant position—accounting for around 70% of global indium production as of 2024—gives it leverage to influence the pace of AI development worldwide.
The export restrictions extend China’s use of material supply curbs as a geopolitical tool, reinforcing a trend first observed in its control of rare earth elements. This granular control over upstream compounds and substrates underlines a strategic approach to slow or condition access to advanced technology components, potentially reshaping global AI hardware supply dynamics and raising costs for manufacturers and data center operators.
What to watch next
Industry observers will monitor how quickly U.S. and allied photonics firms can ramp up domestic InP production capacity to reduce dependency on Chinese exports. Significant investments are underway, including Coherent doubling InP wafer output in Texas by 2027, but new production facilities typically require two to three years to become operational, leaving short-term supply tight.
Additionally, the evolution of China’s export policy on InP and related materials will be a key factor. Any further tightening or easing will considerably affect global AI data center buildout timelines. Stakeholders are also watching how multilateral trade negotiations and diplomatic pressure might influence export licensing delays and whether new supplier sources outside China can scale to meet global demand.