As the global race to build artificial intelligence infrastructure accelerates, a critical bottleneck has emerged not in advanced silicon, but in the raw materials required to connect it. China’s export restrictions on indium phosphide (InP), an essential compound for the high-speed optical chips used in AI data centers, are beginning to bite, sending prices soaring and threatening to delay the rollout of next-generation computing clusters worldwide.
The issue has escalated to the highest levels of international diplomacy. In May 2026, Coherent CEO Jim Anderson flew to China as part of a US business delegation to raise concerns directly about delays in InP export licenses, according to Reuters reporting. The matter was also a key point of discussion during trade talks in Seoul ahead of the mid-May summit between US President Donald Trump and Chinese President Xi Jinping, underscoring the strategic importance of a material that most people have never heard of.
The Photonics Bottleneck
Indium phosphide is a core material for photonics, the technology that uses light rather than electricity to transmit data at high speed. In modern AI data centers, where tens of thousands of GPUs must communicate at lightning speed to train massive models, optical transceivers built on InP substrates are indispensable. There is currently no viable substitute that can match its performance at scale.
China controls approximately 70% of global indium output, according to 2024 data from the US Geological Survey. In February 2025, Beijing implemented export restrictions on InP and related materials, requiring exporters to obtain specific licenses before shipping. The move was widely seen as a calibrated act of retaliation for US export controls on advanced semiconductors and chipmaking equipment, a demonstration that China possesses its own set of supply chain levers.
The impact on the global supply chain has been severe and swift. Since the restrictions began, the average price for a 6-inch InP wafer has surged 250% to $5,000.
“InP export permits represent the most significant challenge we currently face,” stated AXT, the world’s second-largest producer of InP substrates and a major supplier to Coherent. AXT’s Chinese subsidiary only received its first export permits last June and continues to face a significant backlog of pending applications.
Supply Chain Disruptions Across the AI Ecosystem
The constraints are rippling through the entire AI hardware ecosystem. Lumentum, a major optical component manufacturer, is reportedly sold out through 2028 despite quadrupling its output. Taiwanese firms VPEC and LandMark Optoelectronics have also faced disruptions in securing InP substrates for their production lines.
The strategic importance of photonics has not gone unnoticed by the major players in the AI boom. In March 2026, Nvidia announced $2 billion investments each into Coherent and Lumentum, $4 billion in total, to secure its optical supply chain, a clear signal that the world’s leading AI chip company views photonic connectivity as a critical vulnerability.
“InP is one of several supply chain bottlenecks collectively gating AI data center buildouts,” noted Konrad Wang of SemiAnalysis. The constraints highlight the vulnerability of the massive infrastructure projects currently underway, which depend on a complex web of specialized materials, many sourced from China.
Paul Triolo of the Albright Stonebridge Group offered a broader strategic assessment: “Beijing is developing a more granular ‘materials chokepoint’ toolkit.” By targeting specific, hard-to-replace materials like InP alongside gallium and germanium, China is demonstrating its ability to exert leverage over the global tech supply chain, even as it faces restrictions on advanced chips.
China’s Domestic Expansion
While Western firms struggle with export licenses, Chinese companies are rapidly expanding their domestic InP capacity to capture the growing global demand. Yunnan Germanium recently announced a 189 million yuan ($28 million) investment to expand its InP capacity to 450,000 wafers per year, following a 74% surge in shipments in 2025. Similarly, Guangdong Xiandao launched a new project via its subsidiary Guangdong Xianrui, targeting annual production of 40 tons of InP crystals.
Currently, market leaders AXT and Japan’s Sumitomo account for roughly 80% of global InP substrate manufacturing, with JX Advanced Metals holding approximately 10%. Sumitomo has reported no impact on production so far, as it consumes much of its output internally.
Western producers are working to reduce their dependence on Chinese-controlled supply. Coherent is doubling its InP wafer capacity at its Texas plant in 2026, with plans to more than double it again by the end of 2027. However, building out alternative supply chains takes time and capital, and the near-term outlook remains constrained.
A New Front in the Tech War
The InP dispute represents a significant evolution in the US-China technology competition. For years, Washington has used export controls on advanced chips and chipmaking equipment as its primary tool to limit China’s AI progress. Beijing’s response has been to identify and exploit its own leverage points in the global supply chain, not just in finished products, but in the raw materials and specialty compounds that underpin the entire AI infrastructure buildout.
This “materials chokepoint” strategy is proving effective precisely because it targets components that are difficult to substitute and require years to develop alternative supply chains for. The InP crisis serves as a warning that the escalating US-China tech war is increasingly being fought at the molecular level, with consequences that extend far beyond the bilateral relationship to affect the entire global AI industry.