China's Indium Phosphide Surge

A Coming Supply Chain Disruption for Global AI Infrastructure

Indium phosphide has emerged as a critical bottleneck for the AI industry. For this piece, we will focus on China and analyze how Chinese companies’ ramping up capacities will impact global supply and market dynamics.

The Chinese players in this space have not received much attention from global investors. But as in the silicon carbide industry, where companies like Tianyu, Sanan, and TankeBlue have systematically displaced Western incumbents through aggressive capacity expansion and cost reduction, the indium phosphide sector could follow a similar trajectory.

Of course, significant challenges remain for China to achieve dominance—challenges we will examine in detail. But the strategic pieces are falling into place faster than most analysts realize.

The Demand Explosion: AI’s Insatiable Appetite

The numbers tell a compelling story. Goldman Sachs projects that 800G optical module shipments will exceed 34 million units in 2026, with 1.6T modules surpassing 25 million units. These modules are the arteries of AI infrastructure, connecting GPUs in massive training clusters and enabling the data throughput that makes frontier AI models possible.

Each module requires indium phosphide substrates—the crystalline wafers on which laser chips are fabricated. The consumption intensity varies dramatically by technology. EML (Electroabsorption Modulated Laser) solutions are the most substrate-intensive, requiring multiple chips per module with substantial die area. Silicon photonics modules use CW (Continuous Wave) lasers as light sources, theoretically reducing indium phosphide consumption per module.

However, the market has underestimated a critical technical detail: CW laser power is rapidly escalating. Low-power 70mW or 100mW CW sources can yield many dies per wafer. But as power requirements climb to 200mW, 300mW, or even 400mW—driven by CPO (Co-Packaged Optics) and future OIO (Optical I/O) architectures—chip area expands significantly and yield drops. A 3-inch wafer producing 400mW CW sources might yield only 200+ dies, compared to several times that number for 70mW devices.

The shift to high-power CW could paradoxically increase total indium phosphide consumption, partially or fully offsetting the reduction from fewer chips per module.

Translating this into substrate demand is complicated by inconsistent industry metrics. Market estimates ranging from 600,000 to 1.5 million wafers often mix physical wafers, 2-inch equivalents, 3-inch equivalents, and include safety stock orders. With supply chain instability and export control uncertainty, downstream customers are aggressively locking in supply through framework agreements, inflating short-term order data beyond actual consumption. Nevertheless, the directional trend is unambiguous: demand is surging, and 2026 will see new demand reach the range of 100,000 to 200,000+ wafers in 3-inch equivalent terms.

The Incumbent Oligopoly: Concentrated and Constrained

Global indium phosphide substrate supply is extraordinarily concentrated. Three players control over 90% of the market: Japan’s Sumitomo Electric (~42% share, ~800,000 wafer annual capacity), U. S.-based AXT including its Chinese subsidiary Beijing Tongmei (~36% share, ~300,000 wafer capacity), and Japan’s JX Nippon Mining & Metals (~13% share, ~200,000 wafer output).

But capacity figures mask structural supply tightness. The real shortage is in 3-inch and 4-inch high-quality wafers—particularly semi-insulating substrates with stringent consistency requirements for high-end customers. The 2-inch market faces less pressure, while 6-inch remains in R&D and pilot production, unable to contribute meaningful volume in the near term.

Sumitomo’s capacity is largely saturated, with a significant portion now consumed internally as the company vertically integrates into optical chips and modules. AXT’s capacity is running near full utilization, shipping approximately 25,000-26,000 wafers monthly. JX Metals has no announced expansion plans. The incumbents are capacity-constrained precisely when demand is accelerating.

Moreover, technical barriers remain formidable. Indium phosphide crystal growth is far more challenging than market participants appreciate. The equipment itself—single crystal furnaces—is not the primary bottleneck; supply in China is adequate and costs are manageable. The real barrier is process control during crystal growth.