Preface:

The first critical material bottleneck triggered by the AI computing power wave has emerged in indium phosphide. Global leading suppliers have their orders booked through 2027, with industry-wide inventory levels at a historic low of just three months.

AI Demand Explosion: Why is Indium Phosphide Irreplaceable?

In early 2025, a 2-inch optoelectronic-grade indium phosphide (InP) substrate cost just $800, but by April 2026, prices had surged to $2,300-$2,500—nearly tripling. The 6-inch high-end substrate story is even more dramatic, jumping from $1,400 to $5,000—a 250%+ increase. The core driver behind this price spike is the exponential growth in demand for high-speed optical modules from AI data centers.

Indium phosphide serves as the foundational substrate material for 800G, 1.6T, and higher-end optical modules. Whether for EML (electro-absorption modulated lasers), APD (avalanche photodiodes), or high-power CW chips, indium phosphide remains the only viable base material with no mass-production alternatives. Why? Its direct bandgap structure enables near-100% photoelectric conversion efficiency while precisely matching the 1310nm and 1550nm golden wavelength bands critical for fiber-optic communication—a feat silicon cannot achieve.

AI computing power's demand pull on indium phosphide is "multiplicative." A single AI server requires over 10x more optical modules than conventional servers. More critically, 1.6T optical modules consume 2.7-2.8x more indium phosphide substrates than 800G modules. From 800G to 1.6T to 3.2T, each generational upgrade burns through indium phosphide at ever-higher rates.

The numbers speak volumes. Global indium phosphide demand is projected to hit ~2.1 million wafers in 2025, surge to 2.6-3 million in 2026, and exceed 4 million by 2027—representing over 50% annual growth. Overseas leaders even predict 85% annual demand growth in AI applications by 2030. Currently, over 80% of indium phosphide demand comes from AI data centers. Simply put: the hotter AI data centers get, the scarcer indium phosphide becomes.

Supply-Side "Deadlock": Global Capacity Monopolized by Three Players

While demand skyrockets, supply remains rigid. Indium phosphide production is highly concentrated, with over 90% of capacity controlled by three Japanese-U.S. companies: Sumitomo (43%), AXT (35%), and JX Metals (13%). These players directly dictate global pricing, while domestic localization rates for high-end 6-inch indium phosphide substrates remain below 5%.

Worse still, capacity cannot be rapidly scaled. Production requires extreme technical precision—growing crystals under high-temperature/high-pressure conditions. Achieving acceptable yields for large-diameter, low-defect substrates takes 3-5 years, with single production line investments exceeding $170 million. Without a decade of technical accumulation, production remains impossible. Line construction also takes 18-24 months, with core equipment (MOCVD) delivery cycles stretching 1-2 years. Even if expansions start now, new capacity won't materialize until after 2027.

In 2026, global effective indium phosphide capacity will reach only 600,000-750,000 wafers against market demand of 2.6-3 million—a 70%+ gap, meaning four demand units for every one supply unit. Industry-wide inventory covers just three months of usage, forcing downstream players to place orders 2-3 months in advance with 30-50% deposits to lock capacity. Some companies have even pre-booked 100,000 substrates at once. This is the reality of today's "high prices, no availability" market.

Technology Route Battles: Indium Phosphide vs. Silicon Photonics vs. Thin-Film Lithium Niobate

Indium phosphide shortages are pushing the entire optical module supply chain to seek alternatives. For 800G-1.6T pluggable modules, indium phosphide-based EML technology remains the "supply bottleneck layer" controlling the industrial chain (industry chain). Lumentum plans to boost EML capacity by 50%+ by end-2026 versus 2025, while Coherent management teases ‘super-doubling’ indium phosphide capacity between 2026-2027.

But the gap remains vast, opening doors for silicon photonics. China Galaxy Securities projects silicon photonics to account for 50%+ of 800G optical modules and 70-80% of 1.6T modules by 2026. Leveraging mature silicon-based CMOS processes to integrate optical components on silicon wafers, silicon photonics is becoming a strategic path to offset indium phosphide shortages.

Looking further ahead, thin-film lithium niobate is emerging with its ultra-high bandwidth, low power consumption, and low-loss advantages. Huatai Securities calculates the thin-film lithium niobate modulator market driven by 3.2T optical modules alone could approach $420 million by 2031, implying a 271% CAGR from 2029-2031. Indium phosphide, silicon photonics, and thin-film lithium niobate are transitioning from single-material competition to multi-material ecosystem collaboration.

Industry Landscape Fractures: Expansions, Controls, and Localization

The massive supply-demand gap is accelerating global capacity layout (deployments). Overseas giants lead the charge. AXT has initiated expansions targeting capacity doubling by end-2026, but its order backlog already exceeds $60 million—a record high—with multiple clients signing long-term agreements to secure capacity. Lumentum has fully allocated its indium phosphide wafer fab capacity, planning ~40% unit capacity expansion in coming quarters.

Domestically, Yunnan Germanium Industry announced a $27 million expansion of high-quality indium phosphide single-crystal wafer projects in April 2026, reaching 450,000 wafers (4-inch equivalent) annual capacity post-completion. Zhuhai Dingtai Xinyuan is also actively expanding, while Henan Mingjia Semiconductor's Phase II project will produce nearly 30 tons of indium phosphide annually—covering nearly half of domestic demand.

However, indium phosphide's core upstream material is metallic indium, a byproduct of zinc-lead mining that cannot be independently scaled. In early 2025, China imposed export controls on indium, slashing overseas supply and further inflating raw material costs. Chinese refined indium prices soared from ~$400/kg in late 2025 to $707/kg (76%+ annual increase)—a decade high. Indium price changes are now cascading through the supply chain to indium phosphide substrates and then to optical communication modules. AI-driven "computing power scarcity" has now reached "resource scarcity" at the elemental level.

Trend Outlook: Prices to Rise Through 2027, with Disruptions Just Beginning

In the short term, capacity gaps will worsen for two years as downstream demand grows 50%+ annually, creating an irreconcilable supply-demand scissors gap ( scissors gap , scissors gap). Meanwhile, silicon photonics' share is rising rapidly, but high-end EML/CW laser performance remains irreplaceable. NVIDIA has monopolized EML chip capacity, pushing delivery cycles beyond 2027 and opening windows for domestic alternatives.

Long-term, China controls 70% of global indium production and enforces export controls. Against intensifying global AI computing power competition, this resource's strategic elevation is triggering profound realignments in international industry structures.

Conclusion:

The indium phosphide price frenzy under AI's spotlight is merely the tip of a deeper industrial restructuring iceberg. Chinese companies that ultimately secure positions in the high-performance compound semiconductor supply chain while strategically revaluing upstream indium resources may emerge as the biggest winners in this value redistribution.

Online References:

36 krypton : "AI Boom's 'Golden Material': The Deep Logic Behind Indium Phosphide Price Hikes"

DoNews: "Optical Modules Accelerate to 1.6T/3.2T, with Indium Phosphide Substrates and Thin-Film Lithium Niobate as New Bottlenecks"

Securities Times: "Yunnan Germanium Industry Plans $27 Million High-Quality Indium Phosphide Single-Crystal Wafer Project"