InP Supply-Demand Gap Exceeds 70%: Critical Shortage of Key Optical Module Material
Taylor Wilson
Indium phosphide — the core substrate for 800G/1.6T optical modules — faces a supply gap above 70%, with 2-inch wafer prices nearly tripling in one year. This means cost pressure on AI compute infrastructure has no near-term fix.
How wide is the gap?
Global InP substrate demand for 2026 is projected at 2.6–3 million wafers; effective compliant capacity stands at roughly 750,000 — a gap exceeding 70%.
2-inch telecom-grade substrates surged from $800/wafer in early 2025 to $2,300–2,500/wafer by April 2026 — nearly a 3× increase. 6-inch premium substrates jumped from $1,400 to $5,000, up more than 250%.
This means → downstream optical-module makers face not just higher prices but a hard capacity wall — money alone cannot secure supply.
Why can't production keep up?
The InP expansion cycle is painfully long: from crystal-growth furnace installation to customer qualification, the full process takes 18 months to two years — and key equipment must be imported.
Upstream raw materials are equally strained: indium — the rare metal at InP's core — hit ¥5,560/kg as of July 6, double its early-2025 level and a decade-high.
Shenwan Hongyuan estimates InP will drive a 6.77% increase in indium demand by 2027, enough to trigger sharp price swings.
In plain terms = expansion takes two years, raw-material costs keep climbing, and capacity simply cannot match the steep demand curve.
How are export controls deepening the fracture?
In January 2026, China's commerce ministry banned all exports of InP, indium, gallium, and germanium to Japanese military end-users, with civilian shipments subject to strict licensing and end-user review.
Market feedback indicates Japan- and US-based firms now see rejection rates above 80% when applying for Chinese-made InP substrates.
The US had already launched anti-dumping and countervailing-duty probes on Chinese active-anode materials; the EU introduced amendments under its Critical Raw Materials Act to reduce single-country dependence.
This reflects a shift from market-driven shortage to policy-driven fracture — every major bloc is now fighting for supply-chain control.
What are the downstream giants doing?
In March 2026, Nvidia committed $2 billion each to Coherent and a second photonics firm, paired with long-term offtake agreements to lock in years of InP chip capacity. Jensen Huang personally attended the groundbreaking for Coherent's first 6-inch InP wafer-fab expansion.
Lumentum's CEO disclosed that EML laser output — a key laser chip for high-speed optical links — has grown 8× over three years, yet shipments still trail demand by 25–30%.
Huawei's investment arm HiSilicon (Hubble Technology) holds 23.91% of Xinya Semiconductor, making it the second-largest shareholder. In 2025, Huawei locked in 80,000 InP wafers — 53% of Xinya's capacity — with a 40% prepayment, far above the sub-20% industry norm.
This means → top players have shifted from spot buying to capacity lock-ins; latecomers will find it even harder to secure supply.
How is the global capacity race shaping up?
Overseas: AXT plans to add 200 four-inch crystal-growth furnaces and quadruple total capacity by end-2027. Sumitomo Electric is investing roughly ¥18 billion to lift output to 3.1× its FY2024 level by FY2028. Coherent's 6-inch line in Sherman, Texas will hit its 2026 doubling target a quarter early, then double again by end-2027.
China: Yunnan Germanium's subsidiary Xinya has 150,000 wafers/year today; a ¥189 million expansion launched in April 2026 will raise that to 450,000. GRIKIN plans 250,000 wafers/year by late 2027. Pioneer Micro is investing ¥1.7 billion for an eventual 3 million wafers/year, with construction running from August 2026 to August 2029.
Cross-sector capital is flooding in too: leather-goods firm Xingye Tech is bidding ¥55 million for an InP business; Suqian Liansheng announced a phase-one ¥100 million investment in a 120,000-wafer line.
In plain terms = everyone is racing to build, but from groundbreaking to shipment still takes at least two years — these new lines cannot ease today's shortage.
Can technology breakthroughs change the picture?
In August 2025, Jiufengshan Lab and Yunnan Xinya used domestically built MOCVD equipment — the core tool for growing semiconductor thin films on substrates — to achieve 6-inch InP-based detector and laser epitaxy for the first time, with key metrics at international leading levels.
In crystal growth, Huaxin Crystal and Pioneer Micro have both adopted the VGF method — vertical gradient freeze, which yields higher quality and lower dislocation density than traditional liquid-encapsulated pulling.
Jiufengshan Lab and Sun Yat-sen University have demonstrated heterogeneous integration of InP lasers on silicon wafers, validating the feasibility of high-volume production.
This means → the technical path is proven, but fab construction cycles, one-to-two-year equipment lead times, and another one-to-two-year customer qualification window all stack up — the supply crunch will persist at least through 2028, keeping cost pressure on optical modules and AI compute infrastructure firmly in place.
Content is for reference only, not financial advice.