The Indium Phosphide Bottleneck

IQE shares crashed 26% on Huawei fears this week. Huawei exposure is under 5%. The real story is the InP bottleneck every AI optical transceiver depends on.

A UK semiconductor company lost a quarter of its value in a single day this week. The trigger was Huawei ban escalation. The market assumed the worst. It was wrong about where the real exposure sits.

IQE plc is the worlds largest independent foundry for epitaxial wafers the critical intermediate step between raw semiconductor substrates and finished laser devices. It does not sell directly to Huawei. Its Huawei exposure accounts for less than five percent of revenue.

The panic was about the wrong thing. The actual story is where IQE sits in the optical interconnect supply chain for AI data centers.

Where Indium Phosphide Fits in the AI Stack

Every 1.6T optical transceiver shipping into an AI cluster today relies on lasers made from indium phosphide. InP is a compound semiconductor that generates and detects laser light for fiber-optic communication. Silicon cannot lase. InP does.

The supply chain looks like this. Raw indium is a byproduct of zinc mining and refined by a handful of companies globally. It is grown into InP boules and sliced into substrates by players like AXT Inc whose indium phosphide revenue hit thirteen point six million dollars last quarter up sharply. The substrates are sent to epiwafer foundries like IQE which grow the atomic layers that determine laser performance. The finished epiwafers go to laser manufacturers Lumentum and Coherent and increasingly MACOM. The lasers go into transceivers and the transceivers connect GPUs in clusters of a hundred thousand or more.

IQE sits at the third step of this six-step chain. It is the only independent large-scale epiwafer foundry with qualified indium phosphide capacity.

The Qualification Moat

The reason IQE matters is not just that it has the reactors. It is that the qualification process for a new epiwafer supplier takes twelve to twenty-four months. A laser manufacturer must test samples run device fabrication trials and complete reliability qualification before it can switch suppliers. That is a year or two of engineering work. Once qualified switching costs are enormous.

The pattern has historical precedent. As Marc Levinson documents in The Box, the shipping container created a modular interface that collapsed freight costs and reshaped global supply chains. The bottleneck that looks unique always has a structural antecedent.

The MACOM Signal

In April 2026 MACOM a major optical and RF component company invested eighty-one million pounds into IQE. The structure was thirty million in equity at nineteen point eight pence per share plus a fifteen million convertible note. MACOM received eleven point five percent ownership and two board seats. They also signed a long-term strategic supply agreement.

MACOM effectively said: we cannot secure indium phosphide epiwafer capacity without IQE. That is not speculation. It is what a strategic investment with board representation communicates.

Why Silicon Photonics Does Not Kill the Thesis

The most common objection to any indium phosphide investment is silicon photonics. The argument goes that integrated silicon photonics will replace indium phosphide lasers making the entire supply chain obsolete.

Industry analysts at Yole forecast the silicon photonics PIC market at approximately two point one billion dollars by 2030. Cignal AI projects co-packaged optics deployment in scale-up networks in late 2028 to early 2029. Monolithic lasers on silicon a scenario where indium phosphide is genuinely unnecessary is not expected before 2030.

For the current AI infrastructure buildout cycle through 2030 indium phosphide demand is structurally secured regardless of which photonic platform wins.

The Tension Inside the Story

None of this means IQE is a clean investment. The company is loss-making. Gross margins were under four percent last year on 118 million pounds in revenue because the reactors are running at roughly half capacity. The wireless segment which represents fifty-seven percent of revenue is in structural decline. Three CEOs have left in five years. The current chief executive also serves as chief financial officer a governance arrangement that signals distress.

The bull case is that AI photonics demand fills the idle capacity before the cash runs out. MACOMs eighty-one million pounds buys approximately two years of runway. The bear case is that the wireless decline outpaces the photonics ramp.

But the underlying structural thesis is clean: the indium phosphide bottleneck is real and getting tighter. AXT Inc the indium phosphide substrate supplier is doubling capacity again for 2027. Lumentum reported ninety percent revenue growth in its latest quarter. The optical interconnect buildout for AI clusters is not a narrative. It is a capital expenditure cycle backed by the largest technology companies in the world.

The market panicked about Huawei and ignored the structural story underneath. That is the pattern Law I predicts: value migrates upward as lower layers commoditise and the market is slow to update on where the new bottleneck sits.

Subscribe to The Durability Curve on Substack for weekly analysis of AI infrastructure through the lens of structural durability.

Full research reports on Gumroad including the complete AI photonics supply chain breakdown.