The optical stocks rally isn't about a US ban - it's about the InP chokepoint the market finally found

Optical component stocks are flying today.

AXTI is up 13.6% to $68.7 on $1.17B of turnover. Coherent +9.6%. Lumentum +9.2%.

Headlines are calling it a rally on a "proposed US ban on Chinese tech."

That's not the story.

The story is that the market finally found the bottleneck behind the bottleneck in the photonics supply chain - and AXT's Q2 earnings confirmed it beyond doubt.

Then the question became whether the stock still has asymmetric upside after going from $2 to $68 in a year.

So here's the map:

Why photonics replaced copper - and why nobody can design around it

AI clusters are too large to fit on one chip. GPUs need to talk to each other at insane bandwidths. Copper wiring works for short distances, but hits what the industry calls the "Copper Wall" - signal loss, heat, crosstalk, and energy consumption that become unfeasible as data rates climb.

Photonics takes over. Data moves as light through fiber. Optical transceivers convert electrical signals to optical and back.

But silicon - the material that runs everything else in semiconductors - has a fatal flaw for this application. It's an indirect bandgap semiconductor. It cannot lase. Silicon photonics can guide light and modulate it, but it still needs an external light source.

That light source is built on indium phosphide (InP).

InP has a direct bandgap that's efficient at emitting light at 1,300–1,550 nanometers - the exact wavelengths used in fiber-optic communications. It operates at speeds exceeding 100 gigabits per second per channel. There is no commercially available drop-in replacement.

The chokepoint: who controls the substrate

Silicon wafers are everywhere. InP substrates are not.

Global production is dominated by two names: AXT and Sumitomo Electric.

AXT manufactures most of its InP substrates in China. Sumitomo operates in Japan.

Here's where the map gets interesting:

China produces 70% of the world's indium. In February 2025, China introduced export licensing requirements on InP compounds. Not an outright ban - a licensing mechanism that creates administrative friction, slows qualification timelines, and generates procurement anxiety.

Paul Triolo at Albright Stonebridge called it a "granular materials chokepoint toolkit". Beijing doesn't need to block finished photonics products outright. It can throttle the upstream compound and let the downstream ecosystem starve.

The average price for a 6-inch InP wafer surged 250% to $5,000.

AXT: the company that was gatekeeping itself

AXT sat in the strangest possible position. They make the material the West needs most for AI photonics - but their fab is in China, so they need China's permission to ship it to US customers.

They got their first InP export permits in late June 2025. Before that, InP revenue was $3.6M in Q2 2025. Q3 was $13.1M on the first permits. Q1 2026 was $13.6M.

Then Q2 2026 hit and the dam broke:

Total revenue: $47.59M - up 165% year-over-year, 77% quarter-over-quarter.

Management had guided for $34M contingent on permits. They blew past it.

Gross margin: 44.9%. That is up from 8% a year ago and 29.6% in Q1. The kind of inflection that only happens when a company moves from constrained production to pricing power.

Net income: $11.13M. Up from a $7M loss a year ago.

CEO Morris Young called it "a step-function increase."

The capital raise that matters

In May, AXT closed a $632.5M equity raise. Cash and investments as of June 30: $745.8M. Shareholder equity tripled to $912.1M.

The plan: double InP capacity in 2026, double it again in 2027. Move to larger wafer sizes. Expand customer base to tier-1 laser manufacturers and optical transceiver makers.

They're also pursuing a listing of their Tongmei subsidiary on China's STAR Market and working on long-term supply agreements.

$745M in cash. That is not a company worried about surviving. That is a company positioned to scale.

Now the part that makes this harder than it looks

$AXTI is up 320% year-to-date. 43% in the last five days. 147% over the last four months.

Market cap: $4.5B. P/B: 13.3x. Forward P/E: negative (the trailing numbers haven't caught up to the Q2 inflection yet).

The stock went from roughly $100M market cap in early 2025 to $4.5B in eight months.

Structure is real. Price is the problem.

The chokepoint thesis is validated. InP is architecturally indispensable. AXT is one of two global substrate producers. China controls 70% of indium and has export licensing as a lever. Q2 confirmed the revenue and margin explosion when permits flow.

But at $4.5B and 13.3x book, the market has discovered this node. The discovery premium is largely baked in.

Compare that to Coherent at $95B market cap and Lumentum at $45B - those are the tier-1 customers sitting downstream, already priced as established names. AXT rerated from near-zero to a respectable small-cap because the thesis was confirmed, not because anyone priced the substrate layer last quarter.

The question now is whether capacity execution justifies a further rerate, or whether this is the part of the rotation where institutions lock in and the stock consolidates.

The risks that don't go away with a good quarter

Three things to keep in mind:

1. The fab is still in China. China eased InP controls in June 2026, which likely contributed to Q2 beating guidance. But easing is not removal. Permits can tighten again. The export permit risk is structural, not a one-time hurdle.

2. The $632M raise was dilutive. The company went from ~23M shares pre-raise to roughly 76M shares now. That dilution is already reflected in the market cap, but it means early holders got paid and new money came in at higher prices.

3. China's own AI build is a parallel demand source.AXT's CEO noted that revenue from China's InP-based laser market more than doubled in Q1 and is expected to double again. No permit is required for domestic China shipments. That's a bright spot - it creates a demand floor even if export permits stall. But it also means geopolitical risk cuts both ways.

Where Coherent and Lumentum fit

Coherent ($288, $95B MC) and Lumentum ($780, $45B MC) are the customers, not the substrate layer.

Nvidia invested $2B in each of them in March. Both are building their own InP wafer capacity - Coherent is doubling at its Texas plant this year and plans to more than double again by end of 2027.

Lumentum is sold out through 2028 despite quadrupling output.

They're good companies riding real demand. But they're also $45-95B companies where the photonics upside is diluted by legacy telecom and industrial segments. The rerate has already happened for them - Lumentum is up 111% YTD, Coherent up 56%.

The pure-play substrate exposure at a smaller cap was AXT. And the market found it.

TLDR:

The headline about a "US ban" is noise. The real catalyst is the InP substrate chokepoint - architecturally indispensable, concentrated in two producers, with China controlling 70% of the raw material and using export licensing as a throttle.

AXT's Q2 confirmed the thesis: $47.6M revenue, 44.9% gross margin, record InP sales, $745M in cash to fund capacity doubling.

But at $4.5B market cap and 13.3x P/B, the discovery is done. The chokepoint is real. The price reflects that.

The setup now requires flawless execution on capacity ramp and sustained permit flow to justify a further rerate. That's a different kind of bet than the structural discovery that got the stock from $2 to $68.

I personally think the thesis is validated, but the asymmetry has compressed. This is no longer the speculative bottleneck you find before institutions arrive. It's the bottleneck everyone now knows about.

Structure first. Price second. In this case, both have been answered - and the answer to the second one makes me cautious on chasing.