To The Moon

Everyone is talking about NVIDIA chips. Almost nobody is talking about the material that makes AI data center lasers work: 

Indium Phosphide (InP)

According to Reuters, US officials have reportedly made multiple trips to China trying to resolve export restrictions. So far, China is still deliberately slowing approvals and keeping a tight grip on supply.

Why does this matter?

Because every AI data center depends on optical networking. And optical networking depends on lasers. And those lasers depend on InP.

⚠️ The scary part: Switching suppliers isn't like changing a phone carrier. Qualification can take 12-18 months.

China knows this.

Even as Chinese manufacturers rapidly expand production capacity, overseas shipments are expected to remain limited, keeping the bottleneck firmly in place.

🏆 Potential Winners:

✅ $SIVERS SEMICONDUCTORS AB(SIVEF)$   — InP laser specialist. Longer shortages = stronger pricing power.

✅ $Lumentum(LITE)$   — Sources key substrates from Japan, largely avoiding China's export gate. One of the cleanest beneficiaries.

✅ Win Semi (3105.TW) — The world's only pure-play III-V foundry. Existing capacity becomes more valuable as supply tightens.

✅ $Aehr Test(AEHR)$  — Tests InP and silicon photonics wafers. When wafers become more expensive, quality control becomes even more critical.

✅ $Aixtron SE(AIIXY)$  — Supplies the equipment needed to build InP production capacity. More countries seeking self-sufficiency = more equipment demand.

✅ $Soitec, Bernin(SLOIF)$  — Photonics infrastructure play that benefits as optical components become more valuable.

❌ Potential Losers:

⚠️ $COHR — Reuters highlights reliance on China-linked InP substrate supply, creating potential risk.

⚠️ $AAOI — Demand is strong, but laser component shortages could slow growth.

⚠️ $AXTI — A fascinating high-risk, high-reward story. US-listed, but manufacturing sits inside China. Everything depends on export approvals.

🔥 What we're witnessing isn't just a semiconductor story.

It's a real-time geopolitical battle over AI infrastructure.

The companies controlling the physical bottlenecks may end up being the biggest winners as the world fragments into competing technology blocs.

The next AI shortage may not be GPUs.

It may be the tiny material that makes the lasers work.