We’re seeing third-order effects ripple through the market right now — with $Lumentum(LITE)$ , $Applied Optoelectronics(AAOI)$ , and the broader photonics names getting hammered 10%+.
Part of this is the sector-wide selloff triggered by $SanDisk Corp.(SNDK)$ Sandisk’s -10% move and weakness across semis. But my angle is that the real story is upstream: the Japanese supply chain is likely running out of InP inventory/capacity because of Japan’s new export controls — and InP is critical for the Western optical buildout at scale.
I wouldn’t lump COHR in with the other photonics names, but even they’re already stretched on capacity, and this only gets worse with Sumitomo and JX now affected. In turn, this bottlenecks $LITE and others that rely on Japanese-sourced InP to build optics for things like $GOOGL’s TPU Ironwood clusters.
There are offsets that could spark a rally:
→ downstream pricing power (like HBM) could absorb supply shortages
→ Japan relaxing the InP export controls (unlikely)
→ $AXTI delivering substrates at scale once their $100M capacity expansion comes online
Long term, I’m still extremely bullish on the space. But near-term? No clue if the risk-off continues or whether photonics snaps back.
Export controls are the biggest wildcard. With Japan’s two largest competitors effectively halted, $AXTI — and indirectly, China — now control the global flow of InP substrates.
And I think that’s where many analysts are missing the plot: this isn’t about output volume, it’s about pricing power on a finite, critical resource.
It becomes a game-theory problem like HBM — but instead of several players, you basically have one.
InP used to be a few-hundred-million-dollar niche telecom market. Now hyperscalers like $GOOGL would pay billions just to keep Ironwood on track. That pushes TAM from a few hundred million to multiple billions — even $10B+ — once the bidding wars start. And you still have a ~$1.2B market cap sitting on two monopolistic chokepoints.
If pricing went up 3000%, it’d still only be a small percentage of an ASIC cluster’s BOM. You don’t need 3000% more supply under export controls — you just raise prices significantly and aggressively.
Comments