This article takes a look at how the current surge in AI compute is speeding up the shift from electronic to optical interconnects in data centers. Photonics brings much higher bandwidth and dramatically lower energy use per bit.
That means hyperscalers can cluster accelerators more tightly. They can scale systems that would otherwise hit power and heat limits in traditional electronics.
We’re digging into what this all means for supply chain players—especially AXT, Lumentum, and Lightwave Logic. Their strategic moves could shape the future of AI infrastructure, and it’s worth paying attention.
Optical Interconnects: The AI Compute Advantage
The AI compute boom is quietly turning the tables from copper wires to light. Photonics enables a new wave of high-speed transceivers and silicon photonics integration.
Data centers get more throughput and use less energy per bit. That’s absolutely crucial for scaling clusters of AI accelerators without blowing out the power budget.
As hyperscalers chase ever-denser performance, optical interconnects step in as foundational infrastructure. There’s now a whole ecosystem of materials, wafers, and device makers racing to meet a growing backlog with new capacity and fresh partnerships.
AXT: Indium-Phosphide Wafers and Backlog
AXT (AXTI) supplies indium-phosphide (InP) wafers for lasers and detectors in high-speed transceivers. The company’s story right now centers on a record $60 million InP wafer backlog and plans to double capacity by the end of 2026.
That really highlights how crucial InP substrates are for photonic-enabled AI networks. Even though 2025 revenue is dipping to around $88.3 million, investors seem drawn to the backlog and expansion plans as signals of strong demand ahead.
AXT does face geopolitical and export-permit risks, especially from China. That could mess with supply continuity.
Still, management’s Q1 2026 guidance and the multi-quarter backlog suggest sequential growth could keep rolling into 2026 and beyond. The demand from optical interconnects and the high-tech laser/detector market looks robust.
The company’s path is really tied to a technology roadmap that links wafer supply to faster, more energy-efficient transceivers for AI data flows.
Lumentum: Lasers, Transceivers, and Nvidia Alliance
Lumentum (LITE) designs and builds lasers, transceivers, and optical switches. Fiscal Q2 revenue hit $665.5 million, up 65.5% year-over-year, with strong margins to boot.
The components segment jumped 68.3% to $443.7 million, showing solid demand across the photonics stack. A major driver: a $2 billion strategic partnership with Nvidia that really brings together AI compute and optical interconnect supply chains.
Backlog is over $400 million, and management says operations are basically sold out through 2027. They’re targeting a $1.25 billion quarterly run rate within 9–12 months and $2 billion inside two years.
That’s a bold pace, and it matches the rush of AI workloads. The broader optical interconnect market could hit about $90 billion by 2030, but Lumentum trades at a premium and stays sensitive to hyperscaler capital spending cycles.
Lightwave Logic: Perkinamine Polymers for Silicon Photonics
Lightwave Logic (LWLG) is working on Perkinamine electro-optic polymers for modulators that pair with silicon photonics, aiming for 200–400 Gb/s per lane. The stock has absolutely soared, up about 1,304% over the past year.
That reflects investor excitement for materials that could unlock much higher-bandwidth optical links. But Lightwave Logic is basically pre-revenue, reporting just $237,000 in 2025 revenue and a net loss of $20.3 million.
The company does have four Stage 3 design-win customers, a partnership with Tower Semiconductor, and about $70 million cash runway through late 2027 after a $35 million offering. The big risk is all about execution and timing—can these polymers actually scale into real, bankable products fast enough?
What this means for the future of AI infrastructure
Execution timing and commercialization still matter a lot. Right now, backlogs, fresh partnerships, and capacity expansions point to real, lasting winners in optical-AI—this doesn’t feel like just another short-lived cycle.
Data centers keep chasing higher bandwidth and lower energy use. Photonics is quickly becoming the foundational enabler for scalable AI infrastructure, or at least that’s how people are starting to see it.
Investors and technologists should keep an eye on export-control dynamics and capacity milestones. Watching how quickly backlog turns into revenue is also wise.
The optical interconnect ecosystem—driven by InP wafer supply, laser and detector performance, and polymer-based modulators—looks ready to support the AI era for a good while. Of course, individual stocks will still swing with hyperscaler capex cycles, so it’s not all smooth sailing.
- Backlogs signal durable demand for optical interconnects as AI workloads grow.
- Strategic partnerships (e.g., Nvidia, Tower Semiconductor) help de-risk execution risk.
- Market potential could reach up to $90 billion by 2030 for optical interconnects.
- Risks remain from export permits, geopolitics, and the timing of commercialization.
Here is the source article for this story: The Optical AI Boom’s Biggest Winners: 3 Under-the-Radar Stocks Fueling the Revolution