Everyone Has Covered Photonics - These Are the Top 5 I Think Are Actually Worth Owning

Nvidia put $2 billion into Coherent in March. Then it put another $2 billion into Lumentum. Both deals came with multiyear purchase commitments for the lasers and optical components that keep AI data centers from choking on their own copper wiring, and both stocks did exactly what you’d expect an Nvidia-backed company to do. Lumentum is up more than 1,000% over the past year and joined the S&P 500 in March and the Nasdaq-100 in May. Coherent hit a 52-week high near $440, up over 100% year to date, with demand for its AI transceivers booked out through 2028. Ciena’s optical networking business grew revenue 40% year over year last quarter and the stock has traded above $480. Every one of those numbers has been written up a dozen times by now. Coherent, Lumentum, and Ciena are the photonics stocks a casual reader has already heard of, and at current multiples, all three are priced at premiums that I cant justify owning anymore. The problem with the obvious names once everyone’s covered them is that the upside is no longer there and its more likely you’ll end up as exit liquidity. This is why I’ve created this list here for you today, some of these names are popular but still have plenty of upside runway.

The story is true. Copper interconnects run out of runway as AI clusters scale, and every hyperscaler building a bigger cluster needs more optical links to move data between racks without melting the power budget. The optical transceiver market is on pace to jump roughly 60% in a single year, to around $26 billion in 2026. That growth doesn’t stop at the three tickers everyone already owns. Below are five companies with their own unique exposure to the same buildout, picked because each one solves a different piece of the problem and none of them get mentioned in the same breath as Coherent and Lumentum nearly as often as they will in the next few years.

Fabrinet

Fabrinet doesn’t sell you a laser or a transceiver but it does build the ones other companies sell. The Thailand-based contract manufacturer runs the precision optical assembly lines behind a huge share of the industry’s actual output, which means it collects a paycheck no matter which optics vendor wins a given hyperscaler’s business. Fiscal third-quarter revenue hit a record $1.214 billion, up 39% year over year, and the data center interconnect line grew 42% on rising shipments of 400ZR and 800ZR modules, the products carrying traffic between AI data centers today.

That picks-and-shovels position is the whole thesis behind this pick as photonics will increase in demand. Its also a double edged sword due to their client base. Fabrinet’s two largest customers, Nvidia and Cisco, made up roughly 46% of fiscal 2025 revenue between them. If either one brings manufacturing in-house or shifts volume to a rival contract manufacturer, the hit shows up fast. The stock has also re-rated hard, up more than 50% over the past three months, and trades in the mid-30s on forward earnings versus a mid-$13 consensus EPS estimate for fiscal 2026. You’re paying up for a business with added customer concentration risk. You’re also buying the one company that gets paid regardless of whether the AI optics race is eventually won by Coherent, Lumentum, Marvell, or somebody nobody’s heard of yet, because most of them route through Fabrinet’s floor to get built. I like the risk to reward on this set up.

MACOM

MACOM spreads its bets across three markets instead of one. The company sells analog and mixed-signal chips into data center optics, telecom, and defense and industrial customers at the same time. Data center revenue grew 36% year over year in the fiscal second quarter, industrial and defense grew 22%, and total revenue came in at $289 million, up 22.5% year over year and ahead of Wall Street’s estimate. Third-quarter guidance of $331 million to $339 million points to growth accelerating, not slowing down.

The company’s June rollout of hot via chip-scale packaging, a process that eliminates the wire bonds limiting how fast signals move through an optical module, gives MACOM a proprietary manufacturing edge that isn’t tied to any one customer’s roadmap. That diversification, in my opinion, is worth paying for, and the market has: MACOM trades near 50x forward EBITDA, a premium that assumes the data center growth rate holds even as telecom and defense grow more slowly. If AI-linked optical demand cools before those other two segments pick up the slack, that multiple has the opportunity to compress fast. MACOM earns its spot here over a narrower optical component maker because its AI datacenter growth stacks on top of two already-stable businesses making it one of my favorites in the sector.