The AI Boom Has a Second Act—And It's Playing Out in Optics

Key Points

• Optical fiber, connectors, and routing systems—not chips—are emerging as a critical and underappreciated layer of AI infrastructure investment.
• Amphenol, Corning, and Ciena have each recently beaten earnings estimates and secured major hyperscaler contracts tied to data center expansion.
• Analyst Lucas Downey argues that earnings estimate revisions, not stock price momentum, are the key signal for identifying opportunities in optical networking stocks.

The AI build-out conversation tends to start and stop with chips. But the companies quietly landing multi-year, multi-billion-dollar contracts right now aren't making semiconductors—they're making the glass, connectors, and routing systems that hold the entire data center ecosystem together.

Lucas Downey of TradeSmith sees the AI infrastructure stack as five distinct layers: land and site development, power, cooling, compute, and memory, and finally, networking and connectivity.

That last layer—fiber, optics, and high-speed interconnects—is where he's focused, and where he believes analysts are still playing catch-up.

The Shift From Copper to Glass

The transition driving demand here is straightforward: data centers are moving away from copper wiring and toward optical links made of glass and fiber. Optical connections carry data at the speed of light, generate less heat than copper, and reduce cooling costs.

For hyperscalers running massive AI workloads, those efficiency gains compound quickly.

Downey estimates AI infrastructure capital expenditure is approaching $1 trillion globally. The optical networking segment alone is seeing supply bottlenecks—there aren't enough components to meet demand—which is exactly the kind of constraint that tends to reward established manufacturers with the capacity to scale.

Amphenol: Connectors at the Core of Every Rack

The first name Downey highlights is Amphenol Corporation NYSE: APH, one of the world's largest manufacturers of electronic connectors and interconnect systems. Amphenol isn't making the lasers or chips—it's making the physical connections between every component inside a data center: the racks, the servers, the assemblies. Every nanosecond of latency matters in high-speed AI workloads, and every connection counts.

The company delivered one of the biggest earnings beats of the recent season, with revenue of $7.62 billion, well above the Street's estimate of $7.08 billion. Earnings per share (EPS) also beat, and guidance came in above expectations. Downey notes that the magnitude of this latest beat is what's drawing renewed attention from analysts, who he says are still revising numbers higher.

Amphenol carries roughly a $200 billion market cap and has posted double-digit revenue and earnings growth projections extending out to 2028—making it, in Downey's view, a core position in a theme that isn't slowing down.

Corning: The Fiber Backbone of the AI Era

The most widely recognized name on the list is Corning Incorporated NYSE: GLW, a company that has been manufacturing specialty glass for over 170 years and is now at the center of the next generation of optical fiber infrastructure.

Corning recently secured two significant hyperscaler deals. The first is a partnership with NVIDIA Corporation NASDAQ: NVDA to expand U.S.-based optical connectivity manufacturing—Corning expects to grow its U.S. optical connectivity capacity 10x and U.S. fiber capacity by more than 50% under the agreement. The second is a multi-billion-dollar deal with Amazon.com NASDAQ: AMZN for U.S. fiber manufacturing, announced earlier this month.

At a J.P. Morgan conference, Corning's management projected its sales run rate reaching $20 billion by year-end, $30 billion by 2028, and $40 billion by 2030. On the earnings side, Downey points to EPS slated around $3.19 this year, accelerating to $4.21 next year and $5.75 in 2028. He expects those numbers to be revised higher as more hyperscalers follow NVIDIA and Amazon to the table.

The stock has already had a substantial run, and the analyst consensus has been slow to keep pace. Downey draws a direct parallel to Micron Technology NASDAQ: MU, where analysts persistently underestimated the demand cycle before eventually catching up with a wave of estimate upgrades. He sees the same dynamic unfolding in optical.

Ciena: The Traffic Director of the Optical Network

The third name is Ciena Corporation NYSE: CIEN, which Downey describes as the intelligent routing layer of the optical ecosystem. Where Amphenol handles physical connections, and Corning manufactures the fiber cables, Ciena optimizes how data travels through those cables—routing light-speed signals, managing bandwidth, and reducing congestion across distributed AI clusters with thousands of simultaneous data packets in motion.

Ciena's recent earnings reflected the same demand story: EPS came in at $1.64 against estimates of $1.46, revenue beat consensus, and the company raised its full-year guidance to $6.3 billion—above the prior estimate of $6.15 billion—while noting margin expansion.

The stock has pulled back after a roughly 500% one-year run, which Downey attributes to seasonal mechanics—Russell index reconstitution and quarter-end rebalancing—rather than any change in fundamentals. He sees the pullback as a buying opportunity, with full-year 2026 EPS estimates around $6.50, accelerating to $9.65 the following year and $14.28 in 2028.

Following the Earnings, Not the Headlines

The broader framework Downey applies across all three names is the same: track where earnings estimates are going, not where the stock price has already been. When a company is landing multi-year deals with the world's largest cloud providers, the official estimates tend to lag reality. Analyst upgrades follow contracts, and stock re-ratings follow upgrades.

Near-term volatility—whether from AI bubble headlines, sector rotations, or index rebalancing—doesn't change the underlying demand picture. The data center build-out has committed capital stretching to 2030 and beyond. For investors willing to look past the noise, the optical layer may be the quietest opportunity left in an AI trade that's anything but quiet.