These 5 Optical Chip Stocks Are Cashing in on AI Data Center Bandwidth

Coherent’s Datacenter & Communications segment delivered $1.362B revenue (+40.6% YoY), now 75% of total revenue, reinforcing accelerating AI infrastructure spending. Across the optical stack, POET guided >30,000 optical engines shipping in 2026 and expects high-volume 800G production in Q3 2026, while Fabrinet’s Q3 FY26 revenue rose 39.3% YoY and capex nearly doubled to $63.76M (+121% YoY), signaling demand filling capacity. IPG Photonics authorized a $100M buyback alongside Q1 FY26 revenue of $265.5M (+16.6% YoY), and nLIGHT posted a Q1 FY26 beat with revenue up 55.2% YoY to $80.18M and non-GAAP EPS of $0.22 (vs $0.08 expected), as optical bandwidth growth appears to be shifting toward photonics.

Analysis

This is a bottleneck story, not a generic AI story. The economic winner is the layer that converts bandwidth demand into qualified, high-margin hardware: incumbent optical suppliers and contract manufacturers with real capacity, not the lowest-quality narrative names. That favors COHR and FN because they already have customer validation and scale, while the smaller names are more exposed to qualification delays, supply-chain hiccups, and working-capital strain if the ramp is slower than advertised.

The next 1-3 months are about order conversion and capex commentary, not end-market enthusiasm. If hyperscaler spending stays strong, the biggest second-order effect is a pricing/mix tailwind for vendors that can ship 800G/1.6T reliably; if deployment pauses, the market will punish pre-revenue torque first and ask questions later. That makes POET and LASR the highest-beta expressions, but also the most vulnerable to a timing miss into 2H26.

Contrarian take: the market is assuming every name in the optical stack benefits equally, when in reality the gap between “design win” and durable profit is wide. The more interesting setup is quality versus story — COHR/FN can turn the bandwidth wave into cash flow now, while the speculative names need flawless execution and no dilution. The thesis breaks if datacenter growth or forward capex decelerates meaningfully next quarter, or if 800G/1.6T ramp dates slip into late 2026.

Key Decisions for Investors

• Long COHR on a 1-3 month horizon as the highest-conviction quality beneficiary; target relative outperformance vs. NVDA/XLK if datacenter revenue stays above a 30% growth run-rate. Trim if margin expansion stalls or management guides below current consensus.
• Long FN as the picks-and-shovels expression into the next hyperscaler capex print; use pullbacks to build. Risk/reward is attractive if capex remains elevated, but exit if next-quarter bookings or utilization commentary softens.
• Pair trade: long COHR / short POET to express quality vs. speculative torque. This works if qualification and production timing remain the gating factor; cover the short if POET proves volume shipment cadence ahead of Q3 2026.
• Small starter long POET only as a venture-style position, sized for binary execution risk. Treat it as a call option on the 2026 ramp; invalidate the idea if Malaysia production slips or customer concentration becomes a financing concern.
• Watch LASR, but do not size aggressively until defense backlog converts into sustained datacom revenue. The stock can work if the dual-use thesis persists, but it is the most likely to over-earn on headlines and under-deliver on P&L timing.