Markets & Verticals

Segments (FY25; shifting to Datacenter & Communications + Industrial from FY26)

1. Networking - Datacom transceivers (800G, 1.6T, pluggables) - Telecom: coherent transmission, amplifiers, ROADMs, pump lasers - Advanced optics (filters, crystals, assemblies)

2. Materials - Engineered materials (ZnSe, ZnS, SiC substrates, CVD diamond) - VCSELs, EELs, photodiodes, driver ICs - Thermoelectrics, advanced ceramics, metal-matrix composites - Laser optics, IR thermal imaging

3. Lasers - Excimer lasers (UV, display annealing) - Solid-state ultrafast/nanosecond lasers - CO₂ lasers, fiber lasers - Laser systems & subsystems

Key Competitors by Overlap

Note: no single competitor spans Coherent's full stack — from raw materials to finished laser systems. Lumentum and MKS are the closest breadth comps.

Let me break it down in plain terms.

What Coherent's products actually do:

The unifying theme: Coherent makes the stuff that generates, shapes, moves, and detects light — across virtually every wavelength — for anything from a submarine cable to a factory robot to a satellite.

800G, 1.6T

These are speeds — how much data one transceiver pushes per second.

• 800G = 800 gigabits per second (about 500 HD movies streamed simultaneously)
• 1.6T = 1.6 terabits per second (double that). Next gen: 3.2T, 6.4T in the pipeline.

These numbers track the upgrade cycle inside AI datacenters — GPUs are getting faster, so the pipes between them must get fatter.

Pluggable

A pluggable transceiver is exactly what it sounds like: a hot-swappable module you push into a port on a switch or server. Think of it like a USB stick — but for fiber optic cables. You don't solder anything; you just plug it in.

The form factors have names like QSFP-DD and OSFP (different shapes/sizes of the metal cage). "Pluggable" distinguishes them from older approaches where optics were permanently soldered onto the circuit board. The point is flexibility — upgrade the speed by swapping the module, not the entire switch.

Why it matters for Coherent: AI clusters are racing up the speed curve (800G → 1.6T → 3.2T). Each jump is a new product cycle and a pricing lift. Coherent is one of the few that makes the laser chips inside the transceiver too — not just final assembly — so it captures more margin per unit.

No — data moves through fiber optic cables. The laser light travels inside the glass fiber, not through air.

How it works:

Electrical side → The switch/server sends an electrical signal to the transceiver.

Conversion → The transceiver's laser chip converts that electrical signal into light pulses (on/off at insane speed). A driver IC modulates it — think of it like Morse code at a trillion pulses per second.

Transmission → The light shoots down a hair-thin glass fiber. Bounces along the walls of the fiber (total internal reflection), traveling kilometers with almost no loss.

Receiving end → Another transceiver catches the light with a photodetector and converts it back into electrical signals.

So the path is:

Switch → electrical → laser → light through fiber glass → photodetector → electrical → another switch

The laser is just the engine. The fiber is the road. Copper cables exist too but max out at short distances — fiber wins for anything beyond a few meters in a datacenter because light is faster and loses less signal.

Same principle. Different scale.

Home broadband is a bicycle. Datacenter interconnects are a Formula 1 car. Both have wheels and move people, but the engineering underneath is wildly different.

Coherent sits on the Formula 1 end — their transceivers use "coherent optics" (hence the name), encoding data in the light wave's phase and polarization, not just simple on/off. That's how you squeeze 800 billion bits per second through a strand of glass thinner than a hair.

You're right — that wasn't in my summary. These are not products; they are competing technology architectures inside the transceiver. Here's the clean version:

Two ways to build the laser engine inside a transceiver:

The pattern you're describing:

Every speed jump (100G → 400G → 800G → 1.6T), the same cycle repeats:

1. New speed arrives → only EML/InP can hit the spec → margins are great
2. SiPho catches up (barely good enough for shorter links) → hyperscalers switch for cost
3. Margins compress on SiPho → those with InP scale move up to the next speed

This is the structural tension in Coherent's Datacom business: they lead with InP performance, but face constant SiPho "good enough" pressure from competitors (and even within their own portfolio). Their bet is that AI workloads (high bandwidth, need for signal integrity) will push demand toward InP longer than in prior cycles.

1. DATACOM & COMMUNICATIONS (FY26 realigned; was Networking)

2. INDUSTRIAL (FY26 realigned; combines Materials + Lasers)

Bottom line: Coherent spans the full stack — raw exotic materials → laser chips → finished modules → complete turnkey systems. No competitor does all of it.

Here's the combined picture — margins, position, and competitive dynamics across the portfolio:

Segment-Level Profitability (FY25, old segmentation)

Segment profit = after direct R&D, sales, marketing, admin; before corporate overhead, SBC, amortization. FY25 had 1 month left when filed.

Company-level (Q3 FY26): GAAP gross margin 37.7%, non-GAAP gross 39.6%, non-GAAP operating margin ~20%.

Product-Level Competitive Landscape

DATACOM TRANSCEIVERS

Key dynamic: Coherent is a top-3 transceiver supplier but is a generalist (module + component + materials) competing against specialists (Innolight, Eoptolink). Specialists have higher net margins (20–33%). Coherent's edge: own the laser inside, not just final assembly.

TELECOM COMPONENTS

SiC SUBSTRATES

Key dynamic: Chinese competitors doubled share in 2 years. Coherent's response — shift to differentiated bundles (optics + power), but pure SiC is a tough market structurally.

INDUSTRIAL LASERS

MATERIALS & OPTICS

Summary Matrix

The headline trade: Coherent's deep-tech moats (excimer, materials, A&D) are high-margin and stable. Its growth engine (datacom transceivers) is fast but lower-margin relative to specialists. The bull case is vertical integration wins over time; the bear case is "good at many things, best at none."

Good question. The "fiber optic cable" in a datacenter isn't really one product from Coherent — they play in three layers:

Where Coherent sits inside the datacenter fiber link:

The simple picture:

Server --- [Coherent transceiver] ==== fiber ===== [Coherent transceiver] --- Switch
                  ↑                                              ↑
          Laser chip inside                              Laser chip inside
          (Coherent VCSEL/EEL/SiPh)                      (Coherent VCSEL/EEL/SiPh)

Coherent owns the endpoints (the chips + the plug-in modules), not the glass in between. The cable itself is commodity — the value is in generating and detecting light at each end.

Laser — a device that generates a focused, coherent beam of light. Single-function: emits photons.

Transceiver — a module that both transmits and receives data. Contains a laser (to send), a photodetector (to receive), plus electronics/DSP to encode and decode signals.

Simple analogy: Laser = lightbulb. Transceiver = full walkie-talkie (speaker + mic + radio).

In Coherent's world: - They sell lasers to others who build them into systems (cap equipment, medical devices, manufacturing tools). - They sell transceivers directly to datacenter and telecom customers who plug them into switches/routers for optical networking.

Yes. In a datacenter:

• Each transceiver has one laser (transmit side) + one photodetector (receive side).
• The laser converts electrical signals → light pulses. The photodetector converts incoming light → electrical signals.

Coherent's advantage: they make the laser chips themselves (InP, VCSELs) and build the finished transceiver around them. Vertical integration — they're not buying someone else's laser to put in their transceiver.