Coherent DSP Vendors Are Down to Three. Here's What That Means.

Six coherent DSP houses competed in the optical market in 2018. By 2020 the field was five. By 2022 it was four. As of mid-2025, three remain. Two of those three are spoken about as "the leaders". The third is, depending on the quarter, "the safety net", "the disruptor", or "the also-ran" — never the same one twice.

That's not market dynamics. That's a supplier-concentration story, and almost nobody on the procurement side of the optical industry has caught up to what it means yet.

What "three vendors left" actually looks like

Three vendors does not mean three real options. Three vendors means:

• One vendor sets the roadmap. Every other vendor is reactive.
• One vendor is the safety net. Customers buy them to avoid sole-source exposure on the leader, and they get just enough volume to stay in business.
• One vendor exists primarily as a price-pressure chip. They win bids when the leader gets aggressive on commercial terms, and they lose them when the leader is willing to negotiate.

The leader's roadmap defines what 800ZR, 800ZR+, 1.6T pluggables, and the next-generation reach classes look like. The other two implement, with delay, to a spec they didn't draft. Customers experience this as "three options" on a spreadsheet. Operationally, it's one option, one second-source, and a negotiating chip.

How we got here

Coherent DSPs are expensive to develop and expensive to qualify. Each generation requires a new process node, an enormous tape-out cost, and a multi-year engineering team. The market for them is large in revenue terms but small in customer count — maybe forty serious buyers worldwide, of which the top six account for most of the volume.

That economic shape consolidates. The fixed costs of staying in the game eat smaller players first. Acquisitions retire competing roadmaps. The remaining players run a shared incentive structure: keep their existing customers, don't price-cut so aggressively that everyone goes underwater.

This isn't anti-competitive in any actionable sense. It's the equilibrium that follows from the cost structure. The point isn't whether it should have happened. The point is what it means for buyers now that it has.

The procurement consequences

Roadmap leakage. When there were six vendors, a buyer's roadmap conversation with vendor A was confidential because vendor B might win the next bid. With three vendors, the same three buyers see the same three roadmaps and have the same three conversations. The information asymmetry is gone. Vendors know what every major customer is planning, because the customer told all three of them.

Single-vendor-DSP risk on a 5-year platform. Pluggable optics are typically sourced from multiple module vendors. Inside those modules, the DSP often comes from a single house. A switch platform that locks in a 5-year hardware refresh cycle is implicitly betting that the DSP house behind that module's choice will still be on roadmap for those five years. Two of the three are clearly bet-able. The third is a coin flip on any given product line.

Price floors that don't move. Three rational competitors don't drive prices down the way six did. Buyers used to volume-driven price reductions every two years are seeing prices stabilise. The savings now come from architectural moves (LPO, CPO, embedded vs pluggable), not from vendor competition.

Roadmap timing decisions get harder. When one vendor sets the schedule, the timing of your refresh cycle starts being a function of their slip risk. If the leader misses an 800ZR+ ratification milestone by six months, every buyer waiting on the second-source has to decide whether to wait another six months or take the first-mover risk.

What buyers can do that actually works

Three things, in order of how often they get done well.

Structure RFPs around three vendors that pretend to be five. Get every player to respond to the same technical spec, even when you know two of them are not real contenders. The price signals are useful. The roadmap signals are useful. The negotiation position is useful even if the bid award is preordained.

Lock in volume commitments separately from roadmap commitments. Vendors will give you better prices on a longer commitment. Don't accept their roadmap as part of the deal. Lock in the volume, leave the technology choice open at each refresh.

Maintain real second-source engineering teams. Most buyers say they qualify multiple DSP vendors. Few actually do, because the qualification cost is real and dual-running adds operational complexity. The buyers who maintain this discipline pay a small overhead and gain a structural negotiating position.

The buyers who don't do these things end up with line cards from one vendor, a captive optical roadmap, and a procurement team that doesn't know they've lost their position until renewal time.

The strategic question for the next two years

The interesting question isn't whether the field shrinks to two. It might. The interesting question is what happens to coherent DSP pricing when the leader's volume is large enough that their cost structure decouples from the second-source.

When that happens, the second-source either consolidates into a niche reach class (long-haul, submarine, ultra-low-power) or gets acquired by someone outside the optical industry who wants the silicon for a different application. Either way, the "three vendors" math collapses to two real and one captive.

That's the scenario the procurement side should be modeling right now. Most of them aren't.