The Last Three Coherent DSP Vendors and What That Costs You
Six coherent DSP houses competed in 2018. Three remain. That's not really three options. It's one option, one second-source, and a negotiating chip. Procurement teams who treat it like a real three-way market end up paying for the illusion.
This post is about what that costs you, in actual euros and actual roadmap-flexibility-units, and what you can do that matters.
The shape of a three-vendor market
In any market with three rational competitors and high fixed development costs, you don't get three peer competitors. You get a leader, a fast-follower, and a third player whose role rotates depending on who's hungrier that quarter.
The leader sets the technical roadmap. New form factors, new modulation classes, new FEC variants — they all originate from whoever has the largest installed base and the most fab capacity. The other two implement, with delay, to a spec they didn't draft.
The leader's commercial pricing sets a floor. The fast-follower undercuts by 10-15% on volume bids. The third player undercuts the fast-follower by 5-10% on bids where they need the win to keep the lights on.
That looks like price competition. It mostly isn't. The three of them have implicit incentives to keep the floor where it is. None of them benefits from a price war that drops the floor by 30 percent and forces them all to operate near margin.
What that costs the buyer
Roadmap leakage. When there were six vendors, you negotiated roadmap with one of them and the others didn't know what you were planning. With three, you have to talk to all three because the second-source is real. The information you share with vendor B is, in practice, information vendor A's account team will hear within a quarter. Negotiation power that depended on information asymmetry no longer exists.
Floor pricing. A modest price drop per year, indexed roughly to silicon-node maturity. Real competition would deliver more. The historical pattern of 20-30% per generation broke around 2022. Since then it's been single-digit percentages, and the gains have all come from architecture (LPO, embedded) rather than DSP competition.
Slip-risk concentration. If the leader misses an 800ZR+ ratification or a 1.6T pluggable launch by six months, every customer waiting on the second-source has to wait too. The fast-follower can't ship without the spec being stable. Your refresh cycle is hostage to one vendor's quarterly results.
Operational dependency on private MIBs. Modules from each of the three vendors expose diagnostics differently above the CMIS floor. If you've built monitoring tooling around one vendor's private MIBs, switching DSP vendors at module level means rebuilding the monitoring layer. That's a real cost, and it locks you into the DSP choice you made three years ago.
Three things you can do that actually work
Structure RFPs around three vendors, not two. Get every player to respond to the same technical spec, even if you know the bid will go to the leader. The price signals are useful even when the bid award isn't. The roadmap signals — when does each vendor commit to specific milestones — are intelligence that informs your refresh planning.
Decouple volume commitments from technology commitments. Vendors offer the best prices on combined volume-plus-roadmap commitments. Don't accept the bundle. Lock in the volume separately, leave the technology choice open at each refresh cycle. The volume gets you the price. The flexibility gets you the negotiating position for the next round.
Maintain a real second-source qualification. Most buyers say they qualify multiple DSP vendors. Few actually do, because the cost of running two qualification programs in parallel is real and the operational complexity of dual-running adds friction. The buyers who maintain this discipline pay a small overhead and gain a structural negotiating position. The ones who don't, find out at renewal time that they didn't have a position.
What good vendors will tell you (and what they won't)
A vendor who's confident in their roadmap will tell you their next two generations, with rough timing, on a call. The ones who hedge ("we're flexible on roadmap to accommodate customer needs") are the ones whose roadmap is reactive to the leader.
A vendor who's confident in their second-source position will tell you which of the three they expect to compete with them on which class of products. The ones who can't articulate that don't have a strategic position — they're just trying to win bids.
A vendor whose pricing is structurally close to their cost will tell you their volume-tier breakpoints clearly. The ones whose pricing is "wherever the leader's pricing is, minus 10 percent" are the second-source role, and they'll tell you that too if you ask in the right way.
The next two years
The interesting question for 2026-2028 isn't whether the field consolidates to two. It might. The interesting question is whether the leader's volume advantage decouples their cost structure from the other two, at which point the fast-follower either consolidates into a niche reach class (long-haul, submarine) or gets acquired.
If you're planning a 5-year hardware refresh that lands in 2027 or 2028, your assumption needs to be either "two real vendors" or "one real vendor and one captive". Build the procurement strategy for either, and adjust as the market actually unfolds.
The three-vendor framing is convenient and increasingly wrong. The procurement teams who notice early will be the ones who maintain the most pricing power.