Credo pays up to $1.3B for DustPhotonics as photonics moves from R&D to deployment

Credo Semiconductor announced the acquisition of Israeli chip startup DustPhotonics for up to $1.3 billion, marking the semiconductor industry's most explicit bet yet that co-packaged optics has crossed from lab to production timeline. The deal structure includes immediate cash and equity plus earnouts tied to manufacturing milestones, a payment framework that converts DustPhotonics' photonics IP into Credo's product roadmap over the next eighteen months.

DustPhotonics specializes in silicon photonics transceivers designed for co-packaging with switch ASICs, the architecture hyperscalers need when 800G and 1.6T port speeds make traditional pluggable optics a thermal and latency liability. Credo has been shipping SerDes and DSP chiplets for data-center interconnect since 2022, but lacked the photonics layer required to compete with Broadcom and Marvell in fully integrated optical engines. DustPhotonics brings both design IP and early manufacturing partnerships with TSMC's advanced packaging lines, collapsing Credo's time-to-market by at least two years.

The $1.3 billion figure places DustPhotonics among the top five private semiconductor acquisitions in Israel, but more importantly signals that photonics is no longer a hedge. Hyperscale deployment of 51.2T switches begins in late 2025, and those systems assume co-packaged optics availability at volume. Credo's deal timing suggests customer pull rather than speculative positioning. The earnout structure ties roughly $400 million to production milestones in 2026, meaning Credo expects revenue contribution within two product cycles, not the five-year horizon typical of photonics R&D.

For allocators, the relevant shift is cost structure. Co-packaged optics eliminate the pluggable transceiver margin stack, collapsing a $15,000 per-port system cost to under $8,000 at scale. That math changes data-center capex models and threatens the module vendors who have owned optical interconnect economics since 10G Ethernet. Credo's move forces Broadcom, Marvell, and Nvidia to either acquire their own photonics capabilities or accept bill-of-materials disadvantage in 2026 switch generations. The Israeli photonics startup ecosystem holds three other credible co-packaging players, all now acquisition targets.

Watch for TSMC's Integrated Fan-Out packaging roadmap updates in Q2 2025, which will clarify whether co-packaged optics can hit hyperscale yield requirements at 3nm and below. Monitor Broadcom's February earnings call for any acceleration of its internal photonics programs or acquisition commentary. Credo's customer design-win announcements in the next six months will confirm whether this acquisition came early enough to compete for 2026 socket share at Meta, Microsoft, and Google.

The $1.3 billion price is not a photonics premium. It is the cost of staying in the data-center interconnect business after copper runs out of physics.