Credo Semiconductor Pays Up to $1.3B for DustPhotonics, Enters Optical Interconnect Market

Credo Semiconductor announced it will acquire DustPhotonics for up to $1.3 billion, entering the silicon photonics market as hyperscalers push optical interconnects deeper into data center architectures. The deal, Credo's largest, brings the San Jose-based company technology for integrating optical components directly onto silicon chips—eliminating bottlenecks where electrical signals convert to light.

The transaction includes $950 million in upfront consideration, split between cash and stock, with an additional $350 million in earnouts tied to revenue milestones through 2027. DustPhotonics, founded in 2017 and based in Yokneam, Israel, has raised approximately $90 million from investors including Walden International and Entrée Capital. The startup employs roughly 120 people, most in R&D, and holds 40 patents covering photonic integration and packaging techniques. Credo expects the acquisition to close in Q2 2025, subject to Israeli regulatory clearance and standard conditions.

The move addresses a specific constraint in AI training clusters and high-performance computing: co-packaged optics, where lasers and modulators sit millimeters from the switch ASIC instead of plugging in externally. Broadcom and Marvell have signaled investments in the same direction, but neither has shipped production silicon photonics at scale. DustPhotonics's edge is a monolithic process that fabricates both electronic and photonic components on the same wafer, avoiding the yield penalties of hybrid assembly. Meta publicly tested co-packaged optics prototypes in 2023; Google filed patents for similar architectures in early 2024. Both are Credo customers for its existing SerDes chips, which move data at 112 Gbps per lane. Optical integration collapses power consumption—DustPhotonics claims 40% lower watts per bit than pluggable modules—and shaves two nanoseconds of latency per hop, material at the cluster scale where thousands of GPUs synchronize gradients.

Credo's revenue mix shifts with this deal. The company generated $288 million in fiscal 2024, up 64% year-over-year, selling electrical retimers and active cables to hyperscalers and networking OEMs. Gross margin sat at 68% in the most recent quarter, elevated by fabless operating leverage. DustPhotonics adds optical IP but no revenue yet—its first customer shipments are scheduled for late 2025, aligned with the 800G Ethernet transition. Credo's CFO indicated on the earnings call that the earnout structure protects against integration risk: $100 million unlocks if DustPhotonics books $150 million in trailing revenue by December 2026, implying production ramps and design wins need to land within 18 months. The remaining $250 million ties to 2027 thresholds Credo did not disclose, likely keyed to margin profile or customer concentration.

Allocators should watch three things. First, whether Broadcom or Marvell pre-announce co-packaged optics roadmaps before Credo's deal closes, signaling competitive urgency. Second, Israeli regulatory timing—the deal requires Committee for Foreign Investment approval given DustPhotonics's defense-related research contracts, which could stretch close into Q3. Third, Credo's guidance for fiscal Q1 2025, due in early June, will clarify if the company plans to maintain 65%+ gross margins while absorbing DustPhotonics's earlier-stage cost structure. If margin guidance compresses below 60%, the market will reprice integration risk and question the earnout assumptions.

The deal prices DustPhotonics at roughly 14x forward revenue if the full earnout pays, steep for a pre-revenue asset but aligned with Nvidia's 2020 Mellanox acquisition at 13x sales when data center interconnect became strategic. Credo is betting the optical shift happens faster than the market expects, and that owning the full stack—electrical, optical, and packaging—matters more than fabless margin purity. The company's stock closed up 6.8% the day after announcement, suggesting investors agree. Whether that holds depends on customer commitments, likely visible in the next 90 days as hyperscalers finalize 2026 CapEx budgets.