[OFC 2026] Part 1 of 5: 300mm SiPh Foundry: Who Is Actually Ready?

Key Takeaways

• Samsung shows integration, limited performance metrics
• GlobalFoundries offers modular platform, ecosystem immature
• NVIDIA targets AI links, lacks clear foundry roadmap
• System integration now primary competitive factor
• Papers expose gap between announcements and reality

Summary

The OFC 2026 conference showcased seven papers from Samsung, GlobalFoundries and NVIDIA that dissect the state of 300 mm silicon‑photonic (SiPh) foundries. While the headline race to 200 Gbit/s per wavelength still dominates headlines, the papers reveal that success now hinges on system‑level integration and ecosystem maturity rather than pure device performance. Samsung highlights a tightly integrated platform, GlobalFoundries stresses modularity, and NVIDIA targets AI‑driven photonic links, each exposing gaps between press releases and technical reality. The analysis underscores a shifting competitive landscape where completeness of the photonic stack is the true differentiator.

Pulse Analysis

Silicon photonics has moved from a niche research topic to a cornerstone of next‑generation data‑center architecture. The transition to 300 mm wafer production promises lower costs and higher volumes, but achieving 200 Gbit/s per wavelength requires more than fast modulators—it demands a complete, co‑engineered stack of drivers, waveguides, packaging and software. The OFC 2026 papers illustrate how industry leaders are aligning their roadmaps with these broader system requirements, signaling a maturation of the SiPh supply chain that could unlock widespread adoption.

A closer look at the three contributors reveals divergent strategies. Samsung’s work emphasizes a tightly integrated platform, bundling lasers, modulators and drivers in a single process flow, yet it provides scant data on real‑world loss budgets. GlobalFoundries adopts a modular approach, offering interchangeable building blocks, but its ecosystem—design kits, testing infrastructure, and foundry capacity—remains under‑developed. NVIDIA’s papers focus on AI‑centric photonic links, showcasing high‑speed transceivers but offering little insight into long‑term manufacturing scalability. Collectively, the papers suggest that while device performance continues to improve, the decisive factor is how seamlessly these components can be assembled into production‑ready systems.

The market implications are profound. Operators seeking to upgrade to terabit‑scale interconnects will prioritize vendors that can deliver a turnkey SiPh solution, reducing integration risk and time‑to‑market. Investors and OEMs should watch for signs of ecosystem maturity—standardized design libraries, reliable testing services, and robust supply chains—rather than headline‑grabbing bandwidth numbers alone. As the industry converges on system integration as the competitive edge, the firms that can close the gap between research papers and commercial readiness will shape the next wave of data‑center growth.