Tower and Scintil Announce Availability of First Heterogeneously Integrated DWDM Lasers for AI Infrastructure

Why It Matters

The breakthrough lowers power per bit and boosts bandwidth density, accelerating the shift to optical CPO architectures that hyperscalers need to sustain AI model growth and improve ROI.

Key Takeaways

• First heterogeneously integrated DWDM lasers for AI data centers
• SHIP combines Tower silicon photonics with monolithic lasers
• Enables high‑volume, low‑power, high‑bandwidth CPO deployments
• Supports $200 bn AI networking market growth to 2030
• Provides supply continuity via Tower’s multi‑site foundry

Pulse Analysis

Heterogeneous integration of DWDM lasers marks a pivotal shift in silicon photonics, merging active laser material directly onto a silicon waveguide platform. Scintil’s SHIP process embeds monolithic lasers within Tower’s mature SiPho foundry line, delivering a single‑chip solution that meets the tight wavelength stability and channel spacing required for dense‑wavelength‑division‑multiplexing. This approach eliminates the need for discrete laser arrays, reduces packaging complexity, and opens a path to economies of scale previously unavailable to optical interconnect suppliers.

The AI data‑center market is racing toward multi‑petabit per second connectivity, and co‑packaged optics (CPO) is the preferred architecture to achieve that scale. By delivering higher bandwidth per fiber while cutting energy per bit, DWDM‑enabled CPO directly addresses the $200 billion AI networking spend projected for 2030. Hyperscalers benefit from lower tail latency and improved GPU utilization, translating into faster model training cycles and stronger return on investment. The LEAD Light module’s ability to operate at industry‑standard DWDM grids ensures compatibility with existing optical transport ecosystems, smoothing the transition for large‑scale deployments.

Beyond the technical merits, the collaboration secures a resilient supply chain through Tower’s global, multi‑site silicon photonics manufacturing footprint. This capacity is critical for meeting the volume demands of hyperscale operators, who require millions of units per month to keep pace with AI model proliferation. The validation of SHIP on production lines also reduces qualification risk, giving customers a clear path from prototype to mass production. As the industry converges on optical architectures, the availability of production‑ready, heterogeneously integrated DWDM lasers positions both companies as key enablers of the next generation of AI infrastructure, a trend that will dominate discussions at OFC 2026 and beyond.