Lumilens exits stealth with $900M to solve AI's copper bottleneck

The transition from copper to optical interconnects is critical for scaling AI clusters beyond the 1.5-meter distance limitation of electrical signals. Lumilens’ focus on near-packaged optics (NPO) provides an immediate bridge for hyperscalers who require higher density now but aren't yet ready for full GPU-integrated co-packaging. For the streaming industry, this infrastructure shift directly impacts the cost and speed of training the generative AI models used for localized dubbing, content discovery, and real-time video encoding at the edge. Watch for Lumilens' first commercial CPO deployments in late 2026 as a signal for the next phase of high-density rack design.

The emergence of Lumilens comes as the co-packaged optics (CPO) and near-packaged optics (NPO) market is projected to grow from roughly $100 million in 2025 to over $39 billion by 2030, according to TrendForce (June 2026). This growth is driven by the physical limitations of traditional pluggable modules, which face increasing power consumption and thermal challenges at speeds of 400G and beyond. Major Cloud Service Providers (CSPs) have begun prioritizing NPO as a mid-term transition solution because it offers shorter electrical paths than pluggable optics while maintaining the modularity needed for multi-vendor sourcing.

Competitive activity in the optical interconnect space has intensified throughout 2026. Per OpenPR (August 2026), NVIDIA recently invested $2 billion in Marvell to secure high-speed networking and AI connectivity, while startup Ayar Labs closed a $500 million Series E in March 2026 to scale its own optical I/O chiplets. These investments reflect a strategic shift where interconnect technology is now viewed as equal in importance to compute silicon. Hyperscalers like Microsoft and Meta have increasingly backed open standards for NPO to ensure supply chain diversity as they scale clusters to hundreds of thousands of nodes.

Further evidence of the infrastructure ramp is seen in the supply chain for raw materials. Per Reuters (May 2026), NVIDIA invested $300 million to expand Corning's U.S. optical fiber manufacturing capacity, anticipating a massive surge in fiber demand for AI-driven data centers. As data transmission requirements move toward 1.6 Tb/s per lane, single-mode fiber is expected to capture over 60% of the interconnect market due to its superior performance over distance compared to multimode alternatives, according to Mordor Intelligence (April 2026).