AI BIG Idea Series | Patrick Bowen & Preston Woo | Neurophos

Join us for another session of our 2026 Generative AI Big Idea Series with Patrick Bowen, Co-Founder and CEO of Neurophos, and Preston Woo, the company’s CFO and Chief Commercialization Officer.

Power availability and semiconductor manufacturing capacity are becoming increasingly important constraints on the industry’s ability to scale. Neurophos is developing a photonic AI accelerator designed to perform compute-intensive matrix multiplication in the optical domain, while using conventional digital silicon for memory, control and other operations.

We will examine the physics behind Neurophos’ approach, why the company is initially targeting prefill workloads, and what it would take to move optical computing from a promising technology to a commercially deployed accelerator platform. We will also discuss the potential implications for AI infrastructure economics, power consumption and the competitive landscape.

Topics will include:

• Why optical compute, why now: How power, wafer capacity and the slowing pace of conventional semiconductor scaling are creating an opening for new computing architectures.
• Inside the architecture: How Neurophos combines optical tensor cores with digital silicon, memory and conventional packaging.
• The scaling advantage: Why larger optical arrays improve compute efficiency, and the physical breakthrough that enables Neurophos to integrate them at chip scale.
• Prefill versus decode: Why Neurophos is initially focused on prefill, how that workload is evolving and where photonics may offer the greatest advantage in the long run.
• Performance and economics: Performance-per-watt and compute-per-wafer versus traditional compute platforms, and how they translate into system-level performance and cost.
• Manufacturing and integration: The foundry, packaging, precision, reliability and software challenges involved in bringing a hybrid optoelectronic accelerator to market.
• Commercialization roadmap: The path from working technology to engineering samples, volume production and hyperscaler deployment.
• Competitive positioning: How Neurophos compares with GPUs, ASICs and other alternatives.
• Audience Q&A