AUSTIN – Neurophos, an AI chip startup developing optical processors for machine learning inference, raised $110 million in Series A funding led by Gates Frontier, the investment fund managed by Bill Gates.
The round included participation from Microsoft’s M12 venture arm, Carbon Direct, Aramco Ventures, Bosch Ventures, Tectonic Ventures, and Space Capital. The Austin-based company spun out of Duke University with proprietary technology that uses light-based computing rather than traditional electronic processors.
Neurophos has developed what it describes as metasurface modulators that function as photonic tensor cores, enabling thousands of light-based computing neurons on a single chip. The technology aims to dramatically accelerate AI inference workloads while reducing power consumption compared to conventional GPU-based systems.
The company’s approach addresses growing concerns about the energy intensity of large-scale AI deployments. As enterprises move beyond experimental AI projects into production environments, infrastructure power requirements have become a critical constraint for data center operators and cloud providers.
Photonic computing represents a fundamental departure from silicon-based processors. By performing mathematical operations using optical signals rather than electrical ones, the technology promises both speed advantages and significant energy efficiency gains. Industry analysts suggest photonic AI accelerators could reduce inference costs by orders of magnitude if successfully commercialized.
The Series A funding will accelerate development of Neurophos’s ultra-fast, energy-efficient AI accelerators designed for machine learning inference at scale. The company plans to deploy capital toward advancing chip manufacturing processes and building partnerships with hyperscale cloud providers.
Gates Frontier’s involvement signals strong confidence in photonic computing as a viable path for next-generation AI infrastructure. Microsoft’s M12 participation suggests potential integration opportunities with Azure’s AI cloud services, though no formal partnerships were announced.
Neurophos joins a growing field of alternative AI chip architectures competing with established GPU manufacturers. While graphics processing units from companies like NVIDIA currently dominate AI training and inference markets, startups developing specialized accelerators argue that purpose-built hardware will prove more efficient for production workloads.
The funding comes during a period of intense investment in AI infrastructure companies as venture capitalists seek to capitalize on the computing requirements of increasingly sophisticated models and widespread enterprise AI adoption.
Previous
