Photons versus power walls
Neurophos, a Duke University spinout, has closed an oversubscribed $110 million early-stage round to build optical chips that aim to replace traditional electronic pathways in AI accelerators. Its pitch is simple but radical: pack massive optical parallelism onto a single chip to break through the power and bandwidth limits throttling conventional GPUs.
At the heart of its technology are micron-scale metamaterial optical modulators, reported to be up to 10,000 times smaller than existing photonic elements. By tightly integrating these modulators with compute‑in‑memory architectures, the company targets the main bottleneck in modern AI systems: the cost of moving data, not just crunching it.
From lab demo to data center pilot
Neurophos plans to channel the new capital into its first integrated photonic compute systems, including data‑center‑ready optical processing unit (OPU) modules and an accompanying software stack for early adopters. The startup is partnering with Norwegian data center operator Terakraft on a real‑world pilot, with initial systems targeted for 2027 and broader production ramping later in 2028.
The company is also expanding its Austin headquarters and opening a new engineering center in San Francisco to get closer to AI infrastructure buyers and developer ecosystems. That geographic footprint mirrors a broader pattern in deep tech: colocating core physics innovation with the cloud and semiconductor supply chains that can bring it to market.
The photonic spinoff moment
Neurophos is part of a growing cohort of photonic computing startups moving from research prototypes to commercialization, with peers such as UK‑based Optalysys also raising fresh capital to bring photonic AI hardware into US cloud and security markets. These companies are positioning photonics not as a niche accelerator but as a foundational layer for encrypted, energy‑efficient AI workloads.
For universities and corporate R&D labs, the message is clear: photonics is no longer just an enabling technology for communications — it is becoming a strategic axis for AI compute spinoffs. For founders, this shift opens a window to build category‑defining companies at the intersection of materials, optics, and large‑scale AI infrastructure.
