The Center for Data Innovation spoke with Nick New, founder and CEO of Optalysys, a startup based in Leeds, England that uses photonic computing to help companies innovate using sensitive data through novel encryption methods. New discussed how Optalysys’ chip enables fully homomorphic encryption, how this technique could bolster data-driven innovation in medical research, and how using this technique could bolster the training of AI models using encrypted data.
This interview has been edited.
Martin Makaryan: What is Optalysys and what inspired you to start the company?
Nick New: My journey in optical or photonic computing began when I was pursuing my PhD at Cambridge University in the late 90s, where I focused on optical pattern recognition, a technique to identify patterns using light, and the analysis of large video frames and images. This early work laid the foundation for my first company, Cambridge Correlators, which eventually evolved into Optalysys.
At Optalysys, we leverage the unique properties of light to achieve greater computing speeds and efficiency. Unlike traditional computing, which relies on electrons moving through semiconductors, photonic computing uses photons to process, store, and transmit data. Photons, moving at the speed of light, encounter less resistance or physical limitations, resulting in faster data transmission and significantly increased bandwidth. These attributes make photonic computing a powerful method for handling large amounts of data simultaneously and at much higher speeds.
Initially, we focused on processing genetic data and image-based AI models, but we quickly recognized the potential of photonic computing in enhancing data security. We shifted our attention to using this technology for processing encrypted data. Our photonic chips are designed to perform the complex and computationally intensive mathematical functions like Fourier transforms required for fully homomorphic encryption (FHE)—a form of encryption that allows computations on encrypted data without needing to decrypt it, unlike traditional encryption. We believe Optalysys will enable enterprises to adopt FHE at scale, ensuring that sensitive data remains secure while being processed at unprecedented speeds.
Makaryan: What are the benefits of fully homomorphic encryption?
New: FHE allows multiple parties to work on data without seeing or exposing it. This opens up new possibilities for sensitive data use, especially in cases where legal or commercial reasons prevent data-sharing. We believe that FHE can accelerate collaboration and innovation using data. For example, banks could share data for fraud detection without exposing sensitive information. However, processing data under FHE is extremely slow on conventional hardware, requiring a million-fold increase in operations per logic operation.
Makaryan: What kind of hardware does FHE require, and how does Optalysys address this challenge?
New: Conventional processors like central processing units (CPUs) or graphic processing units (GPUs) are not well-suited for FHE. Processing data under FHE is extremely slow on conventional hardware, requiring a million-fold increase in operations per logic operation. New types of processors and custom application-specific integrated circuits are in the works, but they won’t address the challenges alone. Our innovation integrates photonic and digital elements into a single hybrid circuit, combining the best of both worlds. We control light within silicon channels on a chip, using technology that converts data into a form that can be processed optically. This allows us to convert digital data directly into light, making it easier to perform FHE by taking advantage of the speed and efficiency of photonic computing.
Makaryan: How can Optalysys help drive data-driven innovation?
New: FHE enables new ways of collaborating and sharing data, helping enterprises and people extract more value from data. Our role is to make FHE at scale viable for companies by building the powerful hardware this technique requires and accelerating enterprise adoption. For example, in medical research where sharing health data is difficult, especially across borders, FHE could enable sharing of information at earlier stages, potentially accelerating vaccine development, without compromising data privacy or security. Many other industries could benefit from a mass-scale adoption of FHE. Our vision is for FHE to become part of the fabric of secure computing, operating in the background without users needing to understand its complexity. We aim to become the dominant supplier of hardware for secure computing.
Makaryan: How might advancements in AI affect your plans?
New: FHE offers a higher level of security for AI model processing, both during inference and training. FHE can protect the AI models by preventing exposure to end users, and it can encrypt the data fed into models. This is especially important for large language models, where FHE allows developers to securely and privately train and improve models using user queries without decrypting the data. That is why we see FHE becoming increasingly important in ensuring security and privacy in AI processing. This focus on security is becoming a priority for AI developers as well to build trust in society and build on current advances in AI. We think that FHE can be a boost to AI development writ large.