Finchetto – Heterogeneous Photonic Integration for AI-Scale Optical Switching

Finchetto is developing a new generation of all-optical switching technology for AI and high-performance computing infrastructure.

As AI systems scale, moving data efficiently between compute, memory and storage is becoming one of the major constraints on performance, power consumption and cost. Conventional switching architectures rely heavily on electronic processing and repeated optical-electrical-optical conversion, creating increasing penalties in latency, power and scalability.

Finchetto addresses this by switching data directly in the optical domain. Our proprietary architecture combines silicon photonics and thin-film lithium niobate (TFLN) to create ultra-fast, energy-efficient optical switching designed for the bandwidth demands of next-generation AI infrastructure.

Finchetto has already developed, taped out and validated the core photonic chips required for the architecture. The next major commercialisation step is to bring these proven building blocks together through heterogeneous photonic integration, creating a more compact, manufacturable and scalable integrated platform.

The proposed Innovate UK project will accelerate this transition from separately validated photonic devices to an integrated semiconductor platform suitable for commercial deployment. The work will focus on heterogeneous integration, packaging, interconnect, manufacturability and system-level validation of Finchetto's existing photonic technologies.

This is an important value-inflection point for the company. The core photonic technologies have been demonstrated independently; successful integration will convert those proven elements into a more complete product platform and reduce technical risk on the path to commercial scale-up.

The commercial opportunity is driven by the rapid growth of AI compute and the increasing importance of data movement within large-scale systems. Future AI clusters require substantially more bandwidth between accelerators, memory and storage, while simultaneously reducing the energy consumed per bit moved.

Finchetto aims to address this challenge with a differentiated photonic switching platform that combines:

• proprietary all-optical switching architecture;

• silicon photonics;

• thin-film lithium niobate;

• heterogeneous integration of multiple photonic functions;

• very high bandwidth and low latency; and

• a path towards lower-power, highly scalable switching systems.

The investment opportunity extends beyond the scope of the individual project. Finchetto is building a semiconductor and photonics platform for high-bandwidth data movement within AI and HPC infrastructure, targeting a market in which networking, interconnect and power consumption are becoming increasingly important constraints on system economics.

The company is seeking an investment partner alongside Innovate UK funding to accelerate heterogeneous integration, product engineering, manufacturing readiness, packaging, supply-chain development and engagement with strategic customers and partners.

The combination of private investment and Innovate UK funding would allow Finchetto to move rapidly from validated component technologies towards a manufacturable integrated product, while reducing both technical and commercial risk.

The project is strongly aligned with the Semiconductor and AI Hardware theme, particularly integrated photonics, heterogeneous integration, semiconductor system integration and photonic approaches to improving the bandwidth, latency and energy efficiency of AI data movement.

Finchetto's ambition is to establish a leading position in photonic switching for next-generation AI infrastructure, addressing one of the most important emerging bottlenecks in the continued scaling of AI compute.