STL Brings Neuralis to US Data Centre Market for AI Scale

STL Optical Connectivity, a subsidiary of STL, has introduced its Neuralis suite of data centre connectivity solutions in Washington, DC.

The launch focuses on how data centres are adapting to AI workloads and hyperscale computing, which place new demands on fibre infrastructure and interconnect design.

The Neuralis portfolio is positioned as a connectivity layer for modern data centres, where performance depends on how effectively systems move large volumes of data between servers and processing units.

AI training workloads in particular rely on GPU clusters, which require dense, high-speed connections to function efficiently.

STL frames Neuralis as a response to these requirements, drawing on customer collaboration to address constraints around space and deployment speed.

With a growing footprint across the US, STL focuses on supporting operators as they scale infrastructure to handle AI-driven demand.

Rethinking data centre traffic flows

One of the core challenges highlighted in the launch is the shift in traffic patterns within data centres.

Traditional architectures prioritise north-south traffic: data moving between users and servers. AI and hyperscale environments instead generate east-west traffic, where large volumes of data move laterally between servers inside the data centre.

This shift increases pressure on internal connectivity. GPU-heavy systems require thousands of fibre connections to maintain throughput, while also managing power and physical space constraints.

Neuralis addresses this by providing high-density fibre cabling that supports large-scale interconnections without expanding the physical footprint.

The suite is structured around two main areas. The first focuses on optimising usable space within a data centre. It uses pre-terminated fibre cables that are assembled and tested in a factory before delivery, which removes the need for on-site installation work. 

The second area centres on high-speed data centre interconnect (DCI) which links separate data centre facilities or connects different parts of a campus, enabling data to move between locations at scale.

High-density fibre for AI infrastructure

Within the DCI category, STL's Celesta IBR series is designed for high-density deployments.

These cables support up to 6,912 fibres within a compact form factor, allowing operators to increase capacity without expanding rack space or cable pathways.

The design also considers the thermal and safety requirements associated with AI systems, where dense hardware generates higher levels of heat.

Ankit Agarwal, Managing Director, STL says: "AI demands a level of precision and density that traditional cabling simply cannot meet.

"With STL Neuralis, we are providing the high-speed, low-latency foundation that allows GPU clusters to perform at their peak, moving complexity out of the field and into a controlled, high-precision factory environment."

STL's vertically integrated manufacturing model also covers the full lifecycle of fibre production. This begins with ultra-pure preform, where silica glass is prepared for fibre creation.

It then moves through fibre drawing, a process that pulls the glass into thin strands with precise characteristics. These fibres are assembled into cables and finally connectorised into plug-and-play systems that can be deployed directly in data centres.

Moving complexity out of the field

The launch highlights a clear operational priority, shifting work from on-site environments to manufacturing facilities.

By pre-assembling and testing components, STL aims to reduce deployment risks and improve consistency, particularly in large-scale builds where time and accuracy are critical.