NVIDIA
GPUs, DGX and HGX systems, InfiniBand, and Spectrum-X networking for AI training, inference, and cluster fabrics.
CircularTelecom
AI/HPC Infrastructure
Vendor-agnostic support for GPU clusters, high-performance computing estates, 400G, 800G and 1.6T optics, and ARC asset recovery. Shorter lead times, lower cost, and a cleaner route for surplus AI infrastructure.
Data Centre Velocity
Artificial intelligence and high-performance computing are changing data centres faster than anything we have seen in 15 years in this industry. Demand for GPUs, high-speed networking and optical connectivity has pushed lead times out and prices up.
At the same time, organisations are sitting on hardware they no longer need after an upgrade, or bought for a project that never went ahead. Circular Telecom helps you recover value from surplus infrastructure, source hardware faster, and connect clusters with cost-effective, MSA-compliant optics.
Vendor-Agnostic By Design
GPUs, DGX and HGX systems, InfiniBand, and Spectrum-X networking for AI training, inference, and cluster fabrics.
GPU servers and high-density compute platforms for rapid AI and HPC capacity expansion.
PowerEdge AI and HPC servers for enterprise, research, and hosted infrastructure environments.
ProLiant, Apollo, and Cray systems across private AI, research, and high-performance compute estates.
High-radix 400G and 800G Ethernet switching for AI fabrics and data centre networking.
The right blend of new, surplus, and refurbished equipment for your budget, timescale, and technical needs.
ARC For AI/HPC
Send your inventory in any format: GPUs, GPU servers, switches, optics, memory, CPUs, or complete platforms.
We price your equipment against real market demand and provide a clear proposal.
Select consignment on our 50/50 model or an outright purchase, depending on cash flow and timing.
We collect, audit, document, and manage equipment through the route to resale.
You receive reporting on CO2e avoided through reuse to support ESG and Scope 3 reporting.
High-value AI hardware gets back into the market quickly, protecting residual value and returning capacity to a supply-constrained sector.
High-Speed Optics
Optical connectivity is one of the biggest bottlenecks in AI infrastructure. OEM-branded transceivers can carry lead times of six months or more, holding up cluster deployments for neoclouds, enterprises, and service providers.
Circular Telecom supplies MSA-compliant optical transceivers, DACs, and AOCs as a proven, cost-effective alternative. Lead times are typically 4-6 weeks depending on specification and volume.
Sourcing
Data centre GPUs for AI training, inference, and HPC workloads, as individual accelerators or complete GPU server platforms.
High-core-count server processors for GPU host nodes, HPC compute, and supporting infrastructure.
High-capacity DDR4 and DDR5 RDIMMs for new builds, upgrades, or replacement modules, matched to the platform requirement.
400G and 800G Ethernet and InfiniBand switching for spine-leaf, back-end GPU fabrics, and front-end networks.
Vendor-agnostic advice, new, surplus, and professionally refurbished options, quality assurance before shipment, and one partner for sourcing, optics, and asset recovery.
When a cluster is waiting on one component, every week of delay costs money. Our sourcing model is built to keep deployment plans moving.
Optical Connectivity
400G QSFP-DD and OSFP, 800G OSFP and QSFP-DD, and 1.6T next-generation optics for the latest AI fabrics.
Direct Attach Copper for short in-rack links between GPU servers and top-of-rack switches.
Active Optical Cables for longer runs within and between racks while supporting 400G and 800G links.
Multi-Source Agreement standards define form factor, electrical interface, and optical performance, giving buyers availability, cost control, and flexibility.
Tell us your platforms and we will confirm the correct coding route before you order.
A practical option for keeping live clusters running without waiting months for OEM-branded replacements.
Who We Work With
GPU-as-a-service providers scaling capacity quickly and looking for faster hardware and optics supply.
Supplementary supply for high-density infrastructure and constrained deployment timelines.
Private AI infrastructure teams sourcing compute, networking, memory, and optics without being locked to one vendor route.
HPC clusters, GPU platforms, and practical lifecycle support for constrained budgets.
Asset recovery, optics, and sourcing support across high-density hosting environments.
Support for AI projects where delivery dates, specialist components, or recovery routes are under pressure.
Sustainable AI Infrastructure
AI hardware carries a large carbon footprint, and much of it sits in manufacturing. By reusing equipment through ARC and extending the working life of existing infrastructure, Circular Telecom helps you build and run AI capacity that is decarbonised and profitable.
The commercial logic is direct: surplus hardware gets a responsible route back to market, buyers get access to constrained supply, and finance teams get clearer visibility on value recovery.
AI/HPC FAQ
Yes. Through ARC we recover value from surplus GPUs, GPU servers, switches, optics, memory and CPUs, including new and unused equipment from cancelled or paused projects. We offer consignment and outright purchase.
MSA-compliant optics are built to recognised industry specifications, and carrier networks have used them at scale for many years. Every unit we supply is coded to your environment and quality-checked before dispatch.
Lead times are typically 4-6 weeks, depending on specification and volume. OEM-branded equivalents can take six months or more.
Our transceivers, DACs and AOCs are coded to your specific working environment. Tell us your platforms and we will confirm the compatibility route before you order.
We are fully vendor-agnostic and work across NVIDIA, Supermicro, Dell, HPE, Arista and other leading manufacturers.
Yes. ARC valuations can include an estimate of the CO2e avoided through reuse to support ESG and Scope 3 reporting.
Circular Telecom is based in Wales, UK, and works with customers across the UK, Europe and worldwide.
NVIDIA, Supermicro, Dell, HPE and Arista are trademarks of their respective owners. Circular Telecom is an independent supplier and is not affiliated with or endorsed by these manufacturers. Third-party optics are MSA-compliant and coded for compatibility; they are not OEM-branded products.
AI/HPC Project Brief