Getting the right hardware where it needs to be, since 2014.

GPU Direct is the GPU business of Options Direct Ltd, a founder-led datacentre hardware specialist headquartered in Manchester.

Our story

Options Direct was founded in 2014 to supply the latency-critical end of the market: trading firms that measure performance in nanoseconds and cannot afford hardware that arrives late or misbehaves in production.

That work taught us to configure and test every system in our own centres, to ship it ready to install, and to stay with it for its whole life, from design through to buy-back. When AI teams began building GPU clusters, the same discipline carried over.

GPU compute is now the largest part of what we do. GPU Direct brings it together in one place: HGX systems, NVL72 racks, GPUs, networking, optics and cabling, with the engineering, compliance and logistics that go with them.

Founded
2014, Manchester
Ownership
Founder-led
Configuration centres
Manchester, Frankfurt, Dubai
Shipped to
40+ countries
Testing
72-hour burn-in on every system
Quality
ISO 9001:2015
LDC The Top 50 2026

Ones to Watch, LDC Top 50 Most Ambitious Business Leaders 2026

Options Direct’s leadership was named in LDC’s Top 50 Most Ambitious Business Leaders for 2026, in the Ones to Watch category. Featured in The Times.

Offices and configuration centres

Stock is held on three continents, and systems are built and tested in our own centres before they ship to more than 40 countries.

  • MAN

    Manchester (Altrincham)

    Headquarters and configuration centre

  • LON

    London

    Satellite office

  • FRA

    Frankfurt

    EMEA logistics and configuration

  • DXB

    Dubai Internet City

    Middle East, Africa and APAC

  • USA

    United States

    Entity, configuration centre and warehouse

    Opening soon

Flight cases and crated servers ready to ship in a warehouse

How we work

One accountable partner from the first sizing conversation to the day the hardware is bought back.

  1. 01

    Design

    Sized from the workload, with the power, cooling and network design agreed before anything is ordered.

    • Bill of materials
    • Rack elevations
    • Power and cooling model
    • Network topology and port maps
  2. 02

    Compliance

    Every order clears export classification, end-user and end-use checks and screening before it is confirmed.

    • End-user statement
    • Know-your-customer checks
    • Export classification
    • Licences where needed
  3. 03

    Configure

    Built in our own centres, firmware and BIOS set to your estate, then 72 hours of thermal and stress testing.

    • Signed QA report per unit
    • Serial and asset register
    • Firmware baseline
    • Burn-in logs
  4. 04

    Ship

    Packed in the order you will install it, with export and customs paperwork handled, to the data hall.

    • Export and customs documents
    • Palletised in rack order
    • Tracked, insured delivery
  5. 05

    Install and support

    Rack, cable and commission with local crews, then spares held in-region and support after warranty ends.

    • Rack and stack
    • Cabling to schedule
    • Spares in-region
    • Post-warranty maintenance
  6. 06

    Buy back

    At end of life we decommission, erase to NIST standards, audit every asset and buy it back.

    • Certified data erasure
    • Audited asset register
    • Chain-of-custody reporting

Partners

We are an NVIDIA partner and an AMD partner, and we supply NVIDIA networking (formerly Mellanox), switching, optics and cabling alongside systems from the major server makers.

Planning a GPU purchase or deployment?

The GPU Direct Briefing

Once a month: GPU prices and lead times as we see them, new guides, and what changed in export rules.

Monthly. Unsubscribe in one click. We use your email for the Briefing and nothing else.