Isambard AI

A Voluntary Ex-Ante Transparency (VEAT) Notice
by UNIVERSITY OF BRISTOL

Source
Find a Tender
Type
Framework (Supply)
Duration
not specified
Value
£176M
Sector
TECHNOLOGY
Published
10 Nov 2023
Delivery
not specified
Deadline
n/a

Concepts

Location

Emersons Green, Bristol BS16 7FS

Geochart for 2 buyers and 1 suppliers

2 buyers

1 supplier

Description

Isambard-AI is a new GPU-based Artificial Intelligence (AI) computing system to deliver a national AI Research Resource (AIRR) for the UK, as recommended in the “Independent Review of The Future of Compute: Final report and recommendations ”, and to support the Frontier AI safety taskforce. This new AIRR will support several initiatives, particularly on AI safety and foundational models, where the UK aims to establish a leadership position. Isambard-AI will be delivered as a “self-contained unit”. The RFP and Contract covers, in summary, the following mandatory requirements: 1. The scale-out GPU-based production system 2. An early access GPU-based system with AI storage 3. High performance network fabric 4. System management support for multi-tenancy: 5. Support for 5 years of service 6. One or more self-contained units 7. Energy efficiency 8. Storage solutions optimised for AI training and inference 9. AI supercomputing facility access and gateway nodes 10. Data Centre Connectivity and Edge Router for Services 11. UPS 12. Data transfer and mover tool 13. Software and licenses 14. Security monitoring, auditing and compliance 15. Monitoring, reporting and RAS tools 16. Eventual decommissioning 17. Training / handover 18. Environmental impact

Total Quantity or Scope

A call for a mini competition was made on 13 September 2023 under the Southern Universities Purchasing Consortium (“SUPC”) Framework Agreement for the Supply of Servers, Storage and Solutions National Agreement (“SSSNA Framework Agreement”) by issue of a Request for Proposals ("RFP") to all of the Lot 4 (Solutions – HPC and/or DIC) Panel members. The SSSNA Framework was established pursuant to the Find a Tender Notice reference 2021/S 000-027936 published on 8 November 2021. Lot 4 panel members were offered a site visit to view the site for the upcoming Isambard 3 Tier-2 HPC service at the National Composites Centre within the Bristol and Bath Science Park. The RFP identified the mandatory requirements (scored on a Pass/Fail basis), the scored desirable requirements, set out the award criteria and scoring methodology and provided a response template and draft call off contract based on the SSSNA Framework Agreement. The RFP gave Lot 4 panel members 2 weeks to respond (until 5pm Wednesday 27 September 2023), the evaluation concluded on 2 October 2023 and a voluntary standstill period concluded on 31 October 2023. Delivery of an initial 150+ GPUs and site preparation is required in early 2024 to enable Phase 1 of Isambard-AI to go live by the end of March 2024, in order to meet milestones set by the Department of Science, Information and Technology (“DSIT”) in the grant funding of Isambard-AI via UKRI. DSIT’s plans for the new AI Research Resource at the University were announced as part of a £900m investment to transform the UK’s computing capacity, by press release of 13 September 2023. Only one tender was received in response to the RFP (from Hewlett-Packard Limited trading as Hewlett-Packard Enterprise, the provider of Isambard 3). The University proceeded to evaluate the Hewlett-Packard Enterprise tender response, confirmed that it met the mandatory requirements and scored the response to the desirable requirements (see below). Standstill letters were issued to the Lot 4 panel members on 18 October 2023. The call-off contract in relation to the supply of Isambard-AI with Hewlett-Packard Limited (“the Contract”) has been finalised and is ready for signature.

Award Detail

1 Hewlett Packard (Berkshire)
  • Isambard AI
  • Value: £175,875,000
  • Subcontracting opportunities are expected.

CPV Codes

  • 44316400 - Hardware

Indicators

  • Award on basis of price and quality.

Other Information

Isambard AI

Reference

Domains