
AMD MI355X
AMD's CDNA-architecture AI accelerator chip deployed in Cambridge DAWN supercomputer expansion, May 2026.
Last refreshed: 21 May 2026 · Appears in 1 active topic
Why did the UK choose AMD MI355X chips for the Cambridge supercomputer expansion?
- What is the AMD MI355X chip used in the Cambridge supercomputer?
- The AMD MI355X is a CDNA-architecture AI accelerator chip selected for the expansion of Cambridge's DAWN supercomputer into Zenith, delivering sixfold compute capacity as part of a £36m DSIT and UKRI investment.Source: GOV.UK
- How does the AMD MI355X compare to Nvidia GPU chips?
- The MI355X is AMD's direct competitor to Nvidia's H100/H200 GPUs for AI training workloads at data-centre scale, targeting the same high-performance compute market.Source: AMD
- Why did the UK choose AMD rather than Nvidia for the DAWN supercomputer upgrade?
- Public procurement details have not been disclosed, but AMD MI355X was selected for the Zenith expansion announced May 2026; global Nvidia H100 allocation constraints have been a factor in many national compute procurement decisions.Source: GOV.UK
Background
The AMD MI355X is an AI accelerator chip built on AMD's CDNA (Compute DNA) architecture, designed for large-scale machine learning workloads including training and inference at data-centre scale. It was selected as the hardware basis for the expansion of Cambridge's DAWN supercomputer, announced on 19 May 2026 as part of a £36m joint DSIT and UKRI investment to bring the Zenith successor system online with sixfold compute capacity .
AMD's MI-series accelerators compete directly with Nvidia's H100 and H200 GPUs for AI workloads. The MI355X is part of AMD's Instinct series, targeting the high-performance compute market that has been reshaped by demand for AI training infrastructure. UK procurement of AMD silicon rather than Nvidia reflects both supply-chain considerations and AMD's willingness to engage with government-led compute programmes at a time when Nvidia H100 allocation has remained constrained globally.
The choice of MI355X for Zenith has implications beyond the Cambridge facility. It establishes AMD as a hardware partner in Britain's sovereign AI compute stack, creating a procurement reference point for other public-sector compute investments. ARIA's parallel £100m Scaling Compute programme, including the £50m Scaling Inference Lab, will likely face the same silicon-sourcing decisions. At a time when AI chip geopolitics are dominated by US export controls on advanced semiconductors to non-allied states, the MI355X deployment at Cambridge signals where the UK sits in the global chip allocation hierarchy.