Independent report

Future of Compute Review - Terms of reference

Updated 6 March 2023

This was published under the 2019 to 2022 Johnson Conservative government

Context

In the past decade, compute has grown by orders of magnitude as a critical general purpose technology for productivity, prosperity, and innovation. With 90% of UK businesses using the cloud, and the cloud market doubling every 3-4 years, nearly every business now has the potential for immediate access to supercomputer-scale compute.

At the same time, the type and use of compute is changing. Compute is becoming essential to biology, chemistry, physics, and nearly every other area of research. The performance of the fastest supercomputer in the world has grown 200x since 2010, while the compute requirements of the largest machine learning models has grown 10 billion times. For example, artificial intelligence has begun to outperform other methods at traditional areas of high performance computing, such as biological and chemical modelling, climate modelling and cosmology, and financial services.

Recently we have seen the rise of graphics processing units (GPUs), novel custom-purpose chip designs, industry as the largest users of compute for research, plausible quantum computing, questions about semiconductor resilience, and a small number of massive chip developers and computing providers mostly in the United States.

In sum, this government foresees a world where nearly every aspect of business and research is transformed by some aspect of compute – big, small, complex, or simple.

It is therefore important that we review the UK’s future compute needs at this crossroads of a new generation of compute.

We must analyse the supply of inputs like semiconductors and chips, the accessibility of computing systems, their size and efficiency, the skills of its users, and the presence of providers. Interventions could include computing facilities, incentives, access, skills, technology development, partnerships, and more.

Rather than analysing compute as simply a system for research, this review will view compute as a strategic, all-encompassing resource. With that lens, we should ask what our compute needs will be in the next decade across the entire economy and how we should meet them. We acknowledge that other countries, namely the US, China, Japan, and those of the EU, are increasingly thinking similarly.

Objectives

To support the government’s integrated review ambition for the UK to be a science and technology superpower, we need a targeted approach to compute. The UK’s compute capability is falling behind the needs of our world-leading tech sector. The future of compute review will produce recommendations for action to remedy this. The goals of the review are to:

• Understand the UK’s compute needs over the next decade
• Identify cost-effective, future-facing interventions that may be required to ensure research and industry have internationally competitive access to compute
• Establish a view of the role of compute in delivering the Integrated Review and Science Superpower this decade

Outputs

• Building on last year’s Government Office for Science report on Large Scale Computing, the review will provide:

• An analysis of the UK’s compute needs
• An analysis of the role of compute in delivering the Integrated Review and Science Superpower this decade
• A set of discrete, targeted policy recommendations on potential interventions that will form the basis of the UK’s compute strategy over the next decade

Areas of focus (thematic)

The review will provide an assessment of:

• The importance of compute
• International approaches to compute
• Metrics for UK provision of compute and measures of compute competitiveness.
• The structure of the compute market

Building on this assessment, the review will provide policy recommendations on these seven main areas of focus:

• Access - Who will be the users of compute and how will they access it? How can the UK alleviate asymmetry in access to compute? What skills and services are needed by which users to support access?
• Provision - What are the potential use cases for compute?* What provision of compute would best support research and industry needs, increase productivity and strengthen the UK’s international competitiveness?
• Infrastructure design - Which architectures will the UK need in the future? What are the environmental considerations for compute?
• Security - How can concerns over cyber security, economic security and supply chains, privacy and intellectual property be mitigated?
• Enablers - How will this complement existing DCMS priorities across digital skills, tech adoption, federated data and infrastructure?
• Sector needs - What sectors will have growing compute needs over the next decade and what specific actions are required to support their compute requirements.
• Role of government – Is there a rationale for government intervention in compute markets? If so, what government intervention is necessary in the compute ecosystem and how does this compare to approaches taken internationally? What model should potential intervention take: government-sponsored provision; market intervention; or another form?