Yahoo Finance How Britain ceased to be a computing superpower
Britain has fallen behind Russia and China in the global supercomputing race and has a short window to catch-up, the Government has been warned.
Rishi Sunak is being urged to invest billions in building new supercomputers that are capable of training advanced artificial intelligence (AI) programmes, which are seen as crucial to future economic and technological growth.
A lack of investment now “threatens [the UK’s] standing as an international leader in science and technology”, the Government-commissioned Future of Compute Review warned.
The review, published on Monday, found Britain has rapidly dropped down the global rankings for computing capabilities over the last 20 years.
Britain is now tenth in the world by total supercomputer capacity, having ranked third in 2005. No British machines are in the top 25 by power globally and just two are in the top 100.
Russia, meanwhile, has five, and Germany has 12. Russia, China, the US and France now all outrank the UK for total large-scale computing power.
Supercomputers matter because of their ability to conduct complex modelling that can herald important – and sometimes lucrative – breakthroughs. Applications range from climate change modelling to drug discovery and nuclear fusion research.
New markets such as artificial intelligence also rely on these large machines.
Matt Clifford, chair of the Advanced Research and Invention Agency, the £800m fund championed by Dominic Cummings to invest in UK science, told The Telegraph: “The reality today is the UK doesn’t have a very large amount of sovereign compute [computer power].
“This is an area where the UK should be prepared to spend billions over the next decade to ensure the most important AI companies are being built in the UK.”
Since the 1960s, scientists have used supercomputers for research. The vast machines have the power of tens of thousands of personal PCs and can execute hugely complex computer models that can imitate real life.
The US has traditionally led the field, but Britain has had a reputation for computer science and artificial intelligence research.
Today, the UK’s most powerful machine is Archer2, a supercomputer based at the University of Edinburgh. Built at a cost of £79m, it features two dozen cabinets filled with thousands of processors made by US chipmaker AMD. A vast system of pipes and water tanks uses the brisk Scottish air to pump cool water around the machines to keep them from overheating.
Archer2’s processing power is equivalent to a quarter of a million modern laptops and Boris Johnson hailed the machine as the “real horsepower” behind Britain’s science during a visit last February.
However, Archer2 is 100 times less powerful than Japan’s most powerful computer and 56 times less powerful than America’s best machine.
The Met Office’s Cray supercomputer was at one point among the world’s most advanced machines but is now ranked 86th, a decline emblematic of the UK’s drift.
The Future of Compute review, which was led by Cambridge scientist and Google executive Zoubin Ghahramani, concluded: “The UK lags behind other advanced economies in compute.”
Prof Mark Parsons, of Edinburgh University, who runs its supercomputing programme, said: “It would be wrong to say the UK has fallen behind because it hasn't spent money. But over the last decade our peers have been spending more.”
Investment has surged in part because of growing commercial applications. Supercomputers are necessary for training AI tools as well as developing quantum computing.
ChatGPT was trained using supercomputer power from Microsoft. Silicon Valley start-up OpenAI, which developed the chatbot tool, was valued at a reported $29bn at the beginning this year after debuting its new product.
With growing interest in AI and quantum, demand for supercomputers and other so-called large-scale compute systems is fast oustripping supply.
One scientist who consulted on the Future of Compute review said: “It used to be a small group of scientists who needed access to supercomputers: now there is enormous demand for resources.”
Increasingly, academics rely on access provided by tech giants such as IBM, Microsoft or Amazon. However, competition for these commercial resources is fierce and such international systems may not be appropriate for the most complex projects or national security programmes.
Prof Mike Wooldridge, of the Alan Turing Institute, said growth in demand for large-scale computer facilities “has been incredibly rapid, and the UK got left behind.”
The Future of Compute review points out that no UK team has won the Gordon Bell Prize, the most prestigious award for computer research, since 2011.
The UK has some of the world’s leading quantum computing scientists, but it is a nascent industry and requires support.
Prof Parsons, of Edinburgh University, said the UK has lacked a “10-year strategy” on how to invest in computer infrastructure.
Edinburgh’s original Archer computer was eight years old when it was shut down and uncompetitive by the time its successor came online.
“Really there should be a continuous renewal of these systems,” he said.
Plans are in motion to build some new state-of-the-art computers in the UK.
In 2020, Microsoft and the Met Office announced they would build a new weather forecasting system to tackle climate change. The £1.2bn project aims to design the world’s most powerful weather computer that will operate for the next decade.
Nvidia has also been building a super-powered facility in Cambridge.
However, the Future of Compute review said Britain needs “government intervention to provide strategic leadership [and] ensure sovereign compute capability.”
On Monday, Mr Sunak and Michelle Donelan, the technology minister, announced plans to turn the UK into a “science superpower”. The proposals included committing to a new “exascale” computer, the most modern type of supercomputer, which was a key recommendation of the review.
Britain must have a system on this scale ready for action by 2026 to join the artificial intelligence race, the Future of Compute review said. Scientists also said the UK needed to create a national pool of computer power, equivalent to 3,000 AI processors, for scientists to access as soon as this summer.
However, the Government has not set out how much funding has been earmarked for the exascale project. Peer supercomputer systems cost in the region of $600m, the Future of Compute review found.
Without this level of investment, the most promising AI companies could end up drifting from Britain to where they can access the fastest computing resources.
“The need to provide state of the art high performance computing facilities is every bit as essential as the provision of broadband was for businesses and households,” the Turing Institute's Wooldridge said.
“This is not a box that can be ticked and then forgotten about.”
A government spokesman said ministers were “taking action in response to the recommendations of the Future of Compute Review”.
