NVIDIA is smiling! The UK joins the sovereign AI race, aiming to procure 100,000 GPUs in 5 years
According to the plan, by 2030, the UK's sovereign AI computing power will increase by 20 times, mainly for AI applications in academia and public services
Author: Li Xiaoyin
Source: Hard AI
Governments around the world are increasing their investments in sovereign AI computing capabilities.
According to the Financial Times, the UK government will release a report drafted by British venture capitalist Matt Clifford on Monday, planning to significantly expand the AI computing power owned by the government over the next five years, including the construction of a new supercomputer, aimed at establishing a globally competitive AI industry.
The report states that the plan will advocate for the number of GPUs owned by the UK government to reach the equivalent of 100,000 by 2030, meaning that the UK's sovereign AI computing power will increase 20-fold, primarily for AI applications in academia and public services.
The report proposes a total of 50 recommendations, which may also include the following measures: appointing Matt Clifford as a ministerial AI advisor to assist in implementing the report's recommendations; creating AI "growth zones" to expedite the planning and approval of AI infrastructure construction; and establishing an AI Energy Committee to provide advice on AI energy demands (including nuclear energy).
UK Prime Minister Sir Keir Starmer stated:
"Our plan is to make the UK a world leader in AI. This will provide the foundation the industry needs... while also meaning more job opportunities and investment in the UK, more money in people's pockets, and a transformation of public services. This is the change this government is delivering."
Experts analyze that strengthening sovereign computing capabilities in the UK is crucial to ensuring that the country's AI industry reduces its dependence on other countries. As computing infrastructure gradually becomes a battleground for geopolitical competition, the importance of obtaining reliable computing power at reasonable costs is increasingly highlighted