Google DeepMind’s staff costs surge amid AI frenzy

Rising staff costs at DeepMind, Alphabet’s artificial intelligence laboratory, drove up total administrative spending at the business by £350 million to £1.4 billion last year, the company said.
It is a reversal from 2022 when the research business cut employee costs by 39 per cent, after a drop in profit of 40 per cent, mirroring a pattern of layoffs across the tech industry in the wake of the pandemic boom.
Staff costs and related expenses last year rose by £231 million to £826 million, the company said. This figure is still lower than in 2021, however, when staff costs were reported to be £969 million.
A spokeswoman for Google DeepMind said it did not publicly share its staff number and added: “Any attempt to calculate average employee salary using these figures would be highly inaccurate.”
In a sign of the ever increasing interest in the potential of AI, sales at the London-headquartered company rose by £450 million to more than £1.5 billion, according to its latest accounts filed with Companies House for the year to December 2023.
Pre-tax profit was £148.4 million, up from £72.8 million in 2022, and it paid a dividend of £175 million for the year, having paid nothing the year before. The company paid £35 million of tax, up from £12 million in 2022.
Deepmind Technologies was founded in 2010 by Sir Demis Hassabis, Shane Legg and Mustafa Suleyman. The company makes money by developing AI systems and providing research to other parts of the Alphabet empire and externally.
It was acquired by Google for £400 million in 2014. It has been responsible for a number of breakthroughs in AI, most recently in its AlphaFold project, which can predict how proteins behave and could have applications in antibiotics, cancer therapy and materials science.
It also built AlphaGo, the first program to beat the world champion at the game of Go, and recently announced the development of AI robotic hands, teaching a robot to use two at the same time, as a human would, for example, to tie shoelaces or put a shirt on a hanger.
Gemini, Alphabet’s large language model, is powered by some of the techniques that underpinned AlphaGo.
The company’s mission statement is “to build AI responsibly to benefit humanity and improve billions of people’s lives”.
As companies race to develop ever more sophisticated AI models, governments are trying to make sure the technology is used responsibly. The Labour administration has said it plans to introduce laws for the most advanced models while the EU has already implemented an AI Act.
In this latest set of results, DeepMind cautioned that evolving regulation could hit the company’s revenues because of costs associated with the red tape. It also noted that developing AI systems was “an emerging market characterised by continuous change and intense competition”.
At the recent Times Tech Summit, Hassabis warned of the power of the technology and predicted that artificial intelligence with general human cognitive abilities would be reached within ten years.
The company’s aim was to build a general AI system that “is capable of doing any cognitive task that humans can”, he said.
DeepMind made £14.4 million of academic donations and sponsorships last year, up from £9.9 million the year before, the accounts showed.