Jensen Huang, the boss of the chipmaker Nvidia, had some advice for UK ministers last week as they signed a multibillion-pound tech deal with the US: burn more gas.
“I’ve every confidence that the UK will realise that it takes energy to grow new industries,” he said. “Sustainable power like nuclear and wind and of course all of that solar is all going to contribute. But I’m also hoping that gas turbines can also contribute.”
That would be in direct conflict with the government’s determination to wean the UK off gas, as part of the transition to net zero.
The multibillion-dollar tech agreement, signed to coincide with the flummery of Donald Trump’s visit, involves welcoming a slew of vast new energy-hungry datacentres, built by US companies. (The term datacentre occurs 14 times in the official release.)
Demand for these digital warehouses has rocketed with the adoption of generative artificial intelligence, which requires vast volumes of data to train its models.
Appearing recently on comedian Theo Von’s podcast, Sam Altman, the chief executive of the ChatGPT creator, OpenAI, mused: “I do guess that a lot of the world gets covered in datacentres over time,” – before blithely suggesting: “Maybe we put them in space.”
Ministers’ determination to roll out the red carpet for US tech is understandable, given Labour’s urgent need for economic growth and the drive to improve public sector productivity, but it has come with little or no public discussion about potential downsides.
That is particularly worrying given the stranglehold over these technologies of a small number of ultra-powerful American companies, with no reason to stick with the UK if the bubble bursts.
Nick Clegg, who, as a former “president of global affairs” at Meta, knows something about selling out to Silicon Valley, called the US-UK deal “just another version of the United Kingdom holding on to Uncle Sam’s coat-tails”.
A paper published last year by sustainability experts from MIT warned that the “unfettered growth” of generative AI, while having clear benefits for particular sectors, also “incurs significant costs”.
The authors, led by MIT’s Noman Bashir, called for what they described as “the frenetic pace of development akin to a Klondike gold rush” to be replaced by “a more deliberate, thoughtful approach that considers long-term environmental impacts and societal needs”.
The International Energy Agency is expecting electricity consumption by the world’s datacentres to more than double between 2024 and 2030, to 945 terawatt hours – equivalent to the current demand of Japan.
Another recent paper found that as generative AI models become larger and more sophisticated they use dramatically more energy – up to 4,600 times more than simpler models. Calling for “coordinated efforts across the AI value chain”, to tackle the environmental impacts of the technology, the authors warn that rapid adoption of AI could lead to a 24-fold increase in the sector’s electricity use.
Alongside the US-UK tech deal was an agreement on nuclear cooperation, which Labour hopes will lead to a series of small modular reactors being built – some of which could help to meet datacentres’ voracious needs. However, even under the accelerated timetable for construction envisaged in the deal, these are likely to take years to arrive.
Research carried out for the Department for Energy Security and Net Zero paints a relatively reassuring picture, by comparing energy usage by datacentres to an analogue counterfactual: a translation task to the use of a human translator, for example. However, it is far from clear as yet that the use of AI is reducing the demand for other sources of energy demand.
Rapid adoption of the technology is coinciding with increasing demand for electricity, as petrol cars give way to electric, and the government tries to persuade households to switch from gas boilers to heat pumps.
skip past newsletter promotion
Sign up to Business Today
Get set for the working day – we’ll point you to all the business news and analysis you need every morning
Privacy Notice: Newsletters may contain information about charities, online ads, and content funded by outside parties. If you do not have an account, we will create a guest account for you on theguardian.com to send you this newsletter. You can complete full registration at any time. For more information about how we use your data see our Privacy Policy. We use Google reCaptcha to protect our website and the Google Privacy Policy and Terms of Service apply.
after newsletter promotion
It remains unclear how these competing demands can – or should – be reconciled.
Datacentres also rely on water, to help cool the humming banks of hardware. In the UK the Environment Agency, which was already warning about a future water shortfall for homes and farming, recently conceded the rapid expansion of AI had made it impossible to forecast future demand.
Research carried out by Google found that fulfilling a typical prompt entered into its AI assistant Gemini consumed the equivalent of five drops of water – as well as energy equivalent to watching nine seconds of TV.
That may be a small cost for oiling the wheels of a once-in-a-century productivity breakthrough. However, even its most optimistic cheerleaders concede that the economic benefits of generative AI may take quite some time to materialise.
Meanwhile, evidence is mounting of some of the social costs of what amounts to a mass human experiment.
In the UK a paper from the Social Market Foundation thinktank last week warned that the increasing use of chatbots can erode pupils’ cognitive skills – and urged the government to tighten guidelines for schools about their use.
Even more alarmingly, OpenAI recently said it would install “stronger guardrails” after heartbroken US parents claimed their son Adam Raine killed himself after “months of encouragement” from ChatGPT.
Their harrowing story echoes wider concerns about the role of chatbots in mental health crises, and a more diffuse worry about the impact of a tsunami of AI slop on public debate.
None of this is to deny the technology’s potential benefits, from drug discovery to coding.
However, the reality of its energy thirstiness, together with increasingly clear social costs, demands a more clear-eyed look at the trade-offs involved than has so far been evident in Labour’s breathless embrace of the tech giants.