US AI data centres drain power and offer little for UK jobs while in Europe opportunities call says Dave Levy
One of the few substantial outcomes from Donald Trump’s second visit to the UK is the announcement of a UK-US Tech deal, also named the UK-US Tech Prosperity Deal. TechUK states that the key areas for co-operation are Artificial Intelligence, civil nuclear technology, quantum technologies and network technology & cyber-security.
The most eye-catching proposals are to enable the US Big Tech companies to build some very large data centres to host nodes in their AI configurations. The Government press release argues that this will help research significant bioscience advances.
The scale of some of these data centre builds is enormous. The proposed Microsoft/Nscale data centre is planned to be 50 megawatts with a burst capability of 90 megawatts; these metrics do not include the power draw necessary to run the cooling systems. Google Gemini reports that 50 megawatts could be expected to power 30,000 homes.
AI systems and their Data Centres aren’t a public good; the rules-based, i.e. the knowledge is owned by the software company, and the data centres compete with people for power, water, land, and capital. The US corporate domination of the sector also acts as a talent sink.
It is suggested that the deal is worth £31bn of inward investment, but the core IT assets are US-owned and supplied. So these UK-based data centres will be stuffed full of US-manufactured computers.
While they claim that over 5000 jobs will be created and nominate the North East as a location for the new data centres, it doesn’t take very many people to run a data centre and much of the expertise in designing and building them is located abroad.
A final worry is that, in reality, AI has no value. It’s being used to create memes and low-value artefacts. It’s important to understand what makes it an AI solution, as opposed to just more distributed computing complexes. AI would seem to be the layering of heuristic neural networks on top of big data storage and processing systems. The industry has been doing this for a decade or so without calling it AI. It has also been doing complex modelling of weather and physics on supercomputers for even longer.
Given that the outputs of AI systems are based on the captured knowledge, which is current (and popular) today, it remains unclear how AI will innovate. Their protagonists suggest that they can replace labour in white collar work, but this is questionable, and if they do, how will invention occur?
More and more people are suggesting the lack of value means that this is a bubble which Is consuming capital and denying it to other, more potentially worthwhile initiatives. The monopolies are polluting the necessary creative destruction.
At the end of the day, this looks like yet another of Trump’s shakedowns; much of the investment will be spent in the USA, and most of the profits will also be repatriated there.
Another option alluded to in the CNBC article is greater European co-operation. The UK has rejoined the European High Performance Computing Joint Undertaking, and coincidentally reversed Reeves’ 2024 cuts in the supercomputing programme. The EU is now talking about an EU IT stack with which the UK could easily cooperate, and both would increase the UK’s cyber-sovereignty.
The UK’s future is becoming a choice between the USA and the European Union; this treaty is just another piece of untimely Atlanticism, just as is the decision to buy more US fighter planes and their tactical nuclear payload.