NIESR: UK Q3 growth fits early AI pattern, but gilt yields and trade are a drag
UK GDP grew by 0.7 per cent in the third quarter, the second-fastest rate in the G7 behind the US, according to analysis from the National Institute of Economic and Social Research (NIESR). It was the fastest quarterly growth since the end of 2017, excluding the post-pandemic recovery. NIESR examined whether this marks the start of an AI-driven boom, as some in government have claimed.
Investment points to AI, but the evidence is circumstantial
Investment contributed around half of Q3 GDP growth. Spending is concentrated in ICT equipment and buildings, which NIESR says is consistent with a boost from AI infrastructure, though it stresses this is circumstantial evidence.
The benefit to UK output depends on how much of the new infrastructure is imported rather than made at home. Net trade was a drag on GDP in the latest data, and NIESR expects that to continue. It also notes that AI is hard to measure, so the figures should be treated with caution.
Which sectors are growing
Growth has picked up in industries likely to supply AI services, a grouping defined by the Bank of England. Sectors likely to adopt AI early, including financial and business services, have also strengthened this year. Together they account for around three-fifths of the rise in growth over the year. NIESR says this does not measure AI's contribution, only that growth is coming from where AI gains would be expected first.
Data from the ONS Business Insights and Conditions Survey show the ICT sector leading on adoption, with 60 per cent of firms using AI, up from 30 per cent when tracking began in late 2023. Professional services follow at two-fifths of firms. Fewer than 20 per cent of firms in hospitality and construction report using AI. Output declines over the past year have been concentrated in sectors where fewer than one in five firms use AI.
NIESR also points to an "AI J-curve": firms may see measured productivity dip at first while they invest in training, process redesign and organisational change, with gains following later.
Higher borrowing costs a risk
A Brookings paper found around 57 GW of US data-centre capacity in place by mid-year against a pipeline of more than 500 GW, although NIESR doubts all of it will be built. On that basis the buildout could be only a tenth complete. The paper estimates delivering it would take investment of nearly 4 per cent of US GDP.
NIESR's NiGEM model suggests the UK would gain some exports to the US, but only modestly, as it is not a major player in data-centre supply chains. The larger effect would be higher long-term interest rates as global demand for capital rises. NIESR estimates around a third of the rise in gilt yields since February can be attributed to AI investment demand. More upward pressure would tighten financial conditions and dampen UK activity, and NIESR says the Chancellor should not count on lower borrowing costs to help the public finances in the near term.