At the end of yet another dinner party dominated by AI chatter, I often ask myself whether other important topics have been crowded out. Applied to the economy more broadly, it’s also worth asking whether AI mania is sucking resources from other sectors. Data centres and model development require vast quantities of chips, electricity and financing. Does the craze mean others are going without?
Over the long term, this will hopefully seem like a silly question. When AI has solved cancer and climate change, no one will be sniping about whether too much money and electricity were poured into the industry. There is no fixed lump of credit. Historically, innovation has made investment worthwhile, like when electrification made it viable to use a new class of small factory machines.
Here and now, though, constraints can bite. I wouldn’t advise running energy-intensive manufacturing near a data centre, or stuffing cheap phones with the memory chips that make AI run. Financing should be more elastic, and will need to be, as the big tech companies turn to debt markets to help with the AI build-out. But other borrowers hungry for cheap credit could still face a squeeze.
Starting with financing flows, eyeballing the raw data, AI-related companies do seem to be eating up a growing share. Silicon Valley Bank reported that over the past 12 months AI companies have received almost two-thirds of US venture capital dollars. And according to Goldman Sachs, roughly a quarter of US gross investment-grade debt issuance this year has come from AI-related companies, including hyperscalers, software providers and data sector funders.
Also in America, the real economy is showing signs of the AI boom too, reinforcing fears that the less whizzy bits are being starved of resources. As spending on capital in AI-related areas including information-processing equipment and data centres rose over last year, other types of private investment fell. Within office construction, spending on data centres has soared while the rest has sagged.
While suggestive, neither of the trends in financing nor investment proves that AI is sucking anything away from other sectors. The buzz could be stimulating capital. In a recent report, Silicon Valley Bank described how, although higher Treasury yields normally would curb VC investment, more recently those ties have been severed “as the fear of missing out on this step change . . . has pushed VC investment to new heights.”
As for the real investment trends, the AI boom could be masking underlying economic weakness caused by something else entirely. Although some have warned that AI is imperilling a broader revival in US manufacturing, other more obvious culprits include government subsidies running out, tariff uncertainty and tight monetary policy in response to sticky inflation (not itself caused by AI).
International comparisons can offer some reassurance. If AI were simply reallocating a fixed pot of capital spending within countries, one would expect to see no relationship between AI mania and overall investment levels. But researchers at the Bank for International Settlements found that countries more strongly orientated towards AI (or more “prepared”) saw stronger growth in both business and total investment over the two years up to the end of 2025. That suggests that AI has been delivering a real boost.
If crowding out is happening within credit markets, it probably isn’t massive. A recent note from Jessica Rindels and David Mericle of Goldman Sachs argues that non-AI companies haven’t experienced a noticeable uptick in their relative cost of debt. Their financing costs probably are slightly higher overall because of the increase in competition for financing from hyperscalers and their pals. Based on estimates of the effects of quantitative tightening, a restriction in the credit supply, Rindels and Mericle guesstimated a 0.05 percentage point effect.
Signs could also emerge in sectors beyond the corporate world. AI-related borrowers have been demanding long-term funds, of the sort governments typically issue. Just this week, analysts were citing the massive bond issuance to fund the AI build-out as being behind the punishing pricing in long-term sovereign debt markets.
As the AI boom progresses, signs of crowding out could become more obvious. By one estimate, for example, AI investment in the US will rise from 1.8 to 2.8 per cent of GDP between this year and 2028. Perhaps there will be a lot more dinner parties to discuss it all — and perhaps those wanting to talk about something other than AI will have to shout a little louder to be heard.