3 hours ago
6
June 10, 2026 — 12:11pm
Kevin Warsh, who’s just chairing his first rate-setting meeting at the Federal Reserve Board, argues that the boom in artificial intelligence will allow significant cuts to US interest rates. His new colleagues, and plenty of others, aren’t so sure.
Warsh, appointed by Donald Trump with an expectation that he would try to deliver the rate cuts Trump has consistently, and insistently, demanded, has argued that AI will prove “structurally disinflationary” and that, instead of reacting to the data of the moment, the US central bank should make a forward-looking judgement that AI will lower the inflation rate.
He has argued that AI will unleash a productivity boom – “the most productivity-enhancing wave of our lifetimes” – that will enable non-inflationary growth and allow lower interest rates.
Other Fed officials beg to disagree.
The Fed’s vice chairman, Philip Jefferson, said earlier this year that “all other things being equal, persistent increases in productivity growth are likely to result in an increase in the neutral rate, at least temporarily.” (The neutral rate is the rate of interest that neither constrains nor stimulates growth, allowing the economy to expand while maximising growth with a stable inflation rate).
When the Fed concludes this week’s Federal Open Market Committee meeting, it is most unlikely that Trump will get the rate cut he so desperately wants.
Another Fed governor, Michael Barr, has also linked higher productivity growth with higher interest rates, saying AI was unlikely to be a reason to lower the Fed’s policy rate. Meanwhile, the San Francisco Fed president Mary Daly has said an AI-driven acceleration of productivity growth would dictate a higher neutral rate because the demand for investment would rise faster than the supply of savings.
There’s little debate over whether AI will lift productivity rates, but there’s significant disagreement among economists about how that will play out and what it might mean for inflation and interest rates. The weight of opinion appears to be on the side of higher inflation and higher interest rates.
The argument favouring a higher neutral rate is relatively straightforward.
Whether or not AI results in higher productivity, what really matters for inflation and interest rates is whether it raises the rate of productivity faster than the costs of deploying it.
We are in the early phase of AI deployment. The companies developing the technology and its applications are investing unprecedented amounts, at a dramatically accelerating rate, in training their models and building the infrastructure required to provide the necessary computing power.
Last year, investment in AI in the US was about $US375 billion ($533 billion). This year, it is expected to be about $US750 billion. It may top $US1 trillion next year. Goldman Sachs has said $US7.6 trillion could be spent on AI over the next five years.
Most of that spending is on computing power and data centres, which also require investment in energy grids and water supply.
That has flow-on effects, from the rising cost of semiconductors, with a significant shortage in face of the demand producing extraordinary price hikes, to higher prices for commodities like copper and rising consumer costs for energy and anything electronic. It also raises the cost of labour with AI knowledge and skills.
With AI companies competing aggressively for capital – equity and, increasingly, debt – the cost of capital more broadly should rise. A material increase in the yields on 10 and 30-year US Treasury bonds since OpenAI launched ChatGPT in November 2022 may reflect some of that AI influence.
Whatever the eventual impact of AI might be on productivity and inflation, during the transition from here to there it will be inflationary because the costs of rolling out and implementing AI will inevitably rise – during this phase at least – faster than the rate of any productivity gains. That’s what economists call a “productivity J-curve.”
Economists would also say that, even if AI did materially lift productivity, returns on capital and economic growth, it would likely reduce savings rates and push the neutral rate higher, not lower.
It may already be doing that, even if the Fed’s current policy stance and a policy rate that was cut three times last year doesn’t yet reflect it. A neutral rate above the policy rate (currently the target range for the federal funds rate is 3.5 per cent to 3.75 per cent) would mean that the monetary policy settings were stimulatory.
With an inflation rate of 3.8 per cent and rising (thanks to Trump’s tariffs and the war in the Middle East), the Fed shouldn’t be stimulating the economy.
When the Fed concludes this week’s Federal Open Market Committee (FOMC) meeting, it is most unlikely that Trump will get the rate cut he so desperately wants. Indeed, it is more likely that the Fed will signal that the next rate move could be an increase.
The reality that there is likely to be a years-long transition phase before the assumed productivity benefits of the massive investments in AI emerge to a meaningful degree, and that that phase is likely to be inflationary, means that Warsh’s theory on productivity-driven disinflation is unlikely to be tested any time soon.
Even if he wanted to push the case, his colleagues at the Fed would be unlikely to co-operate, and he’s only got one vote on the FOMC.
The risk for the Fed is that it falls behind the curve in tightening monetary policy in line with the probable increase in the neutral rate.
There is, however, a Catch-22 element to monetary policy in the circumstances.
The demand for capital to fund AI developments and the associated infrastructure is vast and growing. It has been satisfied, so far, by the investor appetite for all things AI.
The proposed initial public offerings by the big three AI start-ups – SpaceX, Anthropic and OpenAI – will amount to more than $US200 billion and value them at more than $US4 trillion.
Even the mega-techs like Google’s parent company Alphabet, Amazon, Meta and Microsoft are finding that their spending on AI is soaking up their cashflows, forcing them to raise debt and, in some recent instances, equity.
That’s occurring at a time when the US government, like most developed-economy governments around the world, is awash with debt and adding to that total daily.
US gross government debt will hit $US40 trillion by September – about 125 per cent of GDP – with budget deficits of 6 per cent of GDP ensuring that total debt levels will continue to climb.
The US savings rate has been dropping rapidly, from 3.7 per cent of disposable personal income in the first quarter of this year to 2.6 per cent in April, with households squeezed by the rising costs of living flowing from Trump’s tariffs and the Iran war.
Funding the massive demands for capital from AI companies is doable while rates are relatively low and equity prices (driven by the frenzy around AI) are at record levels.
If interest rates were to rise to counter the rising inflation – and the futures markets have begun pencilling in a rate hike in December – it would increase the costs of capital for the AI companies and those building out the infrastructure of data centres and their energy and water requirements. And potentially, if more rate rises were seen as likely, it could puncture the sharemarket and the cycle of rapid and substantial increases in value that the AI companies have relied on to gain their access to capital.
If the Fed were forced to raise interest rates to counter AI-driven inflation and end the sharemarket boom, the arguments about productivity and interest rates might, at least in the short term, remain academic.
The Business Briefing newsletter delivers major stories, exclusive coverage and expert opinion. Sign up to get it every weekday morning.
