Kevin Warsh focused his first conference as Federal Reserve chair on June 17 on a familiar problem: fast-rising prices fuelled by the Iran war.
Yet another challenge facing central banks goes beyond the immediate inflation shock. As investors place ever bigger bets on artificial intelligence and politicians tout its economic promise, policymakers face a question that would have sounded absurd a few years ago: should interest rate decisions today be influenced by productivity gains that may not materialise for another decade or more?
The US central bank appears to be in no mood to bet on future productivity gains. Mr Warsh signalled that interest rates may need to rise further to fight US inflation that rose to 4.2 per cent in May, prompting investors to scale back expectations of lower borrowing costs. This is not just a US shift.
Across 52 economies tracked by Bloomberg, the number of central banks raising rates matched the number cutting them in May, ending a two-year period in which easing dominated. Recent rate increases by the European Central Bank and the Bank of Japan suggest policymakers are becoming more cautious about declaring victory over inflation.
AI's role in monetary policy
Yet the debate over how AI should influence monetary policy has not gone away. Before taking the Fed’s top job, Mr Warsh argued that AI could become a powerful disinflationary force, echoing former Fed chair Alan Greenspan’s 1990s productivity bet to argue for lower rates.
Similar arguments have emanated from the White House. President Donald Trump’s economic adviser Kevin Hassett has argued that AI-driven productivity gains will put downward pressure on inflation and give the Fed more room to lower interest rates. Also, investors have poured capital into AI-related companies on the assumption that the technology will eventually deliver big productivity gains and speed economic growth.
The question is whether policymakers can afford to act on that expectation before the gains themselves become visible. That argument rests on an assumption that remains deeply uncertain. Expectations about AI are running far ahead of the evidence. Productivity gains are only just beginning to appear, with US output per hour rising by roughly 2.2 per cent last year. That is hardly at a scale that justifies cutting rates today.
As with previous general purpose technologies, the bigger gains depend on changes to business models and organisations, which take time. This means central banks risk treating future productivity as current productivity.
But for now, hawkish signals are emerging from some of those banks. The Bank of England left rates unchanged at 3.75 per cent last week, but a split vote highlighted continued concern that higher energy costs could become embedded in the wider economy.
The disinflationary case for cutting rates is weaker than it appears. And even if that uncertainty fades, another contradiction is that policymakers are mispricing the neutral rate. If AI ultimately delivers the productivity gains promised, the underlying rate of interest should be rising over time, not falling, as higher returns on capital push up the equilibrium cost of money.
There are other factors at play. Higher inflation is not just a risk but potentially an incentive, offering policymakers a way to erode the US debt burden, with gross federal debt now above 120 per cent of GDP. In the 1940s, inflation helped cut the debt ratio by more than 30 percentage points within a decade from levels above 100 per cent of GDP at the end of the second world war.
The irony
The irony is that cutting too soon could backfire even in the markets. Long-term borrowing costs, which are the rates that actually set mortgages and corporate debt, depend less on today’s policy rate than on the bond market’s faith that inflation will be contained.
Ease prematurely and that faith erodes: investors demand a higher inflation premium, long yields rise and borrowing costs climb even as the central bank is ostensibly loosening.
Policymakers would be pressing the accelerator only to find the bond market pumping the brakes.
For businesses, cheap money in this environment can be a trap rather than a gift. If AI genuinely raises the return on capital, the equilibrium interest rate should be rising; so, financing that assumes the opposite is mispriced from the start.
At the same time, a war premium on oil injects volatility into bond, currency and commodity markets, and pulls capital out of energy-importing emerging economies towards the safety of the dollar. None of this is the backdrop of calm in which a central bank can comfortably bet on disinflation.
This leaves central banks making policy against two conflicting signals: near-term inflation driven by energy and demand, and a longer-term productivity story that remains incomplete.
They will need to judge whether AI’s promised productivity gains are real, and when they will arrive. And that decision will matter far more than any single interest-rate move this year.
Karl Schmedders is professor of finance at IMD
