- The five major AI hyperscalers (Alphabet, Amazon, Meta, Microsoft, Oracle) are on track to spend $741 billion on capex this year — up nearly 75% YoY — with total AI build-out spending through 2032 potentially reaching $8 trillion, per Columbia economist Stijn Van Nieuwerburgh.
- The inflation is already arriving: consumer prices for computer software and accessories rose ~15% YoY in May; wholesale electronic components jumped 27% YoY; electrical contractor wages are up 6.5%; Goldman Sachs forecasts electricity prices rising ~6% annually in 2025-26.
- Unlike tariffs or oil — both one-time shocks — AI infrastructure demand could persist for years; 81% of NABE economists surveyed say the AI build-out will add to inflation over the next year.
- The long-run productivity offset exists but is distant: UBS estimates it will take at least a couple of years before AI productivity gains begin lowering inflation — and the May PCE reading due Thursday is expected to show 4.1% YoY, more than double the Fed’s 2% target.
What Happened?
As tariff pressures ease and Iranian oil flows resume, a third inflation driver is taking shape: the AI infrastructure buildout. Hyperscaler capex is running at $741 billion this year — nearly 75% above last year — and analysts estimate total AI-related capital spending through 2032 could approach $8 trillion. That spending is intensely physical: data centers require memory chips, cooling systems, fiber and electrical cables, backup generators, and construction labor. Because those inputs are used across the entire economy, AI demand is creating price spillovers well beyond tech. Apple CEO Tim Cook has warned of chip cost increases “unlike anything he’d seen in over 40 years.” Nintendo, Microsoft, and Sony have already raised device prices. Goldman Sachs forecasts data centers will account for nearly half of US power demand growth through 2030, driving consumer electricity prices up ~6% annually.
Why It Matters?
The inflationary character of AI spending is categorically different from tariffs or energy shocks. Both of those are one-time level shocks that eventually work through the price system and fade. AI capex is a sustained demand shock that scales up over multiple years — and, critically, only a small fraction of announced spending has actually been deployed yet, per Fed Governor Lisa Cook. The Fed’s preferred inflation gauge is expected to print at 4.1% year-over-year for May (3.4% core), due Thursday — well above the 2% target the Fed hasn’t hit since early 2021. Each new inflation driver — pandemic reopening, tariffs, Iran/energy, now AI — has extended the timeline for returning to target. Fed Chair Warsh has previously argued AI will ultimately be “a significant disinflationary force” — the current data is a direct test of that thesis.
What’s Next?
Thursday’s PCE print is the immediate catalyst. The structural question is whether AI productivity gains arrive in time to offset AI-driven input inflation. UBS puts the productivity payoff at least two years out. In the interim, the Fed faces a genuine dilemma: hiking to fight AI-driven inflation risks choking off the investment that could eventually be disinflationary. The broader risk, flagged by Berkeley economist Jón Steinsson, is expectation de-anchoring: if inflation stays elevated long enough through a series of overlapping “temporary” shocks, households and businesses may stop expecting it to return to 2% — which itself becomes inflationary.
Source: The Wall Street Journal
