Swiss National Bank Warns AI Could Push Up Inflation — Before Pushing It Back Down

For most of the past three years, the conventional wisdom among economists has been that artificial intelligence is deflationary: it automates work, cuts costs, and compresses prices. On August 21, 2026, a senior Swiss National Bank official offered a more complicated picture — one in which AI first pushes prices up before it pushes them down.

Petra Tschudin, a member of the SNB’s Governing Board, told a newspaper interview that artificial intelligence could push up inflation in the short term, although the overall effect of the technology on prices remains uncertain and could run in both directions. Her comments, reported by Reuters and picked up across financial media within hours, mark one of the clearest statements yet from a major central banker that the AI investment boom carries a near-term inflationary sting.

What Tschudin Actually Said

According to the Reuters report, the central bank is looking closely at the impact of AI on prices and believes the technology could have an effect in both directions. The short-term channel is demand-side: an enormous wave of capital expenditure on data centers, chips, and power infrastructure is flowing through the economy right now, bidding up the price of everything from semiconductors to electricity to construction labor.

The long-term channel is supply-side and deflationary: as AI diffuses through the economy and genuinely lifts productivity, the cost of producing goods and services should fall, easing price pressure. Tschudin’s framing — inflationary first, disinflationary later — echoes comments made earlier this year by other European central bankers, including Dutch central bank governor Olaf Sleijpen, who argued that short-term inflationary AI spending will eventually be countered by productivity gains.

Why a Low-Inflation Country Is Worried

Switzerland might seem like an odd place for this warning. Swiss inflation was just 0.6% in May 2026, comfortably inside the SNB’s 0% to 2% target band, and the bank has held its policy rate at 0% since its June 2026 monetary policy assessment. In its June forecasts, the SNB projected average annual inflation of 0.6% for 2026, rising only slightly to 0.6% in 2027 and 0.7% in 2028 — each revised up by just 0.1 percentage point from March.

But Switzerland is precisely the kind of small, open, energy-importing economy where global AI-driven price shocks show up quickly. Swiss data center electricity consumption has already climbed to roughly 2.1 terawatt-hours per year according to the Swiss Federal Office of Energy, and grid operators are warning about overdemand for data center capacity — including in Switzerland itself. When global gas and power prices move because hyperscalers are competing for electrons in Texas and Virginia, Swiss households and importers feel it too.

That is the quiet subtext of Tschudin’s warning: even a central bank with inflation firmly under control cannot fully insulate itself from an AI capex cycle that is now large enough to move global commodity and energy markets.

The Bigger Macro Picture

Tschudin’s comments land amid a broader reassessment of AI’s macroeconomic footprint. Earlier the same day, the LA Times reported that economists had raised their forecasts for third-quarter US economic growth, citing upward adjustments to consumer spending and private investment — much of the latter tied to AI infrastructure. The Atlanta Fed’s GDPNow tracker has been running at around 4% annualized for the third quarter, a pace that would have seemed implausible without the current buildout.

Harvard economist Jason Furman has estimated that AI infrastructure investment accounted for the overwhelming majority of US GDP growth in the first half of 2025, and the St. Louis Fed has been tracking AI’s direct contribution to measured GDP growth for over a year. The uncomfortable implication is that AI spending is no longer a rounding error in macro data — it is, in some quarters, the main event. And spending of that magnitude has a habit of showing up in prices before it shows up in productivity statistics.

The energy channel is the most concrete. US wholesale power prices have risen by as much as 267% in five years in data-center-adjacent regions, and researchers project household electricity bills rising 15% to 40% by 2030 in affected areas. Electricity inflation is notoriously visible and politically salient — the kind of price movement that shifts inflation expectations, which is exactly what central bankers like Tschudin are paid to watch.

Why This Matters for AI Watchers

For the AI industry, central bank attention cuts both ways.

On one hand, it is a sign of arrival. When the SNB — an institution that manages one of the world’s most stable currencies — feels compelled to discuss AI in an interview about price stability, the technology has graduated from tech coverage to monetary policy coverage. AI capex is now a variable in rate-setting models the way housing or oil once was.

On the other hand, it introduces a new kind of policy risk. If AI-driven energy and investment inflation forces central banks to keep rates higher for longer, the cost of capital for the very data centers driving the boom rises too. A negative feedback loop — AI spending lifts rates, higher rates choke AI spending — is the scenario no builder wants, but it is now being discussed openly in Frankfurt, Amsterdam, and Zurich.

There is also a timing problem embedded in Tschudin’s framing. The deflationary payoff of AI depends on productivity gains materializing broadly across the economy, which historically takes years to diffuse. The inflationary costs — energy prices, construction materials, scarce engineering labor — arrive with the construction cranes. Central bankers have to set policy for the economy they have today, not the one productivity statistics might describe in 2029.

The Bottom Line

Tschudin’s warning is not a rate-hike signal — Swiss inflation at 0.6% gives the SNB no reason to move, and markets did not treat her comments as such. It is something more interesting: a first draft of how monetary policy will think about AI over the next decade. The technology’s effect on prices, in the SNB’s telling, is not a simple “deflationary miracle” story. It is a race between a demand shock happening now and a supply shock that has not fully arrived.

Whoever wins that race — the capex boom or the productivity dividend — will determine whether the 2020s end as an era of AI-driven price stability or AI-driven price pressure. On August 21, 2026, one of the world’s most conservative central banks put that question on the record.