Panetta hangs euro-area rate policy on whether AI demand arrives before productivity

The two channels do not arrive at the same time. Spending on computing infrastructure and on the energy to run data centres lands as demand while it is being spent, and can lift relative prices until supply catches up ^[8]. The productivity comes later, after the capacity is built and used, and Panetta said the timeline and scale of those gains remain "highly uncertain" ^[6]. On his own sequencing question, then, the demand side starts ahead. I would hold that view loosely. Neither account of his remarks puts a figure on euro-area AI investment ^[19], and if a large share of the spending is imported hardware, the euro-area demand impulse is smaller than the headline spend.

The disinflationary case does come with a number. The Bank of Italy study he cited put widespread adoption at more than 1 percentage point a year on Italian labour productivity growth ^[7]. One point a year compounds to roughly 5.1% more output per hour over five years ^[17]. Against a 2% target ^[11], a gain of that size would drag measured inflation lower for years, provided wages did not rise to match.

The valuation warning is the same question seen from the other end. Panetta said technology valuations reflect "overly optimistic expectations" about AI profitability ^[9]. The gap between those expectations and near-term profits leaves equity prices open to sharp corrections, he said, citing recent fluctuations in tech prices ^[10]. Take one more step, which the sources do not take: if the data-centre and energy capex is being committed on the strength of those expectations, an equity correction and a shrinking demand impulse are one event. A selloff would then argue for cuts.

Cryptobriefing's second account places Panetta on the ECB's Executive Board and reports the valuation warning as remarks to Bloomberg Economics ^[13]. Its first puts him at the Bank of Italy and on the Governing Council, speaking in Kyiv on September 21 ^[1][2]. The same publisher's market evidence is thinner than it reads. It puts the probability of Anthropic reaching a $600 billion valuation by December 31 at 1.8% ^[14] and says related thresholds lost confidence following his remarks ^[15]. What it does not show is that a central banker's speech is what moved it. At 1.8%, the market is about 55 to 1 against ^[16], pricing that moves on rounding.

What would falsify the demand-first case is euro-area inflation drifting under target while data-centre investment keeps running. That would say either that productivity is landing first or that the spending is not touching euro-area prices. Panetta's framing supplies the other test himself: governments that accelerate adoption through subsidies and deregulation would amplify the demand channel before productivity materialises, and make the ECB's job harder ^[18].

Cryptobriefing sets out the two scenarios this way: demand-first means holding rates higher for longer than markets expect, and automation-first means displacement weakening consumption, inflation below 2%, and an easing cycle ^[12].