Europe's AI dividend arrives as equipment exports while its own build costs rise

The Commission credits AI investment abroad for lifting EU exports while cutting investment growth and flagging dearer energy and credit. Europe is selling to the AI buildout more than hosting it.

Power & GridThe Current

Key argument

The Autumn Forecast shows Europe earning from AI mainly as an equipment supplier to builders elsewhere, while the energy and borrowing costs that decide domestic data-center economics moved against EU projects this autumn.

At the same press conference, the European Commission credited Artificial Intelligence for part of the EU's surprise growth and warned that tighter credit would hold back investment at home. Commissioner Dombrovskis said strong AI-related investment is lifting demand for EU exports, with the bloc benefiting as "a significant global supplier of power, cooling and chip-making equipment." In the same remarks, the Commission put EU investment growth at 1.5% this year, down from 2.6% last year, with a slow climb to 1.6% in 2027 and 2.0% in 2028. Energy is the main headwind in the Autumn Forecast. Brent passed 100 dollars in October, EU gas storage sat at 75% of capacity at the end of that month, the lowest seasonal level since records began in 2011, and euro area retail energy prices rose almost 21% year-on-year in October, according to the Commission.

My reading is that the forecast casts Europe as a vendor to the AI buildout rather than a venue for it, and that the two inputs a European data-center project lives on, delivered energy and the cost of debt, turned against builders this autumn. The evidence boundary is wide here. The remarks are pitched at the whole economy. They never mention data centers, give no level for wholesale electricity, and do not break investment down by asset type. Applying their energy and financing figures to an AI campus is my inference, not something the Commission found, and the essay should be read with that gap in view.

What the remarks actually establish

Start with what is stated rather than implied. The upside surprise in the first three quarters came largely from a rebound in exports, Dombrovskis said, while consumption tracked expectations and investment came in slightly weaker. That export strength is tied to AI-related investment, and global goods trade grew robustly in the first eight months of 2026, led by Asia's advanced and emerging economies, per the Commission. It also calls the export lift temporary. That word matters. The AI channel into EU growth, as described, runs through orders from builders elsewhere, and the institution forecasting it does not expect it to last.

On financing, the remarks describe a turn. The ECB and other EU central banks have started a new tightening path. Bank lending to corporates is still firm, but banks report tighter credit standards and weaker demand, and the Commission expects further tightening over the forecast horizon to slow lending and, with it, investment. The 10-year German Bund yield peaked at 3.65% at the end of September and eased to close to 3.3% by the 29 October cut-off, which the Commission notes is still 30 basis points above its spring level.

On energy, the picture is about duration rather than a spike. Futures point to a gradual fall in oil, but to levels above prewar prices, because damaged infrastructure means supply will not fully recover even if the Strait of Hormuz reopens. Diesel rose more than Brent because of refining bottlenecks. Wholesale electricity surged above expectations in spring on gas prices and weather shocks, though the speech attaches no figure to that surge, so the size of any hit to a large-load power contract cannot be measured from it. Demand for European equipment is evidenced. Demand for European capacity is not addressed.

Equipment sellers and site developers face opposite signs

The same shock pays one group and charges another. A maker of switchgear, cooling plant or chip-making tools in the EU is booking demand created by capital spent mostly outside the bloc, if the trade description holds. Its revenue line benefits from the AI cycle without carrying the cost of building and powering a campus in Europe. A developer trying to put AI capacity on European soil faces the other side of the ledger: power priced off gas that the Commission expects to stay dearer for longer, and debt priced off sovereign curves that have moved up.

The fiscal section sharpens this, though here I am reasoning past the text. The Commission projects EU interest expenditure rising from 1.9% of GDP to 2.4% by 2028, deficits widening, and new spending to shelter households and firms from energy prices. Governments carrying that load, alongside higher defence outlays, would plausibly have less room to discount power for large new industrial loads. The remarks say nothing about data-center incentives, so that is an inference about fiscal space, not a reported policy change. The diesel point is narrower but concrete: backup generation runs on refined fuel, and the Commission says refined products rose faster than crude. That is a cost on testing and outages, not on baseload, and it should be sized that way.

None of this proves European AI capacity is uneconomic. It shows that the evidenced AI benefit and the evidenced cost pressure land on different balance sheets.

A screening rule for European capacity bets

What follows is analysis, not something the Commission proposes. A lender or investor looking at an EU AI capacity project underwritten in this window should treat two items as gating rather than as assumptions. First, the power contract: a fixed or hedged price with a stated term, rather than exposure to a wholesale market that the Commission says surprised to the upside in spring. Second, the debt: fixed-rate or hedged financing sized against yields that are, by the Commission's numbers, above spring levels even after easing. Projects without both are carrying the forecast's headwinds directly.

For equipment exporters, the rule runs the other way. Their question is concentration and duration of offshore orders, because the Commission labels the export lift temporary. An order book weighted to a handful of foreign AI builders is a cyclical exposure, however strong this year's number looks.

The practical discipline is to keep the two stories in separate models. Price the vendor story and the venue story apart.

The case that the energy shock is already fading

The strongest objection comes from the Commission's own text. It says market moves after the 29 October cut-off, if sustained, could lower inflation projections and add some growth. Gas futures point to swift declines in spring next year and again in 2028. The Bund yield had already slipped from its 3.65% peak to close to 3.3% by the cut-off. Bank lending is described as holding firm despite higher funding costs, and the Commission lists faster productivity gains from AI adoption as an upside risk. A developer signing today for capacity that energizes years out cares more about the 2028 price curve than about October retail energy inflation, and on that horizon the forecast itself sees inflation falling back toward target and investment growth recovering to 2.0%.

That case narrows my thesis rather than overturning it. It concedes that the cost pressure may be a timing problem for projects contracting power and debt now, not a structural verdict on Europe as a host. I still weigh it less, for two reasons stated in the remarks. The Commission expects oil to stay above prewar levels because damaged supply will not fully return, and it pushes the euro area's return to target out to 2028. A shock that the forecaster expects to persist for most of its horizon is a pricing input for any project financed inside that horizon.

Signals that would move Europe from vendor to venue

The speech gives no date for the next forecast update, so these are conditions rather than deadlines.

The first is the investment line against the export line. If a later forecast revises EU investment growth upward from 1.5% while the export boost fades as the Commission expects, domestic building is absorbing the energy and credit costs and Europe is becoming a venue. If investment stays near that level as exports normalize, the AI dividend was a vendor dividend only.

The second is the spring energy path. If gas prices fall as swiftly as the futures cited by the Commission imply, and storage rebuilds from the 75% low, the energy half of my argument weakens and power contracts signed now will look expensive in hindsight. If the decline stalls because damaged supply stays offline, fixed-price power becomes the scarce input for any European AI project.

The third is credit. The Commission expects tighter financial conditions to slow bank lending over the horizon. If lending to corporates does slow, refinancing risk is real and unhedged projects should be marked down. If lending stays firm while yields ease, financing is a cost to absorb rather than a gate, and the case against European capacity rests on energy alone.

Sources

The Current is written by Josh Jiwoon Inn with AI assistance in drafting. It does not claim interviews or first-hand reporting, and every number is attributed to a named source. See the AI-assisted Disclosure and Methodology.