Euro area non-food retail volume fell 1.2% in a single month. AI spending across European retail kept climbing. The distance between those two lines is the margin arithmetic every pilot programme must now survive.

Euro area headline retail volumes fell 0.5% month-on-month in December 2025; non-food dropped 1.2%.

January extended the run: headline down another 0.1%, non-food down 0.2%. Year-on-year the volume index still reads 2.0% positive. The monthly direction does not.

The 2026 outlook offers no recovery in volume. Bain & Company forecasts flat-to-negative non-food volume growth across the UK, France, and Germany for the full year. UK non-food turns slightly negative. France manages 1.5% in nominal sales with real volumes flat; Germany posts 2.5% nominal growth on fragile consumer confidence.

Retail AI budgets run on a different ledger. Global AI spending will reach $2.52 trillion in 2026, up 44% year on year, according to Gartner. Much of the increase concentrates in infrastructure rather than application deployment. The same forecast placed AI in its trough of disillusionment — the phase where spending persists and conviction does not.

Europe follows the global trajectory. IDC projects European AI spending at a 33.7% compound annual growth rate through 2029, heading toward $290 billion. Retail ranks third among European sectors in AI investment behind banking and software. The allocation is not slowing.

Adoption confirms the commitment. Honeywell’s 2025 survey of 450 retail executives found 85% had developed AI capabilities and 60% were actively expanding their deployments, with just 1% not yet evaluating the technology. The number flatters: it measures commitment, not return. Budgets are growing into a market that is not.

This is where the maths turns. AI spending at 33.7% compound growth against a flat-volume market means every line item’s share of cost rises automatically. A deployment does not need to fail to become a drag. It only needs to grow faster than the margin it generates.

The pressure falls unevenly. Customer-facing AI that lifts conversion can justify itself when volume is flat. Back-office cost reduction needs no volume at all. Personalisation without attribution and content generation without measurement sit between those poles and will not survive.

Protecting a budget is not the same as earning a return. In a flat-volume environment the cost of protection itself rises quarterly. AI spending that holds its line while revenue stalls demands compounding justification from every deployment underneath it. Boards that shielded the budget will demand proof equal to what they shielded.

Few programmes have produced that proof. Retail Economics and Voyado surveyed 300 European retail leaders and found 95% have trialled AI — but only 5% report clear, scalable returns. The study named skills gaps, internal resistance, and integration complexity among the main barriers. The gap between adoption rate and return rate is where pilot budgets die.

The failure rate is not new. MIT’s GenAI Divide study, reported by CIO, found 95% of enterprise generative AI projects produce no measurable return within six months. That was the pass rate before volume started falling.

The forecast matches the arithmetic. Gartner predicts over 40% of agentic AI projects will be cancelled by the end of 2027, citing escalating cost, unclear business value, and inadequate risk controls. Volume contraction supplies the first two conditions at once.

The cancellation wave will not land evenly. Retailers with production-scale AI and margin data will absorb the budgets that others release. The gap between leaders and laggards widens precisely because the laggards were still in pilot when volume turned. Selection pressure compounds advantage.

A pilot without a margin number is a cost centre with a presentation deck.

What survives is production work with receipts. Deployments in search and demand forecasting keep their line items because they produce a number a CFO can read without interpretation. As we reported this week, AI-powered search remains the one budget category that answers the ROI question without rehearsal. The rest auditions in a room that has stopped clapping.

European retailers chose to protect AI spending through a volume decline. The market offered no reciprocal protection. Every deployment without proven margin is now borrowing time — from the deployments that have it.