ECB finds euro area firms funding AI with bonds and their own cash

Europe’s companies are spending on artificial intelligence, but they are not asking their banks to cover the bill.

New analysis from the European Central Bank shows euro area firms leaning on market-based financing, especially debt securities, to fund AI. According to the ECB, that shift may reduce their sensitivity to interest rate changes, which is a quietly big deal for an institution whose main tool is setting those rates.

The findings come from the ECB’s Survey on the Access to Finance of Enterprises, known as SAFE. It is the central bank’s regular check-in on how businesses raise money. This time, AI got its own spotlight.

Lots of experimenting, not much commitment

Approximately 70% of euro area firms say they use AI in some form as of 2026. Only 7% describe their usage as significant or intensive.

Planned AI investment is expected to make up about 9-10% of total firm investment for 2026.

Technologies and tools account for 49% of AI investment budgets. Employee training takes 46%.

Cash first, bonds second, banks fading

Approximately 72% of firms planning AI investments are expected to pay for them with internal funds. That means cash flow and retained earnings, not loans or outside investors.

Among high AI-active firms, debt securities issuance rose by 13% as of January 2026. Meanwhile, these firms’ reliance on traditional bank loans has started to decline.

AI-related borrowing made up about a quarter of credit growth to firms in the first quarter of 2026, according to the ECB research.

The hardware problem

Firms report it is easier to secure outside financing for tangible AI assets, such as hardware. Intangible assets are a harder sell.

A server rack can be repossessed and resold. A fine-tuned model, a retrained workforce or a redesigned internal process cannot be easily put up as collateral.

Much of AI’s value lives in those intangibles. 46% of budgets goes to training, which is about as intangible as investment gets.

A patchwork map of adoption

AI uptake is uneven across the currency bloc. Usage is highest in the Netherlands, Finland and Austria. It is lower in Italy and Ireland.

Previous SAFE modules found no evidence of aggregate job losses linked to AI adoption.

What this means for the ECB and for markets

If AI investment is mostly funded from retained earnings, and the most active adopters are moving toward bond markets, the monetary policy transmission channel weakens for this category of spending. The ECB’s own framing is that this may reduce firms’ sensitivity to rate changes.

For bond investors, the 13% rise in debt securities issuance among high AI-active firms signals a new source of corporate supply tied to technology spending. Investors buying that paper will need to judge how much of the underlying investment sits in tangible hardware versus harder-to-value intangibles.

For banks, the trend is a mild warning. The companies spending most aggressively on AI are the ones drifting away from loans.

The gap between 70% experimenting and 7% committed is the number to track. If that intensive group grows, the shift toward market-based funding, and its implications for how monetary policy reaches the real economy, could grow with it.