Country-by-country: where European agentic commerce activity is concentrating in 2026, identifying five leading

Agentic commerce, as an operational category, refers to the execution of purchase, payment, and fulfilment decisions by software agents acting under delegated authority from a human principal, without requiring transaction-by-transaction human confirmation. The agent receives a mandate, accesses payment credentials or tokenised payment instruments, selects counterparties, initiates transfers, and closes loops across multiple rails and jurisdictions, often in sub-second sequences.

Geographic clustering in this infrastructure follows the logic described for fintech more broadly: concentration occurs where the cost of accessing the enabling layer is lowest and where the regulatory cost of operating at the boundary of existing authorisation categories is contained [1]. For agentic commerce specifically, three enabling conditions must be co-present: a live open banking or instant-payment rail that exposes programmable payment initiation through standardised APIs; an identity and authentication framework that permits non-human principals to satisfy strong customer authentication requirements; and a regulatory environment that has either explicitly accommodated, or chosen not to penalise, automated delegated payment authority. Jurisdictions satisfying all three conditions attract disproportionate investment and serve as the registration and operational home for platforms that then transact across the wider EU single market [2].

Regulatory sandboxes as filtering mechanisms. A regulatory sandbox does not simply reduce compliance risk; it provides a documented, time-bounded permission to operate outside the standard authorisation perimeter while generating a supervisory record. For agentic payment systems, this matters because the existing PSD2 authorisation categories were designed for human-initiated payment instructions. An agent acting under delegated payment authority sits awkwardly across the payment initiation service provider and account information service provider categories, and potentially outside both. Jurisdictions with sandboxes that have admitted agentic AI participants gain three structural advantages over those with only general fintech sandboxes: operators obtain precedent-setting supervisory guidance; regulators develop the institutional knowledge to write agentic-specific rules; and the sandbox cohort itself becomes a co-location incentive for infrastructure suppliers, legal counsel, and institutional investors who need proximity to the regulatory frontier [5].

Payments infrastructure as a rate-limiting constraint. The platformisation of financial transactions [3] is the prerequisite on which all agentic payment constructs rest. An autonomous payment agent requires, at minimum, a payment initiation API that accepts machine-originated instructions, a real-time settlement rail that confirms finality within the agent's decision loop, and a tokenisation or credential-delegation mechanism that allows the agent to act on behalf of a human principal without transmitting primary account credentials. These are identified architectural requirements rather than settled regulatory categories: whether a given agent configuration satisfies PSD2's PISP definition, and on what terms machine-to-machine strong customer authentication exemptions might apply, remains unresolved in most jurisdictions. What is resolved is that higher per-transaction friction for agents compounds as agent transaction volumes scale. The consequence is that infrastructure maturity, measured by API uptime, error-rate standardisation, and settlement latency, concentrates agentic commerce activity in the jurisdictions that moved earliest from compliance to operational quality.

Industrial verticals as demand anchors. Regulatory permission and payment rails are necessary conditions; they are not sufficient to generate cluster density. The demand signal that fills available infrastructure capacity comes from industrial verticals whose existing transaction patterns map directly onto agentic automation. Multimodal mobility payments, where a single journey triggers fare calculation, seat reservation, carbon-offset purchase, and expense reconciliation across multiple operators, constitute a structural fit with agentic orchestration. Logistics procurement, retail omnichannel fulfilment, and high-frequency B2B supply-chain settlement carry the same property. Clusters form where a nationally significant industrial vertical generates enough agentic-compatible transaction volume to justify the fixed cost of building and operating the bridging layer between existing instant-payment rails and agentic-native payment constructs.

Market access becomes jurisdictionally gated. A platform registering its agentic payment services in a sandbox-forward jurisdiction gains supervisory clarity in that jurisdiction and, through EU passporting, the right to operate across the single market from a single authorisation. The consequence is that the competitive geography of agentic commerce does not map onto where transactions occur; it maps onto where the platform is registered and where its supervisory relationship is maintained. Platforms domiciled in the Baltic corridor or Dublin can execute agentic transactions in markets whose own regulators have not yet issued agentic-specific guidance, importing the permissive interpretation of the home regulator [2]. This creates a structural asymmetry: jurisdictions that delay sandbox formation or agentic authorisation precedent do not protect their markets from agentic commerce activity; they redirect the registration and tax base to the jurisdictions that acted earlier.

Regulatory arbitrage acquires a specific operational form. The arbitrage available in agentic commerce is narrower and more technical than the broad-category regulatory shopping that characterised earlier fintech waves. It consists of the selection of a home regulator that has documented, through sandbox interaction, a position on delegated payment authority, machine-to-machine strong customer authentication exemption, and multi-agent liability attribution. Each of these three technical questions carries direct consequences for platform operating costs. A jurisdiction with published supervisory guidance on all three reduces per-transaction compliance overhead substantially relative to a jurisdiction where all three remain unaddressed. The cluster jurisdictions identified in this brief each offer resolution on at least one of the three; none currently offers authoritative resolution on all three.

Identity binding as a fourth concentration factor. The eIDAS 2.0 digital wallet rollout timeline introduces a compounding structural condition that extends beyond the three drivers examined in the mechanism section. Agentic platforms require identity binding at the moment of delegated authority grant; without a machine-readable, cross-border identity assertion that satisfies the relying party's authentication requirement, the agent cannot execute payment instructions in a foreign jurisdiction's banking environment. Clusters with accelerated wallet issuance rates, specifically Estonia's existing digital identity infrastructure and the Netherlands' advanced eIDAS 2.0 pilot posture, gain an identity-layer advantage that compounds the payment-rail and sandbox advantages already in place.