Build, Buy or Partner: The Bank Decision Most Get Wrong

When UniCredit closed its €370 million acquisition of Aion Bank and Vodeno earlier this year, the press read it as another European bank doing what European banks do, buying a piece of fintech because the in-house roadmap could not deliver it fast enough. The more interesting reading is structural. Bain's 2026 Banking M&A report frames the moment as a "double helix" of scale and scope, with capability-driven deals delivering roughly 30% better valuation gains than scale-only transactions. IDC, looking at the technology layer, projects that 40% of global banks will run a sidecar core in 2026, rising to 70–80% by 2028. Banking M&A hit $212 billion in 2025, with the largest share now flowing toward capability

acquisitions rather than market consolidation.

Underneath the deal flow is a question every innovation lead, head of strategy and CIO inside a financial institution is already wrestling with: build, buy or partner. The wrong answer is rarely catastrophic. It is usually expensive, slow, and quietly corrosive, the kind of decision that does not blow up but does not compound either.

The problem with the build, buy or partner question inside a bank

In a software company, build versus buy is a Tuesday afternoon conversation. Inside a regulated financial institution, the same question carries a different gravitational pull.

First, the delivery pressure. Innovation teams, transformation directors and the CIO's office are already running ISO 20022 finalisation, instant payments rollouts, DORA's ongoing third-party register obligations, FIDA preparation, and AI Act timelines for high-risk systems. The build, buy or partner conversation rarely happens with a clean desk. It happens in the margins of an annual plan that is already over-committed.

Second, the regulatory weight. Each path attracts a different supervisor's attention. A build attracts model risk and operational resilience scrutiny. A buy attracts prudential supervision, fit-and-proper reviews, and integration risk. A partner attracts outsourcing rules, the EBA's guidelines on outsourcing arrangements, DORA's ICT third-party risk requirements, and, for in-scope cloud providers, the watchful eye of the ECB. None of these are blockers. All of them change the unit economics of a decision that, on a single slide, looks identical.

Third, vendor noise. The number of fintechs presenting to European banks has grown faster than the number of hours in a quarter. Innovation scouts at the largest institutions report taking 80 to 150 vendor pitches a year and short-listing fewer than ten. Most of the noise is not bad, it is just not decision-grade. The teams making build, buy or partner calls are operating with weak market signal precisely when the cost of being wrong has gone up.

Fourth, internal stakeholder complexity. A build decision pulls engineering, architecture and product. A buy pulls M&A, treasury, legal, the supervisory board and integration management. A partner pulls procurement, vendor risk, information security, business continuity, compliance and the line of business. Each path has a different "yes" coalition and a different veto list. Picking the path is, in practice, picking which internal politics you can survive.

Common approaches, and what they cost

The cleanest pattern is full in-house build. The pitch is strategic control, IP retention, alignment to the bank's risk appetite, and the long-term cost curve. The reality is that mid-tier European banks rarely have the engineering depth to ship modern, cloud-native capabilities at the cadence the rest of the market is moving at. The Aion/Vodeno deal is, among other things, an admission that buying a stack is faster than hiring one. UniCredit did not lack talent; it lacked time.

The buy path looks the most decisive. It is also the most expensive to get wrong. The post-merger integration of a fintech into a regulated bank is where most of the value either compounds or evaporates. Bain's data suggests that scope-only deals, buying capability without integrating it into the operating model, underperform. The acquirers who win are the ones who have a coherent value thesis before the term sheet, not after the press release.

The partner path is the modal choice. Most banks default to it because it sits closest to existing procurement infrastructure. The cost is rarely captured on a single budget line. It shows up in vendor sprawl, in shadow IT, in a third-party register that grows faster than the team maintaining it, and in DORA-mandated exit plans that, when stress-tested, are not actually executable. Partnering is not free. It just spreads its cost across more cost centres.

A fourth pattern is increasingly visible: the sidecar. Rather than choosing between cores or between vendors, banks stand up a parallel modern stack, often partner-built, sometimes acquired, to handle new products or a narrow customer segment while the legacy estate runs the existing book. IDC's projection that 40% of global banks will run a sidecar in 2026 reflects how attractive that compromise looks under delivery pressure. The honest read is that the sidecar is excellent at deferring the hard decision and dangerous if the migration path back to a single core is not specified at the outset.

A smarter route: decide the question before you decide the answer

The teams making good build, buy or partner decisions are not better forecasters. They are better at framing the decision before the options arrive.

Three questions reframe the conversation:

First, is this capability strategic or scaffolding? A capability that differentiates the franchise, pricing engines, risk models, the customer interface, has a different default than one that is table stakes for the next regulatory cycle. Banks routinely over-build the scaffolding and under-invest in the differentiator. The build, buy or partner answer changes once that distinction is honest.

Second, what is the half-life of the underlying technology? Agentic AI, payments rails, identity infrastructure and fraud detection are moving at different speeds. A capability with an 18-month half-life is rarely a build candidate inside a bank, by the time the platform team ships v1, the market has moved. The reverse is also true: a stable, well-understood capability is rarely worth the integration and exit cost of a vendor partnership.

Third, who else has solved this, and what did it cost them? Peer benchmarking is the most under-used input into build, buy or partner. Most banks do not know with any precision which of their peers tried the same vendor, in which configuration, with which integration cost, and what they would do differently. Without that signal, every bank is funding the same learning curve in parallel.

Closing the gap on the third question is where structured peer exchange earns its place. The Discovery Innovation Meeting format that The Connector runs across European financial institutions is one workable answer, a small, closed-door conversation between innovation leads from non-competing banks who are looking at the same build, buy or partner question from different angles. Peer Forum and Roundtable formats sit alongside it for the deeper, market-wide debates. The Finance X Magazine editorial track captures the same signal in a written form for teams who cannot travel. None of these replace internal diligence. They sharpen it before the term sheet is drafted.

Why this matters right now

The cost of getting build, buy or partner wrong has gone up in the past 18 months for three reasons.

DORA is in force. The third-party register, the concentration risk reporting, and the exit-plan obligations have moved partner decisions out of procurement and into the board's line of sight. A partner choice that was tactical two years ago is now a governed one.

The AI Act is starting to bite for high-risk systems. A build decision in a fraud, compliance or credit-decisioning context now carries a documentation and conformity assessment burden that materially changes the build-cost curve. The buy-or-partner side is not exempt, the obligations follow the deployer, not the developer.

Modernisation deadlines are converging. ISO 20022 finalisation, instant payments scheme rules, FIDA preparation and core migration plans are all landing in overlapping windows. A bank that gets one build, buy or partner call wrong absorbs the loss. A bank that gets three wrong in the same year runs out of organisational bandwidth to fix any of them.

Closing thought

Build, buy or partner is not a one-off decision. It is a recurring capability, the muscle that institutions either develop or outsource to whoever is loudest in the room. The banks that develop it have three things the others do not: a clear view of which capabilities deserve a build, a disciplined buy thesis when scope acquisition is the right answer, and a real peer signal on what partnering has actually cost their competitors.

The deal flow in 2026 will reward the banks that have done that work and punish the ones that confuse activity with decision-making. The question is not whether to build, buy or partner. The question is whether your innovation function is set up to answer that question better next quarter than it did last quarter.