Last week five of the leading voices in European financial market infrastructure, including our CRO Richard Vosper, came together for a conversation that challenged a founding financial market infrastructure assumption: that speed and uptime are the only measures of network readiness.
The feedback has been strong. Below are the standout contributions from each panellist – and their response to five key questions arising from the discussion.
What the panel said
Richard Vosper, Chief Revenue Officer, Zayo Europe
Richard reframed the industry’s core infrastructure metric. The conversation has moved, he argued, from latency benchmarks to something harder to measure but more commercially significant.
“Speed is important, but predictability – how the network performs under stress and under unknown events, route changes, dependency degradation, things like that – these conversations are much more prevalent now.”
Most firms are sitting on infrastructure assembled piecemeal over decades. The seams between suppliers are precisely where resilience fails first.
Francisco Santana, Senior Advisor, Market Intelligence, Ciena
Francisco made the broadest argument of the session: connectivity is no longer an afterthought, it is where differentiation is built or surrendered.
“The abundance and flexibility of the network is equal to the capacity of delivering more and differentiating more.”
As AI moves from experimentation to production scaling, firms with well-governed, adaptable infrastructure will have a shorter path to deployment. Those that have not will be held back by the weight of what they have accumulated.
Eleni Coldrey, EMEA Business Development Director, Equinix
Eleni tackled the assumption that public cloud is the answer to scale demands head-on.
““If you are intending to architect on the public internet or public resources, just think about every one of your counterparts, every one of your competitors, everyone in your value chain that’s also trying to scale on those same elastic resources at the same time. The reality is that at some point they’re not elastic anymore.”
Another headline: DORA has formally named Equinix as a critical third-party provider – a signal, she suggested, that regulators are now thinking about the physical infrastructure layer in the same terms as the institutions it serves.
Michael Barbera, Chief Product Officer for Trading, Iress
Michael focused on what accumulated infrastructure does to data – and by extension, to any firm that believes it is ready to deploy AI. Fragmented infrastructure produces fragmented data. Fragmented data is precisely what AI amplifies rather than corrects. His other observation was more human:
“There’s always that single person dependency. Someone who’s been in that role, built it, nothing documented and they’ve moved on to the next firm. There’s a problem, and in the midst of it, the failover doesn’t work. That’s a major catalyst for change.”
Diana Stanescu, Keysight Technologies
Diana drew one of the sharpest distinctions of the session: Availability is not resilience.
“An infrastructure which is fast but has some bursts when you don’t expect them is actually worse than an infrastructure which is a little bit slower, but it’s constant.”
Her test-monitor-retest methodology – full-path testing at peak load, millisecond recovery measurement, emulating AI traffic alongside live market data – reframes what it means to say an infrastructure is resilient.: Evidence, not assumption.
Top five FAQs
1. How do you triage an accumulated estate when rebuilding overnight is not realistic?
Start with the critical path, not the full estate. Map the workflows where failure directly impairs revenue or settlement, identify dependencies that have never been formally documented, and test those first under realistic load. Build a prioritised roadmap from that evidence rather than attempting to modernise everything at once.
2. Isn’t consolidating to a single integrated provider just swapping one concentration risk for another?
The argument is not for single-provider dependency – it is for route transparency. The challenge with accumulated multi-carrier environments is that physical diversity claimed on paper does not always exist in practice: Carriers share ducts, building entries and access providers. The test is not how many providers you have; it is whether you can evidence genuine physical route separation at a duct and building level.
3. When a workflow fails across multiple providers, whose SLA actually governs?
This is one of the most under-resolved problems in financial market infrastructure, and DORA is beginning to surface it. Most SLAs describe component performance, not end-to-end workflow outcomes. Until firms have timestamped evidence correlating network events with order-level outcomes, the accountability conversation at supplier level defaults to everyone pointing at everyone else.
4. How do you make the business case for investment in things that may not happen?
Start with the cost of failure, not the cost of investment. Quantify the value of trades at risk per minute of downtime, the regulatory exposure under DORA and other resilience regulations’ incident reporting obligations, and the cost of a missed T+1 settlement window. The conversation shifts when infrastructure is framed as a revenue creating platform – which is increasingly what it is – rather than operational overhead to be managed down.
5. What does a proportionate approach look like for a challenger bank or mid-tier buy-side firm?
The principles scale even when the resources do not. Managed colocation and managed network services give smaller firms access to enterprise-grade resilience without building it from scratch. The governance discipline – mapping dependencies, testing end-to-end, establishing a performance baseline – requires rigour, not headcount. Buying the service and managing the risk are two different things.
Missed the session? The full recording is available on request. Our full report on infrastructure resilience in European financial markets will be published shortly.