DataCloud Global Congress: Key Takeaways

Key takeaways from Datacloud Global Congress 2026

Europe must think like a continent, not a collection of countries

The opening theme across several panels was scale, and the gap between European ambition and US execution. A panel on building Europe’s AI gigafactories put the question plainly: does a roughly 20-billion-euro consortium and four or five gigafactories touch the sides next to American spend?

The consensus was that Europe’s real advantage lies in public and private collaboration, a complete supply chain, and policy tools such as the Cloud and AI Development Act, accelerated permitting zones, and sovereignty frameworks that still let global providers in. Poland emerged as the standout energy story, with around 3.2GW of capacity being developed across the Baltic and new nuclear in the mix, redesigning how power is generated and opening fresh opportunities across the continent. Portugal’s Sines region and the Nordics continue to attract investment on the back of cheaper land and power.

On how the facilities get built, the message was that stick build only scales so far. Pre-fabricated, modular design is more stable and repeatable, and the unit of measurement has shifted from megawatt data centres to megawatt racks. The closing point was blunt: act as a continent, streamline permitting, and revenue will follow, increasingly measured in dollars per token rather than megawatts deployed. Water, often imported as a concern from US coverage, was largely answered on stage, with modern cooling approaches now consuming little to no water in operation.

Capital is following conviction, not waiting for certainty

A dedicated financing session made clear that the market is still early in its capital formation. The most bankable structure right now is a bankruptcy-remote vehicle in which customer contracts self-amortise the debt over the contract term, leaving very little risk if deployment milestones are met.

The sharpest argument came from a developer who described deploying around 1 billion dollars of proprietary capital to unlock roughly 45 billion dollars of hyperscaler investment, built on a Pennsylvania campus that connected a data centre directly to its power generation. The thesis: execute first, and let that execution unlock the financing. The wider view was of a world being re-architected around energy abundance, with West Texas moving from a 7.5GW market toward 25GW in short order.

Elsewhere, operators described leveraging long-term contracts to bring down the cost of purchasing GPUs, with deals becoming more favourable as supply builds. Notably, expansion in Europe and the Nordics is being driven less by lowest-latency demands and more by timeline to power and the cost of that power, with around 80 percent of site selection resting on that model. Canada is showing strong growth, Iceland is intriguing for its green power if you can live with the volcanoes, and APAC is opening up. The prevailing read was that the US remains the dominant market, with no shortage of opportunity still to come.

Building up, and insuring like a capital provider

A session on vertical construction confirmed that going up is use-case driven and far from new. A nine-storey facility built in Singapore some 14 years ago was offered as evidence, and London’s docklands carrier hotels make the same point about latency and proximity. Where land is scarce, as in Japan, up is the only option.

The economic sweet spot sits around four to five storeys, before roof space for heat rejection becomes the constraint. As rack density rises, buildings may actually get shorter while delivering the same output, and cheaper with it, since you cannot save much on the below-ground work. Speed is the other driver, which pushes the case for offsite manufacturing, although steel volumes are far higher in a vertical build.

The insurance lens was just as practical. Water, not fire, is the primary risk in a vertical build, where a single event has a far bigger effect than on a single-storey site, and the value of GPU equipment now dwarfs everything else on the floor. Lithium batteries are a flagged concern. The advice from the insurance side was to treat your insurer like a capital provider, engage them early in design and construction, and bring them along on the journey. As one panellist put it, insurance does not see vertical as more challenging, just a different conversation.

Floating data centres, and a talent gap that follows the tide

One of the more striking sessions presented floating data centres converted from second-hand merchant vessels, developed in partnership with a floating-power specialist already running power ships off the coast of Africa. The pitch is proven technology, reconfigured: a vessel of roughly 120 metres, with a modular IT load of 20 to 73MW per ship, sea water cooling on open or closed loop, and power supplied either by an adjacent power ship or from shore.

The appeal is speed and flexibility. Existing ships have short lead times, deployment in industrial port zones sidesteps much of the community resistance and permitting that slows land-based builds, and the asset can be moved if needed. Up to 1GW of capacity could be available within around 18 months, with target markets including Japan, Malaysia and the United States, and first deployments potentially live from 2027. For a recruitment business, the implication is direct: this sector will need people who can bridge maritime engineering and critical-facility operations, a blended skill set that barely exists today.

The power architecture question: AC or DC for the AI era

One of the sharper technical debates on the agenda came from a panel asking a deceptively simple question: is AC or DC power better suited for tomorrow’s high-density computing needs? As racks push past 100kW, a distribution architecture that sat unquestioned for years is suddenly back on the table, with real consequences for efficiency, footprint and how power is delivered to the chip.

The thread connecting the room was that the UPS is no longer a passive emergency asset. AI training and inference create extreme, millisecond-scale load swings, jumps of tens of megawatts in fractions of a second, that traditional backup systems were never built to absorb and that can trigger voltage instability, breaker trips and generator stress. The direction of travel is towards lithium battery architectures that turn the UPS into an active, real-time load-smoothing layer. For anyone building critical-power teams, the message is clear: power electronics and energy-storage expertise are moving from a specialist niche to a core hiring priority.

Other conversations that defined the floor

Cooling was everywhere. With grey space steadily displacing white space and racks crossing 100kW, heat is now a design question rather than an afterthought, and direct-to-chip and immersion liquid cooling have moved from the fringe to the baseline assumption for AI halls. The people who can design, install and maintain those systems are in short supply, which is why the line about needing as many plumbers as electricians landed so hard.

Power supply dominated, and the conversation has widened well beyond the grid. Behind-the-meter generation, gas, and a serious thread on nuclear, including small modular reactors, ran through the programme, with a dedicated Nuclear for AI Summit sitting alongside the main agenda. Grid connection queues stretching out for years are pushing operators to secure their own power, and that is reshaping where sites get built and who is needed to run them.

Commissioning earned a session of its own, billed as the invisible process that controls a data centre’s fate. It is the discipline that quietly protects schedules, budgets and uptime, and it is one of the hardest areas to staff. As build programmes run concurrently across multiple sites, demand for experienced commissioning managers is outstripping supply in every European market.

And the talent question had a platform of its own. A panel on building a long-term pipeline for a resilient workforce confronted the problem every other session circled: specialised labour is scarce, especially in remote locations, and the industry has to build its next generation rather than simply compete over the current one.

People remain the constraint no technology solves

The session that closed the congress, a women in leadership panel captured by the phrase “women in white, not men in black”, a nod to technical site roles over boardroom suits, brought the three days to a fitting point. One speaker noted that growth is outpacing the available talent pool, with a wave of retirements compounding the gap, and that the industry needs as many plumbers as electricians. Another spoke about women having the courage to lean into technical roles and push through the myth that they do not belong there.

The energy challenge was framed starkly: most grids are running at around 50 percent and cannot continue as they are, so flexible connections and microgrids will matter. The line of the session was that you should not chase the puck but anticipate where it will be, and that serviceability and customer obsession are everything. The day also set out a five-layer AI infrastructure stack, from energy through infrastructure, GPUs and models to the applications on top, where every layer needs the right people to function.

The thread running through all of it was consistent: technical adaptability, the courage to move into new roles, and the ability to anticipate demand rather than react to it are the qualities that will define the next generation of data centre leadership.

The talent throughline

The coming months and years will bring significant development across the European data centre market, whether that is gigafactories in the Nordics and Iberia, vertical campuses where land is tight, or vessels moored off an industrial port. Every one of those routes runs into the same bottleneck, and it is not steel, capital or even power. It is people. That is exactly where we focus: finding the specialist talent that lets ambitious companies turn this market’s potential into operating capacity.

Want to know more? Contact Michael Aspinall.