Europe’s data centre market is growing at an unprecedented pace, but that growth is no longer happening where it traditionally did. The FLAP-D region (Frankfurt, London, Amsterdam, Paris, Dublin) remains the heart of the sector, but new capital — particularly that linked to AI model training — is visibly shifting towards the periphery: Spain, Portugal, Italy, the Nordic countries, Poland and, increasingly in public debate, Romania.

The starting point: FLAP-D is full, but far from obsolete

For almost two decades, Europe’s data centre map has had five capitals: Frankfurt, London, Amsterdam, Paris and Dublin — collectively known by the acronym FLAP-D. This concentration was no coincidence: mature power grids, high fibre density, interconnection hubs (such as DE-CIX in Frankfurt or AMS-IX in Amsterdam) and proximity to financial and corporate markets made these cities the default choice for any operator seeking minimum latency and fast access to customers.

Two indicators published this summer, however, show that the model is under serious pressure:

• The JLL – EMEA Data Centre Mid-Year 2026 report shows that new development planned for 2026–2028 is shifting significantly towards greenfield projects located away from traditional hubs.
• An analysis by The Next Web, based on CBRE data, shows even more clearly that only 37% of the new capacity delivered in Europe in the second quarter of 2026 went to the five FLAP-D markets — the remaining 63% went elsewhere.

Importantly, this is not an “exodus” scenario. FLAP-D continues to grow — but the pace of absolute growth in the core markets is being outstripped by growth in secondary and emerging markets.

What the JLL report actually shows

The JLL EMEA Data Centre Report is probably the most detailed tool for understanding this trend through the numbers. A few key points stand out:

FLAP-D is growing steadily, but it is no longer where the marginal investment decision is being made. Live capacity across the region reached approximately 3.8 GW by mid-2026, compared with 1.8 GW in 2019 — a doubling in seven years. A further 1.4 GW is under construction and 2 GW is planned. Paris led deliveries in the first half of the year with 72.5 MW, already exceeding its target for the full year, supported by France’s nuclear capacity. London added 49 MW, Frankfurt 45 MW, Amsterdam 16.3 MW and Dublin 11.4 MW.

Vacancy rates are close to historic lows, while divergence between markets is increasing. The average vacancy rate in FLAP-D stood at 6.4% in Q2 2026 — well below the 16.9% peak recorded in 2021, but almost unchanged over the past two quarters. Frankfurt is the tightest market, with only 3.1% of space available — less than half the FLAP-D average. In practice, large contiguous high-density space has become a scarce resource in the core markets, while pre-leasing several years ahead of delivery has become the rule rather than the exception.

The cost of “powered land” (with allocated electrical capacity) has surged. The average price of land with access to power in FLAP-D has increased by 82% since 2021, from €1.24 million to €2.26 million per MW. Primary markets (FLAP-D) now cost 2.3 times more than secondary markets and four times more than tertiary markets.

The most important structural signal is that greenfield investment is rising from 8% to 39% of projects planned for 2026–2028, while the average distance between a hyperscale site and a hub city has increased from 46 km to 175 km. In other words, new projects are not simply moving to the outskirts of cities — they are moving to the outskirts of entire regions, wherever electrical power can be secured quickly.

JLL’s conclusion is nuanced and useful for any strategic discussion: demand is splitting into two different categories. Latency-sensitive workloads (financial applications, enterprise cloud, everyday digital services) remain “anchored” in FLAP-D, where project delivery times can reach 10 years. Large-scale AI model training (100 MW and above), by contrast, does not require minimum latency — it requires rapidly available electrical power, which is becoming increasingly scarce in traditional hubs.

What CBRE and RaboResearch add

Other analyses confirm the picture outlined by JLL, while adding another dimension that is essential to any business analysis: cost is not the only major factor driving relocation — time matters as well.

Analysts at RaboResearch show that, in congested core markets, a new grid connection can take “10 years or more” — compared with significantly shorter waiting times in Belgium, Portugal, Spain, the Nordic countries and Italy. Amsterdam is the most extreme case: Dutch grid operator TenneT has publicly stated that no additional capacity will be available across most of the Noord-Holland region for the next ten years, while the Netherlands has banned hyperscale projects exceeding 70 MW and 10 hectares across most of the country. Frankfurt’s central grid will not see major upgrades before the 2030s, while work on the West London substation (which supplies the London market) is not expected any earlier — which is why a flagship project such as Nscale’s campus in Essex has yet to get underway. The situation is similar in Denmark.

On the cost side, the CBRE analysis confirms JLL’s figures (“powered” land at €2.26 million/MW, up 82% since 2021) and adds useful context: nothing is cheap in absolute terms in the new markets — but everything is cheaper than FLAP-D, which is a different proposition. Under this combined pressure of cost and time, the share of greenfield land has jumped from 8% to 39%.

According to the source, the clearest indicator of relocation comes from the neocloud segment (providers specialising exclusively in AI infrastructure): new capacity contracts reached 420 MW in Q2 2026, compared with just 89 MW a year earlier — an increase of almost 4.7 times, concentrated particularly in the Nordic countries, where renewable energy is cheaper.

The reports cite specific projects that confirm the trend:

• Spain — installed capacity of 439 MW at the end of 2025, with the Spain DC association projecting approximately 2,537 MW by 2030; Amazon has committed €33.7 billion to the Aragón region alone.
• Finland — Nebius is building an €8.5 billion, 310 MW site in Lappeenranta.
• Norway — the Stargate Norway project near Narvik will start with 230 MW powered by hydroelectricity, with a further 290 MW planned.
• Ireland — the connection moratorium ended in December 2025, but the “price of entry” remains high: proprietary generation (“behind-the-meter”) sized to match the entire grid connection, together with 80% of annual demand being covered by own investments in renewable energy.

CBRE confirmation: the market is growing overall, but its centre of gravity is shifting

The Europe Data Centres Figures – Q2 2026 report independently confirms this direction:

• Europe’s total data centre market is expected to grow by around 20% in 2026, reaching 13 GW of capacity — growth intrinsically linked to AI infrastructure.
• The hyperscaler self-build segment (Google, Microsoft, Amazon, Meta and others that build their own facilities) is expected to grow by 22% in 2026, reaching 4.3 GW — the seventeenth consecutive year of double-digit growth. Importantly, 70% of this operational capacity is concentrated in just four markets — Ireland, the Netherlands, Sweden and Belgium — evidence that the “old guard” is maintaining its installed base even as new capacity moves elsewhere.
• Of the nearly 700 MW delivered in Q2 2026 (colocation + self-build), only 252 MW (37%) went to FLAP-D markets, with the remainder going to secondary and tertiary markets.
• The neocloud segment quadrupled capacity signings compared with the previous year, reaching 420 MW — again concentrated in the Nordic countries, due to the lower cost of renewable energy.
• Even as secondary markets grow rapidly, overall take-up remains higher than new supply across Europe — the reason being, as CBRE explicitly states, “the lack of available electrical power, which is causing delays in data centre construction”.

CBRE’s overall message is important for any investor or developer: this is not a market that is shrinking or relocating from a weak region to a stronger one — it is a rapidly growing market in which the marginal investment decision is shifting to locations where power is available quickly, regardless of the brand or age of the location.

What, then, are the business arguments?

Taken together, the information from the five sources outlines an extremely coherent investment rationale that can be summarised in four factors:

1. Grid connection time has become the critical variable, not cost. A project that waits 10 years for a connection in Frankfurt or Amsterdam loses the AI race before construction even begins. A similar project in Aragón, Sines or Kouvola can be connected several times faster.
2. Workloads have bifurcated. Enterprise cloud, financial applications and everyday digital services remain “tied” to hubs because they depend on minimum latency. Large-scale AI model training does not face the same constraint — it can be located wherever sufficient, affordable and rapidly available electrical power exists.
3. The cost of “powered” land has become a price of entry rather than an absolute barrier. The higher costs (2.3x and 4x) paid to locate in FLAP-D reflect the scarcity of power and space in the core markets — not any structural superiority of those markets for AI workloads.
4. Countries that align public policy with this logic benefit disproportionately. The examples of Spain (Declaration of General Interest for accelerating projects), Ireland (a specific legal framework for renewable energy linked to grid connection) and Poland (deliberate positioning as a Sovereign Cloud hub) show that the speed of administrative decision-making has become, in 2026, a competitive advantage just as important as the price of land or energy.

Could the same logic apply to Romania?

The question raised by all the data above for the local market is straightforward: if, in Western Europe, capital is shifting from major established hubs (Paris, Amsterdam, Frankfurt) towards locations that are cheaper and provide faster access to power, could a similar shift follow the same logic — this time on a continental scale — from Western towards Central and Eastern Europe?

The arguments supporting this hypothesis are not speculative — they are exactly the same four variables discussed above, applied on a larger scale:

• Connection time. Romania does not currently face the 10-year waiting queues seen in Amsterdam or Frankfurt. Within the logic described by RaboResearch and CBRE, this is exactly the kind of advantage that has already shifted billions of euros towards Spain, Portugal and the Nordic countries.
• The cost of land and energy. Structurally, Romania starts from a price level well below that of primary Western European markets — a situation similar to that of Spain or Poland before the current wave of investment.
• The energy mix. Projects in Romania could rely on nuclear power from Cernavodă, but also on a hydro-solar-gas mix, just as Norway/Finland (hydropower) have already attracted large-scale AI capital.

The market signal has already been visible for several years. As has been repeatedly discussed at DataCenter Forum Romania — organised by Tema Energy — the local industry is openly considering the idea of Romania becoming the “new Norway” of data centres: a historically modest market that could gain a disproportionately strong position through the combination of available energy and a limited window of opportunity. At the May 2026 edition of the event, this was precisely the central discussion: the future of the sector is no longer defined by construction capacity, but by rapid and predictable access to energy — exactly the area where Romania still has room to move.

All the successful examples in Europe (Aragón, Sines, Kouvola, Warsaw) share one common element in addition to lower costs and available energy: active alignment by the state with this economic interest — whether through accelerated permitting procedures (Spain), a dedicated legal framework for energy and grid connection (Ireland), or an explicit public policy position supported by regional offices and partnerships.

The logic of investment relocation described by JLL and CBRE is not specific to Western Europe — it is a universal market logic that will repeatedly seek out the fastest available combination of land, energy and administrative predictability. Structurally, Romania already has the first two elements. If it can secure the third — sustained, coherent alignment around a national interest, without repeated delays — it could genuinely compete for the position currently occupied by Spain or Finland on this rapidly reconfiguring map.