Scored to Lose

Somewhere in Aragón this summer, a data-centre developer with money in hand, a signed letter of intent from a hyperscaler and a plot zoned for industrial use will apply for the one thing that actually matters — a connection to the electricity grid — and lose it to a hydrogen project that may never be built. This is not a failure of the developer. It is the system working exactly as designed.

For two years the industry treated grid access as a queue: file early, wait your turn, connect. That queue is gone. In its place is a competitive tender, the concurso de capacidad de acceso de demanda, run by the state at congested nodes on the transmission network. And in the scoring rubric that decides who gets a node, data centres do not sit at the front. They sit, by explicit design, at the back.

The mechanics were the subject of a recent ATA Insights webinar, Tramitación estratégica en 2026, in which two people who spend their days inside these files — Daniel Vázquez, who runs public and regulatory law at Simmons & Simmons in Madrid, and Humberto Carbajal, a senior infrastructure adviser at Colliers — walked through how the game is actually scored, and how a developer plays a hand the rules were not written to help.

The queue that turned into an auction

Start with why the queue died. Under the old rules, anyone could request a connection point, and — as Carbajal put it — "everyone could ask for a connection point, and just in case, they asked." The result was a phantom pipeline: nodes clogged with reservations for projects that existed mostly as options. When real demand arrived, there was no room, because the room had been booked by speculators.

So Spain did what Spain does. It regulated. Since Royal Decree-Law 8/2023, access at a congested transmission node is no longer first-come, first-served. When the grid operator receives a request at a node of 220 kV or above, it publishes that request and opens a one-month window for anyone else to apply for the same capacity. If the applications exceed what the node can carry, the state pulls the requests out of the queue and runs a tender.

Carbajal, whose instinct is to be sceptical of Madrid's regulatory reflexes, was unexpectedly warm about this one. "We consider it good for the sector," he said. "It stops a great deal of speculation, above all in connection-point requests — the thing that created those bottlenecks and that collapse." Then came the caveat, delivered with the weariness of a man who has watched this film before. "In Europe we're junkies for regulation. We can push out rules fast, very deep ones, often without talking enough to the sector — and then they turn out to be hard to apply, and business gets a little stuck."

The tender fixed the speculation problem. What it did not do was decide the contest in the data centre's favour. Quite the opposite.

The rubric, and the trapdoor inside it

Here is the part every developer needs tattooed somewhere visible. A concurso is decided on three criteria: the start date of consumption, the volume of investment, and the greenhouse-gas emissions avoided. That last one — avoided emissions — carries the most weight of the three. As Vázquez laid it out, it is "the preponderant one over the three I've just mentioned."

Read that again with a data centre in mind. A data centre is a magnificent thing to build and a hopeless thing to score on avoided emissions. It does not displace a coal plant. It does not decarbonise a steel mill. It is, in carbon-accounting terms, a very large, very hungry new load that happens to buy renewable power. So when the single heaviest criterion is "how much CO₂ does your project avoid," the data centre walks into the exam already failing the question worth the most marks.

Vázquez was blunt about the consequence. The avoided-emissions criterion, he said, "isn't one that particularly benefits — not to say it may actually harm — those entities that had applied, within this competitive process, for access to one of these nodes. Because it seems to be favouring projects like hydrogen or renewables, and electrification projects." When capacity is scarce and priority goes to projects that cut emissions, "it doesn't seem to be a criterion designed to benefit the data-centre sector very much."

The design is not a rumour to be read between the lines. Underneath the scoring sits a formal hierarchy of consumption types. Type 1 is renewable hydrogen and renewable gases. Type 2 is industrial electrification and mining. Type 3 — the back of the room — is data centres. In a contested node, types 1 and 2 have structural priority over type 3. A developer can arrive with the deepest pockets and the earliest energisation date and still be outranked by a project whose chief virtue is that it fits the government's decarbonisation story better.

The scoreboard bears this out. The first round of these tenders, resolved in early 2026, awarded 928 MW across a handful of transmission nodes — and it went to industrial and green-hydrogen projects. Data centres, the loudest and best-capitalised demand story in the country, did not walk away with the prize. A fresh round of tenders is now working its way through roughly four to five regions — Aragón, Andalucía, Castilla-La Mancha, Cataluña, Galicia and the Basque Country were the names in play — and the rubric has not changed.

So the strategic problem is not "will there be capacity." It is: the scarce capacity that exists is allocated by a formula that structurally down-ranks you. Vázquez's verdict on the tenders as a rescue for the sector was quiet and final. This "hoped-for, one-off — certainly not structural — solution" to the access problem, he said, "all points to it not being the solution."

Why the grid is empty in the first place

None of this would bite if there were spare capacity to go around. There is not. When the distributors published their capacity maps a few months ago, roughly 84% of the distribution network was already saturated. In the Basque Country, La Rioja, Navarra, Aragón and Andalucía it is effectively full. Only in a few corners — Asturias, the Canaries, the Balearics, Galicia — is there much room left. "If there's no capacity," Vázquez said, "there's no possibility of connecting projects" — not data centres, not anything.

The demand curve explains the panic. The current network plan assumed around 2 GW of new demand. The plan being drafted for 2026–2030 pencils in 27 GW. That is not a trend; it is a step change, and for the first time the whole planning exercise pivots on demand rather than on generation. Where earlier plans were built to evacuate renewable output onto the grid, the next one is built to feed loads pulling off it — data centres prominent among them.

Madrid has started to move, prodded, Vázquez suggested, by April's blackout. A draft royal decree — its public-consultation window closed the very week of the webinar — earmarks more than €7bn, close to €8bn, for the distribution network and over €3.5bn to reinforce transmission, lifting the investment ceiling by around 62%. Real money, at last.

But money planned is not money spent, and here Vázquez put his finger on the coordination that Spain keeps failing at. The environment ministry writes a plan that raises investment; the energy regulator, the CNMC, drafts a remuneration order aimed at cutting network tariffs — and tariffs are what pay for investment. "On one hand you're attending to investment," he said, "and on the other, to cutting tariffs." Grid build-out is, in any case, a matter of "months, of years. This is not an immediate answer. It's a medium-to-long-term answer." A developer who needs power in 2027 cannot wait for a substation that arrives in 2031.

The paperwork tax, and the town that had never heard of you

Even a developer who clears the energy hurdle then meets the bureaucratic one. A draft decree transposing EU rules on data-centre energy efficiency and sustainability layers on new reporting duties — water use, refrigerant types, waste heat, socio-economic impact, all of it public and open to third-party scrutiny. Projects above 100 MW must prove they sit in the top 15% of performers on energy and water metrics; and compliance with all of it becomes a condition of the very access-and-connection permit the developer is already fighting for through the tender. Laudable in intent, Vázquez noted, but "far from easing the bureaucratic burden, it seems to increase it."

Then there is the ground beneath the building. Data centres are a genuinely new use, and Spanish planning law was mostly written before anyone had heard of them. By Carbajal's count, essentially "one municipality in Spain, and one other" have a specific land-use category for a data centre. Everywhere else, the developer is arguing by analogy — is this industrial use, or tertiary? — with a local authority that, as Vázquez put it, often finds the whole thing generates "more fear than anything else, because it's unknown." In practice, he said, "90% of data-centre projects are developed on industrial land, because the only alternative is tertiary" — and tertiary zoning drags in requirements Carbajal called, with feeling, "stupid for a data centre": bathrooms mandated in halls where no human ever sets foot.

How you actually play it

For all the gloom in the diagnosis, both speakers were describing a game that can be won — just not by brute force. The moves follow from the rubric.

Play the demand type, not just the node. Since types 1 and 2 outrank data centres, the developers most likely to secure a node are those who can dress their demand in something the formula rewards — co-locating with industrial electrification, or pairing with on-site generation and hydrogen so the project reads as more than a bare load. It is telling that the winners of round one were industrial and hydrogen projects; the lesson is to be, or to stand beside, one of them.

Play location as a function of power, not prestige. Carbajal was withering about site selection driven by a map rather than a customer. "You can't just say, right, I'm going to build a data centre in Albacete because I think I've got land," he said. "First you have to look at the business model — who are you aiming at, what do you want to bring there." Madrid, Barcelona and the Aragón corridor between them are the obvious plays; the strategic opportunity lies where power is available and a specific end-customer can tolerate the trade-offs. An AI training campus that can live with higher latency can chase 200-plus MW in places a latency-sensitive colocation client would never accept. Aragón, he noted, won its position not by geography alone but because "part of the administration was agile enough" to hand the hyperscalers real facilities fast.

And play the regions against each other. This is where the strategy turns from defence to offence. Where the national tender ranks data centres low, some autonomous communities are doing the opposite — declaring data centres projects of strategic regional interest, creating special-actuation instruments (Madrid's proyectos de actuación especial, Aragón's regional-interest plans) that authorise the use on land where ordinary planning never contemplated it, and compressing permit timelines that would otherwise run to years. The map of where to build, Vázquez said, "depends first on where there's capacity in the grid, and then on the regulations" — and those regulations are anything but uniform. The developer's edge is to read that patchwork as a menu rather than a maze.

The kicker

Underneath the rubric is a bet the state has not quite admitted it is making. By weighting avoided emissions above all, Spain has decided that scarce grid capacity should preferentially decarbonise the old economy rather than power the new one. It is a defensible bet. It is also the kind of bet that, if the arithmetic stays this lopsided, sends the hyperscalers elsewhere — and Carbajal named the elsewhere without flinching. Miss this, he warned, and the operators "go to Milan, to Marseille, or even to Latin America — we've already seen some go."

Spain has the renewables, the subsea cables, the geography and, for now, the queue of investors. What it does not yet have is a scoring rubric that lets a data centre win a node on its own terms. Until it does, the smart developer will stop trying to beat the formula head-on and start doing what winners of rigged games always do: change what they are being scored as. As Carbajal put it, with the flat certainty of someone who has watched the meter run, on the day of the tender "one thing is having a letter" of capacity — "and another is having the power at the plot gate. That's a different battle."

Watch Tramitación estratégica en 2026 on demand, and bring the harder cases in person: the whole debate reconvenes at RENMAD Datacenters in Zaragoza, where the people writing and losing these tenders will be in the same room.