“The decisive competition will not be won by the country that announces the most gigawatts.”
Today brings one development consequential enough to trigger the Watch: Nabiax has formally started construction of an €800 million, 80 MW AI/cloud data center in Alcalá de Henares, Madrid. The first stone was laid on September 18. The facility is designed for AI and cloud workloads, will use electricity supplied from 100% renewable sources, and is expected to support about 1,000 jobs during construction. It will become Nabiax's fifth data center in the Madrid region.
This is not an EU AI Gigafactory award. At 80 MW, it is materially smaller than the eventual Gigafactory class. But strategically, it may be more informative than another multi-gigawatt announcement — because it has crossed the line from plan → permit → capital commitment → construction. That is precisely where much of Europe's announced AI capacity still fails.
The Execution Signals
from PowerPoint to concrete
-
Our framework has increasingly distinguished announced MW from bankable MW. Today Spain adds something more valuable: MW under construction. Madrid now has 41 operating data centers and 15 under construction, with another 40 projects envisaged that the regional government says could represent around €15 billion of investment.
Every step in the chain — land → grid → capital → contractors → construction → operating compute — eliminates another layer of execution risk. A theoretical 500 MW project five years from operation may be economically less valuable than an 80 MW project already pouring concrete. signal: Add EXECUTABLE MW to the Watch — not just announced or financed capacity.
-
Spain is simultaneously tightening the discussion around data-center grid access, renewable additionality and economic value creation, while private infrastructure capital continues to enter Madrid. The relevant question is no longer will regulation stop investment? but which type of investment survives the new regulatory environment?
A selection mechanism is emerging: speculative grid reservations face greater difficulty, while capitalized, permitted and executable projects continue progressing. Spain could end up with fewer headline GW but a higher percentage reaching operation — validating our Capacity Conversion Velocity thesis. The empirical test over the next year: Operating MW / Announced MW. signal: If the denominator shrinks but the numerator rises faster, the system has become more productive.
-
One project creates capacity. Many projects create an ecosystem. Madrid's accumulation of operating and constructing facilities generates circular effects: more data centers lift fiber density, energy infrastructure, specialized construction, the technical labor pool and supplier density — cutting time-to-build and attracting still more capital.
This is industrial clustering applied to accelerated computing. Once the flywheel begins, it becomes increasingly difficult for a competing geography to replicate quickly. signal: Madrid is accumulating execution capability — that is why today's €800 million matters beyond Nabiax.
-
Europe increasingly has capital, demand and announced projects. The scarce variable is becoming time-to-power. An AI infrastructure investor asks four questions in sequence: When can I obtain grid capacity? When can construction begin? When can accelerators arrive? When can customer revenue start?
Every additional year between committed equity and operating compute raises financing cost, delays cash generation, increases technology-obsolescence risk, and raises the probability that competing capacity enters first. signal: The Gigafactory race is partly becoming a competition in calendar time.
This Changes The Meaning Of Europe's €30+ Billion Gigafactory Programme
pre-solved infrastructure wins
Italy — A Different Form Of AI Value Capture
supplying the physical stack
The Hidden Pattern
layers of the framework
Announced MW
Where most headlines still live — capacity that exists on a slide.
Bankable MW
Capacity with financing structure behind it.
Bankable Intelligence Ecosystem
Financeable AI cash flow supported by a working ecosystem.
Execution Velocity
The new layer: none of the previous variables matter economically until capital becomes operational compute.
what the Watch should track
- The decisive metric is AI Infrastructure Execution Velocity: operating MW added per year ÷ bankable MW in the pipeline.
- Executable, capitalized and permitted projects will survive Spain's tightening regulatory environment; speculative grid reservations will not.
- Time-to-power is the scarce asset — the Gigafactory race is partly a competition in calendar time.
- Europe can capture value across the physical AI stack, not only in compute ownership.