On 30 July the European Commission launched a call for proposals to establish up to seven AI Gigafactories across Europe, an initiative the Commission says will unlock more than €30 billion in investment. Up to €10 billion of that is EU and national public money; the Commission expects it to unlock at least €20 billion in private investment.
Most of the coverage has treated this as a chips-and-sovereignty story, which it is. But read the EuroHPC Joint Undertaking’s own call notice and a second story appears underneath it. What Europe has actually just tendered is a construction programme.
What Is Being Procured
The Commission describes the facilities as combining “advanced AI processors, software and cloud technology stacks, high-speed connectivity and energy-efficient data centres.” That last clause is the part that involves excavators. These are not server rooms retrofitted into existing buildings — they are purpose-built compute campuses, at a scale the “gigafactory” branding is meant to imply, and they will need siting, grid connections, cooling infrastructure and someone to pour the concrete.
Successful projects will receive different levels of EU and national co-funding, which EuroHPC says will reflect “the size and range of sovereign services that they will be deploying.”
The Timeline Is the Story
Here are the dates, from the EuroHPC notice:
- 12 November 2026 — submission deadline
- Early 2027 — competitive evaluation concludes, projects selected
- Within 18 months of selection — the gigafactories “are expected to begin operations”
That last line is the one to sit with. A project selected in, say, February 2027 would be expected to be operating by roughly mid-2028. For a facility that has to be sited, permitted, powered, built and then filled with accelerators, that is a compressed schedule — and it is worth being precise about what it means, because the notice says begin operations, not complete construction, and does not specify how much of each site must be new-build versus expansion of existing infrastructure. Consortia bidding on brownfield sites next to existing EuroHPC facilities face a very different job from consortia proposing greenfield campuses.
There is a funding wrinkle in the same direction. EuroHPC states that part of the financing comes from the current Multiannual Financial Framework, but “the bulk of the funds will come from the next Multiannual Financial Framework (2028–2035).” The construction schedule and the money supply are therefore not perfectly aligned: sites are meant to be operating around the time the larger tranche of public funding begins.
The Contrast With New York
Sixteen days before this call opened, New York did the opposite. Executive Order 62, signed on 14 July, froze state permitting for data centers of 50 MW and up, and fifteen days after that an analyst cut Caterpillar’s price target citing it.
The contrast is not quite “Europe builds, America blocks,” and it would be lazy to write it that way. New York is one state, and the US data center buildout is proceeding at enormous scale elsewhere. But the two decisions do point at the same underlying constraint from opposite ends. New York’s moratorium was a response to permitting and grid pressure arriving faster than the state could absorb. Europe’s call is an attempt to manufacture exactly that pressure on purpose, with public money as the anchor tenant.
For contractors, the practical read is that the binding constraint on AI infrastructure is increasingly regulatory and electrical rather than financial. Capital is not scarce in this market. Sites with power and permission are.
What to Watch
Three things will tell you whether this becomes real construction work or stays a press release.
Who bids. The composition of the winning bids is the tell. If they are led by cloud operators and chipmakers with construction partners attached, the buildout will look like the US hyperscale model. If they are led by national research institutions extending existing supercomputing sites, the construction content will be much smaller than €30 billion suggests.
Where they land. Siting decisions will follow power availability, and the countries with spare grid capacity and permitting willingness are not necessarily the ones with the largest AI ambitions.
Whether the 18 months holds. This is the number to watch after selection in early 2027. If it slips, it will slip for the reasons construction schedules always slip — grid connection queues, permitting, and the availability of people who can build this kind of facility — and that slippage will be the most honest available measure of Europe’s capacity to build AI infrastructure at speed.
Proposals close 12 November. Until then, this is a €30 billion intention rather than a project pipeline. But few construction signals in the European AI buildout have been this concrete, and the industry that will actually pour it has barely registered that the tender is open.
