ANALYSIS: Brussels Launches Seven AI Gigafactories to Boost Digital Sovereignty

A call for tenders that commits tens of billions

What the Commission announced on July 30

The European Commission launched, on July 30, 2026, a call for tenders to establish up to seven "AI Gigafactories" across Europe, according to the Commission itself. The initiative is to be backed by up to €10 billion in public funding from the EU and member states, and is expected to unlock at least €20 billion in private investment. Ten billion in public money to attract twenty in private funds: that is the central bet of this arrangement.

A figure that keeps climbing, according to Le Monde

According to Le Monde, seven projects are now planned, up from five announced in February 2025 at the AI Action Summit in Paris, and the total envelope now stands at €30 billion, two-thirds of which comes from private sources. This ramp-up, in just over a year, illustrates the speed at which Brussels has revised its ambitions for AI computing infrastructure, in a context where every major technological power is racing to secure its own processing capacity for tomorrow's models.

Advanced chips and high-speed cloud, the promised technical base

Infrastructure open to multiple users

The "AI Gigafactories" are meant to give start-ups, scale-ups, SMEs, industry, universities, and public authorities access to infrastructure for training, inference, and fine-tuning "frontier AI" models, according to the European Commission. They will combine advanced AI processors, software, cloud technologies, high-speed connectivity, and energy-efficient data centers.

A timeline stretching into 2027

According to Reuters, the call for tenders will close on November 12, and the Commission plans to announce winners in early 2027, with facilities coming online within 18 months of contract signature. This timeline means no gigafactory will be operational before mid-2028 at the earliest, a horizon that few official communications highlight. Between the announcement and the first delivered computation, nearly two years will pass. This delay is not abnormal for a heavy infrastructure project of this scale, but it contrasts with the pace at which AI models themselves evolve, often refreshed within months.

The expected role of public-private consortia

Who can apply

Reuters notes that companies or consortia applying will need to include technology providers, cloud providers, public entities, and investors. This mixed-composition requirement aims to ensure each project has the technical expertise, financing capacity, and institutional grounding needed to carry a project of this scale through to completion. A consortium that only gathered one of these profiles, even a very strong one in its own segment, would likely not be deemed eligible under the general criteria described by the Commission.

An architecture inspired by earlier AI antennas

The Commission had already announced, in October 2025, the creation of "AI Factories Antennas" across member states and partner countries, a more modest scheme than the gigafactories but one that foreshadowed the same logic of territorial coverage. The gigafactories did not come out of nowhere: they extend a project already underway. That earlier antenna program received far less international press attention than this week's announcement, even though it laid the institutional groundwork the Commission is now building on.

Dependence on American chips, the project's blind spot

European infrastructure, still-foreign components

None of the sources reviewed specify the exact origin of the AI processors that will equip these gigafactories. In a global AI chip market dominated by a small number of mostly American or Taiwanese manufacturers, the question of technological sovereignty remains unresolved at a level the budget announcement does not address: funding European infrastructure alone does not guarantee independence from semiconductor suppliers.

What this dependence means for the sovereignty ambition

This piece names this limit rather than passing over it in silence: a gigafactory built in Europe but powered by chips designed and manufactured elsewhere represents only partial sovereignty, covering hosting and operations rather than the full technology chain. Hosting the computation at home is not enough if the processor remains imported. This nuance does not devalue the European project, but it clarifies its actual scope: operational and geographic sovereignty, not yet full industrial sovereignty over the most critical component of the chain.

A three-way digital race against Washington and Beijing

An explicitly named three-way race

According to Reuters, the European Commission wants to fund seven AI gigafactories across the Union with €10 billion, and hopes to attract at least €20 billion in private investment, in a project explicitly presented as a response to competition from the United States and China. Politico describes a text in which the EU is preparing for a "leap forward" on artificial intelligence, amid already well-established global technology competition. This phrasing is not trivial coming from a European institution usually cautious in its industrial announcements, suggesting an acknowledged sense of urgency in the face of a lag accumulated over several years in the global race for AI computing.

The investment gap that persists despite the announcement

Even with €30 billion mobilized, this amount remains below the cumulative investments announced by several major American technology groups for their own AI infrastructure. Closing a gap is not measured only in announced billions, but in speed of execution.

Bloomberg confirms the public figure, with a methodological nuance

A consistent figure across major outlets

Bloomberg reports that the EU "pledges $10 billion in public funding" for new AI data centers, confirming the order of magnitude given by the Commission and repeated by Reuters. This convergence among several reputable agencies on the central figure of €10 billion strengthens the reliability of this data point, even as precise breakdowns vary from source to source. For an industrial announcement of this size, such convergence on the central order of magnitude, despite diverging details, is fairly reassuring about the solidity of the initial figure put forward by the European Commission itself, especially given how often first-day figures in large public programs get revised weeks or months later.

Where the figures still diverge

Politico, for its part, mentions a first billion already committed, topped up by an additional four billion, a breakdown different from the one put forward by Le Monde, which cites a figure of five billion coming directly from the EU budget. These presentational gaps are not necessarily contradictory: they may reflect different budgetary scopes, but none of the sources reviewed allow us to reconcile them with certainty. Three serious newsrooms, three breakdowns of the same figure.

Euronews details the format of the seven announced centers

A territorial distribution still to be clarified

According to Euronews, the EU is opening a call for seven "gigafactories" meant to train the next generations of AI technologies. None of the sources reviewed yet specify in which member states these facilities will be located, a decision that will depend on the outcome of the tender closing in November. Several member states that already have advanced energy or digital infrastructure could logically position themselves as natural candidates, but no source reviewed confirms any official move in that direction at this stage, and the Commission's own announcement stops short of naming any frontrunner country.

A disagreement on the number of sites, according to Politico

Politico describes a different configuration, with four small hubs and three large centers, a breakdown not repeated by any other source reviewed. This piece names this divergence rather than arbitrarily settling on one presentation or the other. Seven gigafactories or seven plus four hubs: the count itself remains unclear.

Smaller digital players, the announced big winners

Access promised to the most modest players

This scheme explicitly aims, according to the Commission, to give computing infrastructure access to players that could never finance such capacity on their own: early-stage start-ups, SMEs in digital transition, or university labs conducting fundamental research that is costly in computing power. For these players, access to world-class computing power is often the main barrier to entering cutting-edge AI research, well ahead of talent or training-data availability.

The risk of concentrated access despite the stated intent

No source reviewed details the precise criteria for allocating computing capacity among these different users once the gigafactories are operational. This gray area deserves to be named: infrastructure that is open on paper can, in practice, favor players already best equipped to file complex funding applications. A door open to everyone benefits first those who already know how to walk through it.

From five to seven projects, a budget escalation in a year

A progressive escalation trajectory

The shift from five projects announced in February 2025 to seven today, with an envelope that has nearly tripled over the same period, traces a trajectory of progressive budget escalation rather than a plan fixed from the outset. This kind of upward revision, in barely eighteen months, is rare for a European institution whose budget cycles generally stretch over several years of negotiation among member states. A project that grows every year is adjusting under competitive pressure.

What this evolution does not allow us to claim

No source reviewed explicitly documents the internal reasons for this upward revision between 2025 and 2026. This piece refrains from attributing this evolution to a single cause, for lack of direct proof in the available documents. It is possible that competitive pressure from American and Chinese announcements on their own AI infrastructure played a role, but this remains, given the sources gathered, a reasonable inference rather than an established fact, and readers should treat it as such rather than as a confirmed causal chain.

Silence, or opposition, from private cloud players

A notable absence of reaction in the sources

None of the sources reviewed reports a detailed public reaction from major American cloud providers, potentially competitors or partners in this European project. This silence could be explained by commercial caution, by waiting for the tender's outcome, or simply by the absence, at this stage, of a direct stake for these players. None of these hypotheses is confirmed by the available sources, and this piece treats the absence itself as a fact worth recording rather than a gap to be filled with speculation. The silence of America's cloud giants is not a negligible fact in this dossier.

What this silence could mean going forward

If American cloud providers ultimately join candidate consortia, as technology partners rather than competitors, the "sovereign" scope of the European project would be nuanced. A European gigafactory funded by public money could remain dependent on non-European technology partners, a tension official communications do not resolve and that this piece prefers to name rather than ignore.

Lessons from the 2025 AI Antennas precedent

A network already under way before the gigafactories

In October 2025, the Commission had announced the creation of AI antennas across member states and partner countries, a lighter initiative aimed at democratizing access to existing computing resources. This earlier initiative provides a first indicator of the Commission's ability to actually deploy this kind of scheme, though the sources reviewed do not allow a precise assessment of its results to date. An antenna often precedes the tower; the tower still has to get built.

What the absence of a numerical record means here

None of the sources gathered for this dossier provides a numerical record of these AI antennas after nearly a year of operation. This lack of follow-up data calls for caution about the Commission's ability to turn a budget announcement into a measurable operational outcome within the stated timeline.

The French precedent from the primary French-language source

Le Monde documents a rapid ramp-up

Le Monde, in its English-language edition, reports that Europe "commits €5 billion to fund seven megafactories" to catch up with the United States and China, with a figure that, according to this source, comes directly from the EU budget rather than a combined arrangement with national funding mentioned elsewhere. This methodological precision from Le Monde is useful: it distinguishes strictly community-level funding from the national co-financing that other sources aggregate into a single figure, which partly, though not fully, explains the discrepancies found among agencies. The same project, told by three serious sources, produces three different breakdowns of public funding.

Why this French-language source remains valuable despite the gap

This primary French-language source confirms that the information circulates independently in the French press, with a narrative angle centered on the technology race rather than solely on budget mechanisms, which enriches understanding of the dossier without erasing the numerical divergence that must be named.

The documentary limits weighing on this dossier

A dossier still largely prospective

This dossier rests on eight converging sources on the existence and general scale of the project, but diverging on the exact breakdown of public funding. No source reviewed provides the full text of the tender nor the detailed selection criteria that will be applied to applications received by November 12. This lack of complete technical detail limits any outside observer's ability to assess, at this point, which types of consortia stand the best chance of being selected, or how heavily the Commission will weigh technical merit against geographic balance among member states.

What must remain conditional until selection

This piece treats as conditional everything related to actual implementation: selected sites, winning consortia, compliance with the 18-month timeline after signature. An announced tender is not a project under way, a distinction this dossier maintains from start to finish. Between the press release and the poured concrete lies an entire tender process to get through.

Why this project deserves to be tracked beyond the announcement

A test of European execution capacity

This project is not just a budget matter: it is a real-world test of the European Union's ability to turn a political ambition into operational physical infrastructure, in a sector where execution speed matters at least as much as the amount invested. The tight timeline — selection in early 2027, commissioning 18 months later — leaves little room for the usual delays of major European infrastructure projects, whether high-speed rail lines, cross-border power grids, or joint space programs.

What failure or success would mean for the future

If this timeline holds, Europe would have, by 2028 or 2029, significantly strengthened AI computing capacity. If it does not, this delay would feed an already-present debate about the execution gap between European digital ambitions and their concrete delivery, a gap already flagged by several analysts of the European tech sector in recent years. Between the announcement and commissioning, eighteen months remain during which everything could still go off track.

Observers of the European digital sector often note that several ambitious initiatives from previous years, announced with similar enthusiasm, saw significant delays between the initial press release and the actual delivery of the promised infrastructure. Nothing in the sources gathered here allows us to claim the gigafactory project will necessarily escape this pattern, but nothing allows us to predict its failure either: the dossier remains, at this stage, open in both directions. Readers who want to track this story honestly should treat every future milestone — the November tender close, the early-2027 selection, the eighteen-month build-out — as a separate test, rather than assuming today's confident announcement guarantees tomorrow's delivery.

Conclusion: a figure with ambition, an execution still to be proven

Brussels chose a spectacular figure — up to €30 billion mobilized in total — to signal its determination not to fall behind in the global race for artificial intelligence. This communication choice worked: the project was widely picked up by the international financial press within hours, from Reuters to Bloomberg to French-language newsrooms, all drawn by the scale of the announced figure and by its explicitly geopolitical dimension against the United States and China. That speed of pickup, on its own, says something about how starved international coverage currently is for a European counter-narrative to the American and Chinese AI infrastructure race.

But an announced figure is not a built gigafactory, and the gray areas documented here — dependence on foreign chips, unclear allocation criteria, a tight timeline, diverging breakdowns of public funding — are a reminder that the distance between budget ambition and effective digital sovereignty remains, at this stage, largely uncharted. This dossier will keep being tracked at every key milestone: the tender's close in November, the announcement of winners in early 2027, then commissioning within the following eighteen months, with each stage offering a fresh chance to confirm or contradict today's optimistic framing. Thirty billion on paper; the real measure of success will be taken in delivered gigaflops, not press releases.

Sources

Primary sources

European Commission — EU launches AI Gigafactories call to boost Europe's computing capacity

European Commission — AI Gigafactories

European Commission — EU announces AI Factories Antennas across Member States and partner countries

Secondary sources

Reuters — EU aims for seven AI gigafactories with €10 billion plan in race with US, China

Le Monde — AI: Europe commits €5 billion to fund seven megafactories and catch up with the US and China

Euronews — EU opens call for seven 'gigafactories' to train AI technologies

Bloomberg — EU Pledges €10 Billion in Public Funding for New AI Data Centers

Politico — EU readies for a leap forward on AI