Product 1 publisher2 min readPublished September 14, 2026 at 1:53 AM PDT

Europe's new AI strategy asks for triple the data centre capacity by 2030

A group including Philippe Aghion and Margrethe Vestager puts Europe at 5% of global AI data centre capacity and wants 15% by 2030. The Commission's roughly EUR 30bn attempt to buy capacity is already late.

The Product Desk · Product desk

Illustration accompanying Europe's new AI strategy asks for triple the data centre capacity by 2030

What happened

Europe runs about 5% of the world's AI data centre capacity, and the Transformative AI Strategy for Europe sets 15% by 2030, meaning three times what exists there today.
The strategy comes from a group of senior European economists, policymakers and computer scientists, among them Nobel economics laureate Philippe Aghion and Margrethe Vestager, who ran competition policy in Brussels for a decade.
The compute target carries a condition: companies that build on European soil have to show local communities a benefit.
The Commission has already opened bidding for seven AI gigafactories in a programme worth about EUR 30bn, with French companies putting $10bn behind a single site.
That programme has since run into delays serious enough to alienate the partners it was meant to attract.

Compiled by The Product DeskSomething wrong?How this is made

Why it matters

decision A team signing an EU residency commitment this quarter has to price it against capacity that exists now, because the target sets a share for 2030 and leaves 2027 supply with the same providers as today.
constraint TNW's argument about rented European GPUs caps what any residency promise can mean: the data location changes, and the operator stays under its home jurisdiction.
cost The local-benefit condition lands on the developer, who has to make that case to a planning authority site by site before any of the extra European capacity exists.
contradiction The same document asks for three times the data centre capacity and calls a European frontier lab too expensive, so the capacity Europe wants is for running models built elsewhere.

The easiest row on a security questionnaire is the one asking where the data sits. Pick a European region in the console and the inference runs in Frankfurt. The auditor moves on to key rotation. The company on the other end of the contract is incorporated where it always was. Europe has been trying to buy its way out of that for a year already, according to TNW ^[21].

Tripling only holds if the rest of the world stands still. The strategy's own framing has the global total climbing underneath the target ^[2]. If global AI data centre capacity doubles by 2030, 15% of it works out at six times what Europe operates today ^[20].

TNW describes the document as a warning ^[22]. "We cannot allow Europe to become a playground for foreign powers," said Vestager ^[7]. "From Washington to Beijing, every other major power is treating AI with the seriousness that it deserves," said Aghion ^[8].

The plan keeps the rulemaking and adds the buying. It asks that enforcement of the Code of Practice on systemic risk rest on technical assessment, and that member states shield workers from labour market shocks before the shock arrives ^[19]. Frontier AI expertise should move inside European institutions, so that officials can follow what they are regulating ^[17]. And Europe should lead on assurance technology, meaning hardware whose security and use can be proven, which the authors argue is the price of hosting anyone else's frontier models ^[18].

The plan is a purchase order without funding. Nobody has costed it, and no legislative vehicle is attached ^[15].

A team planning EU-resident inference has to know whether the requirement names a place or an operator. The other thing to know is what breaks, and how fast, if the operator's home government orders it to stop serving you. Requirements that name a place are satisfied by today's European regions, and the 2030 share target is irrelevant to them. Requirements that name an operator are contract work this year: notice periods and exit terms, and what the provider owes if its home government changes the rules.

What to watch

Whether the Commission's next work programme, due within months, names a capacity number.
Whether the seven-gigafactory bidding wins back the partners the delays pushed away, or the programme shrinks.
Whether member states form the alliance of the willing the strategy wants around ASML's lithography chokepoint.

Claim ledger

Ranked by verification strength, evidence, and original report placement.

[1]
Europe runs about five percent of the world's AI data centre capacity.
ReportedSupportedView cited source
[2]
The Transformative AI Strategy for Europe sets a target of fifteen percent of global AI data centre capacity by 2030, which means building three times what exists in Europe today while the global total keeps climbing underneath.
ReportedSupportedView cited source
[3]
The Transformative AI Strategy for Europe was produced by a group of senior European economists, policymakers and computer scientists, including Philippe Aghion and Margrethe Vestager.
ReportedSupportedView cited source
[4]
Philippe Aghion won the Nobel Prize in economics.
ReportedSupportedView cited source
[5]
Margrethe Vestager ran competition policy in Brussels for a decade and served as the European Commission's executive vice-president.
ReportedSupportedView cited source
[6]
Monika Schnitzer chairs Germany's council of economic experts, and Marta Kwiatkowska is a professor of computing systems at Oxford.
ReportedSupportedView cited source
[7]
"We cannot allow Europe to become a playground for foreign powers."
ReportedSupportedSource: Margrethe Vestager, former Executive Vice-President of the European CommissionView cited source
[8]
"From Washington to Beijing, every other major power is treating AI with the seriousness that it deserves."
ReportedSupportedSource: Philippe Aghion, French economist and Nobel Prize laureateView cited source
[9]
The compute target comes with a condition attached: that companies building on European soil show local communities a benefit.
ReportedSupportedView cited source
[10]
The strategy says an alliance of willing member states should exploit the chokepoints Europe already owns in the AI supply chain, ASML's lithography machines chief among them.
ReportedSupportedView cited source
[11]
The European Commission opened bidding for seven AI gigafactories in a programme worth about EUR 30bn, and French companies put $10bn behind one site.
ReportedSupportedView cited source
[14]
The strategy treats access to foreign frontier models as the main task and a European frontier lab as too expensive to attempt.
ReportedSupportedView cited source
[15]
The strategy arrives with no costing and no legislative vehicle attached.
ReportedSupportedView cited source
[16]
The European Commission's next work programme is due within months.
ReportedSupportedView cited source
[17]
The strategy says frontier AI expertise should move inside European institutions, so that officials can follow what they are regulating.
ReportedSupportedView cited source
[18]
The strategy says Europe should lead on assurance technology, meaning hardware whose security and use can be proven, which the authors argue is the price of hosting anyone else's frontier models and the only practical way to check that international red lines are being kept.
ReportedSupportedView cited source
[19]
The authors argue that enforcement of the Code of Practice on systemic risk should rest on technical assessment, and that member states should be shielding workers from labour market shocks now rather than after the shock arrives.
ReportedSupportedView cited source
[21]
Europe has been trying to buy capacity for a year already.
ReportedSupportedView cited source
[22]
TNW says the Transformative AI Strategy for Europe reads closer to a warning than to a white paper.
ReportedSupportedView cited source
[12]
That gigafactory programme has since run into delays serious enough to alienate the partners it was meant to attract.
ReportedInsufficientView cited source
[13]
TNW argues that renting European-hosted GPUs from American providers reproduces the dependency it was supposed to cure, and that Europe's cloud reliance is a political exposure rather than a technical inconvenience.
ReportedInsufficientView cited source
[20]
If global AI data centre capacity doubles between now and 2030, a fifteen percent European share would equal six times the capacity Europe operates today.
DerivedInsufficient1 source— sign in to open

Sources

Publishers with included, body-backed reporting in this cluster.

thenextweb.com
1 article · September 13, 2026
A Nobel laureate and a former EU commissioner say Europe has one year to act on AI

Ask Clarity

Have a question this piece left open? The answer comes from its reporting and claim ledger — nothing else.

Try

Clarity analysis

How the coverage leans and what the record actually supports. Read the methodology

Perspective Coverage

How much of the coverage speaks to builders, operators and investors. It measures emphasis, not whether the story is right.

The Next Web's account gives builders a brief with no schedule. The strategy puts its weight on guaranteed access to foreign frontier models, treats a European lab of that kind as too expensive to attempt, and asks Europe to lead instead on assurance technology: hardware whose security and use can be proven. The authors call that the price of hosting anyone else's frontier models. The report does not say what standard such hardware would be certified against, or who would run the certification. Alongside it sits a recommendation that frontier expertise move inside European institutions, which would have regulators and the labs they oversee hiring from one small pool.

Reality

Evidence, adoption, hype gap, incentives and confidence, each read independently from the current record on the scale shown beneath it. How these are measured.

Evidence44

0100

One outlet's reading of one report

The five percent, the fifteen percent target, the EUR 30bn tender and the $10bn French commitment all reach us through The Next Web's account of a document our coverage never links to. The firmest part of the record is the quotations: Vestager and Aghion are quoted verbatim and named. The weakest is the claim that the gigafactory programme's delays alienated partners; the story never says who is making that claim.

Adoption21

0100

A tender running, a strategy going nowhere yet

The strategy has no uptake to measure: it was written outside the institutions and comes with no price attached. What is actually moving is the earlier gigafactory tender, seven sites and about EUR 30bn on paper with $10bn of French money behind one of them, and The Next Web says that programme is late. The tender has run for a year; the capacity figure has not moved.

Topics and entities

Follow any of these and your For You feed starts watching them — no settings page required.

Topics

Entities

product
Chip and lithography makers have now led two consecutive Mistral rounds
5 publishers
product
Brussels proposes default 70% price cap in EU public tenders, with exceptions
1 publisher
product
Made in EU starts at 5% concrete and says nothing about the software
1 publisher
product
Two-thirds of large organisations have redefined digital sovereignty as resilient interdependence
1 publisher

What happened

Why it matters

What to watch

Claim ledger

Sources

Let Clarity write the post for you.

Topics and entities

Topics

Entities

Related stories

Claim ledger

Let Clarity write the post for you.

Topics and entities

Topics

Entities

Perspective Coverage

Reality

Related stories

Sources