NextFin News - Europe will need to spend about $3 trillion to build its own digital infrastructure and cut its reliance on foreign technology companies, an estimate that lays bare the depth of the bloc's dependence on American and Asian tech. The figure spans artificial intelligence, data centers, cloud services and software - the full stack that keeps hospitals, power grids and government services running - and it arrives as Brussels rolls out the most ambitious policy package yet aimed at reclaiming technological sovereignty.
The number is not a budget request. It is a diagnosis of a structural gap that two decades of underinvestment, fragmented markets and missed platform shifts have dug across the continent. And it comes with a complication: Europe is trying to reduce its dependence on Washington at the same moment it is aligning more closely with Washington on Beijing.
The $3 Trillion Diagnosis
This is the first judgment the package forces, and it is the one that determines everything that follows: Europe's technology gap is structural, not cyclical. A cyclical shortfall - a supply-chain kink, a temporary funding squeeze, a single missed product cycle - corrects itself. Mean reversion does the work. What Europe faces is a regime-level deficit: a market structure in which US platforms capture the developers, the data and the economies of scale that make the next generation of technology cheaper to build in America than in Europe. That kind of gap does not self-correct. It compounds, because every year of dependence raises the switching cost for the next year. The package is, in effect, an admission that the market will not close this gap on its own - which is why the bill is measured in trillions rather than billions.
The European Commission's European Technological Sovereignty Package, unveiled June 3, 2026, is the policy vehicle behind the spending debate. It bundles two legislative proposals - the Chips Act 2.0 and the Cloud and AI Development Act - with an open-source strategy and a roadmap for digitizing the energy grid. Commission President Ursula von der Leyen framed the stakes directly in a statement accompanying the package:
We cannot afford to depend on others for the technologies that keep our hospitals running, our energy grids stable and our services secure. This is about protecting our citizens, defending our interests and making our own choices.
The dependence the package targets is measurable. US cloud providers control more than 70% of the EU cloud market. The bloc produces less than 10% of the world's semiconductors. Europe imports more than 80% of its digital services, infrastructure and intellectual property, mostly from the United States. And the EU spends €264 billion a year - mostly on US proprietary IT products and services, according to the Commission's own assessment.
The $3 trillion estimate sits within a range of serious calculations. The Commission's package breaks down into component estimates of roughly €120 billion for semiconductors, €200 billion for data centers by 2036, €100 billion for cloud and AI, and €2 billion for open-source software over seven years - with an annual energy investment gap of €400 billion alongside. A separate study by the Center for European Policy Analysis, cited by the Information Technology and Innovation Foundation, estimated in late 2025 that full digital sovereignty would cost the EU €3.6 trillion - about $4.2 trillion - over ten years, or roughly 20% of annual GDP. The $3 trillion figure in circulation is a rounded midpoint of that conversation, not a single line item.
The timing carries its own signal. The package landed two days after the EU signaled it would join Pax Silica, a US-led chip alliance coordinating semiconductor export controls on China. Brussels is simultaneously deepening one transatlantic tie while trying to loosen another.
The Arithmetic of Catching Up
The first thing to understand about the $3 trillion number is that Europe is not chasing a static target. It is chasing competitors whose spending accelerates every year.
Consider the data center math at the center of the Cloud and AI Development Act. The legislation aims to at least triple EU data center capacity within five to seven years; the Commission's energy directorate states the target as tripling capacity by 2035. Current capacity sits around 12 gigawatts, with the Harvard Kennedy School's Belfer Center projecting a target range of 32 to 60 gigawatts and a cost of up to €300 billion by 2036. That is before counting the chips to fill the racks, the software to run them, and the power plants to feed them.
The energy dimension is where the arithmetic turns brutal. The Commission puts the annual energy investment gap at €400 billion. Data centers already consume 22% of Ireland's electricity, and new facilities in saturated hubs face grid-connection rules requiring them to build their own generation. A data center strategy that does not also solve power generation is not a strategy - it is a permitting exercise with a ribbon-cutting ceremony.
And the gap is not only against the United States. The CELIS Institute estimates the investment gap between US and EU ICT companies at roughly $1.36 trillion - not a one-year shortfall but the accumulated result of American firms spending hundreds of billions annually on capital expenditure while European peers scale more slowly. Global investment in data centers nearly doubled since 2022 to reach half a trillion dollars in 2024, per the International Energy Agency. Hyperscalers are on track to borrow around $400 billion in 2026 alone to fund AI infrastructure, more than double the $165 billion raised in 2025, according to Morgan Stanley estimates reported in European business coverage.
The takeaway: the $3 trillion is a moving target measured against rivals whose annual capital expenditure can approach the entire European package in a single year.
Regulation Can Ring-Fence Workloads. It Cannot Build a Cloud.
Brussels's instinct is to legislate, and the Cloud and AI Development Act creates a four-tier sovereignty framework for cloud services used by public authorities. The tiers rank providers on ownership and control, supply-chain dependencies, data processing location and cybersecurity. All public bodies must use at least tier-one services. The most sensitive functions - national security, defense, law enforcement, border management - require the highest tier, which would effectively exclude providers subject to foreign jurisdiction.
Here is the constraint: the Commission estimates only about 1% of Europe's public services are sensitive enough to require the strictest tier. The sovereignty framework protects the crown jewels while leaving the other 99% of public workloads - and the entire private sector - to a market reality in which US hyperscalers are cheaper, deeper and more capable.
The Commission knows rules alone cannot create a European AWS, which is why the package also includes supply-side measures: "data center acceleration zones" promising permit decisions within 12 months, a EuroCloud Federation to pool unused capacity across borders, and procurement criteria favoring providers that demonstrate added value inside the EU. Commission Executive Vice-President Henna Virkkunen has said the goal is to strengthen Europe's "digital autonomy and resilience" without undermining openness to "like-minded partners."
But procurement preference is not technological parity. European cloud providers held roughly 15% of the European cloud infrastructure market as of 2025, according to industry estimates. OVHcloud, IONOS, Deutsche Telekom and SAP exist - but none offers the global scale, AI toolchain or developer ecosystem of the American three. A government buyer can be told to prefer European. A hospital chief information officer running AI diagnostics cannot be told to accept slower models.
Early signs suggest the package formalizes a shift already underway. The European Parliament said it would replace Google with the French search engine Qwant as its default, citing sovereignty and privacy concerns. The Netherlands blocked a US attempt to acquire Solvinity, which operates the Dutch DigiD digital identity system. These are real moves - but they are defensive, and they are small relative to the market being defended.
The takeaway: the package can ring-fence the most sensitive 1% of workloads. On its own, it cannot reverse a 15% market share.
The Second-Order Outcome Nobody Is Pricing In
The first-order read of the package is straightforward: Europe spends $3 trillion, Europe builds sovereignty. The second-order effect is more awkward, and it cuts against the stated goal.
The package's most likely outcome is not that European champions displace American hyperscalers. It is that American hyperscalers build more inside Europe. AWS, Microsoft and Google are already pouring capital into European data centers. Oracle pledged $3 billion for AI and cloud infrastructure in Germany and the Netherlands in 2025, and the American three are all expanding AI-dedicated campuses across the Nordics and Southern Europe. Sovereignty rules that require data to be processed and stored in Europe, under European governance, are entirely compatible with American-owned infrastructure sitting on European soil.
That outcome is not nothing. Local buildout brings jobs, tax revenue, grid investment and some operational resilience. But it is not the sovereignty the package promises. If the $3 trillion ends up financing infrastructure that remains owned and controlled by US firms, Europe has paid for redundancy without buying autonomy.
There is a deeper second-order risk: fragmentation. The four-tier framework, national cloud strategies and member-state acquisition screening create a patchwork of compliance obligations. That raises the cost of operating in Europe for everyone - including European startups trying to scale across the Single Market. The Commission says the package keeps the market open to like-minded partners. The market will judge that claim by how many deals get blocked and how many permits actually issue within 12 months.
The takeaway: the package may produce more American infrastructure on European soil rather than more European infrastructure anywhere.
The Counter-Thesis: Maybe $3 Trillion Is the Wrong Goal
The strongest argument against the package's premise is that full technological sovereignty is neither achievable nor desirable - and that Brussels, in its design choices, knows it.
The evidence is in the architecture. Only 1% of public services fall under the strictest tier. The framework explicitly keeps the market open to like-minded partners. The EU is simultaneously joining a US-led chip alliance. These are not the moves of a bloc trying to decouple. They are the moves of a bloc trying to de-risk - to ensure that in a crisis the lights stay on and hospitals keep running, without pretending Europe can replicate the entire US technology stack.
From this angle, the $3 trillion headline is less a funding target than a negotiating position. It tells member states why they must accept centralized permitting, pooled procurement and Brussels-level coordination. It tells Washington that Europe has options. It tells voters the problem is being taken seriously. The actual policy is narrower: ring-fence critical workloads, accelerate infrastructure deployment, and nudge the market toward European alternatives where they are genuinely competitive - chips, open source, energy digitization.
There is real logic here. Europe does not need a European Google to be sovereign. It needs European control over the systems that matter in a crisis: a sovereign cloud tier for defense and health data, a domestic chip design ecosystem for automotive and industrial AI, and open-source software that cannot be switched off by a foreign court. That is a defensible version of sovereignty. It probably costs a fraction of $3 trillion.
But the counter-thesis has a weakness. De-risking works only if the risk is intermittent - a supply shock, a sanctions episode, a temporary outage. It does not work if the dependence is structural and compounding. Every year that European firms build on US platforms, the switching cost rises. Every AI model trained on American infrastructure embeds that dependency deeper. De-risking today does not prevent re-dependence tomorrow. If the goal is genuine optionality rather than crisis insurance, the spending gap is real, and it is large.
The takeaway: de-risking is cheaper than sovereignty - but it may only delay the bill rather than cancel it.
Who Benefits, Who Is Exposed
The beneficiaries are clearer than the sovereign outcomes. European data center developers, utilities with grid capacity near digital hubs, semiconductor equipment makers and industrial chip designers stand to gain from the buildout regardless of who owns the final infrastructure. The market has already begun pricing this: European semiconductor names have been among the strongest performers of 2026, with Soitec up 371%, AT&S up 330%, Technoprobe up 123%, Aixtron up 116% and STMicroelectronics up 101% for the year, per market data compiled through early August. ASML Holding and Infineon Technologies have each gained more than 60% in 2026. The pan-European STOXX 600 touched a record high in early August, up 10% for the year, though it lagged its North American counterpart.
The exposed are European software and cloud firms that compete head-on with US hyperscalers without comparable scale. Sovereignty procurement helps at the margin, but it does not close a capability gap - and if the package raises compliance costs across the board, smaller European players bear a disproportionate burden. There is also a currency and rates channel: a multi-trillion investment program, even if mostly privately financed, implies sustained capital demand in Europe, which supports the euro's long-run investment case while pressuring European savings to fund domestic infrastructure rather than flow abroad.
What to Watch
Short term (12-18 months): the story is legislative and permitting-driven. Watch whether the European Parliament and the Council dilute the sovereignty tiers, and whether the first acceleration zones are actually designated. Expect volatility in European technology names as the package moves through Brussels.
Medium term (3-5 years): the story is capital expenditure and capacity. The falsifiable test is concrete: does EU data center capacity move measurably toward the triple-by-2035 target, and does European cloud market share rise from roughly 15%? If capacity builds but share stays flat, the package is financing American expansion, not European sovereignty.
Long term (5-10 years): the story is structural. Either Europe has built a self-sustaining digital stack and the spending was the down payment on autonomy, or it has built American-owned redundancy and the dependence has merely changed address.
Three scenarios frame the range. In the base case, the package passes in diluted form; sensitive workloads migrate to certified providers; US hyperscalers expand their European footprints to comply; European cloud share edges up modestly; and a large share of the $3 trillion flows to non-European owners. In the upside case, acceleration zones deliver fast permits, energy investment catches up, and European open-source and chip-design ecosystems scale enough to capture a meaningful slice of the AI infrastructure buildout. In the downside case, fragmentation wins: member states interpret the tiers differently, deals stall on national-security grounds, permits lag, and compliance costs choke the very European startups the package aims to help - so the $3 trillion never materializes because private capital waits for clarity that never comes.
The falsifying signal is specific: if, three years from now, US cloud providers still control more than 65% of the EU market and European data center capacity has not at least doubled, the sovereignty-through-spending thesis has failed. At that point the package will have bought resilience, not independence.
Europe's $3 trillion question is not whether it can afford to spend the money. It is whether spending it buys ownership - or just a more expensive lease on the same American technology.
Explore more exclusive insights at nextfin.ai.
