The EU Commission’s first 100 days
Trump, DeepSeek, transatlantic rifts, and market corrections - how is the EU responding?
Welcome back!
It’s been 100 days since the new Commission officially started their work. Back in December, we laid out some predictions for things to watch out for. Time to take stock.
Much has changed and shifted in these past 100 days:
DeepSeek led to a renewed hope that Europe has a fleeting chance to “compete” on AI (although we’re more sceptical of this changing the underlying market structure on large-scale AI)
Trump came to power, leaving transatlanticism dead, or at least disrupted, with JD Vance siding with the far-right in Germany’s February election in an unprecedented intrusion to European domestic policy-making.
Deregulation is now firmly established as a necessary condition for Europe’s economic survival, taking in the politically and institutionally more palatable part of Draghi's proposal. This leaves the European Union in an awkward position, where there is a pressure backtrack on existing regulations without admitting deference to the US threats. Leevi wrote about this recently with professor Daniel Mügge here.
Investment is still lagging. Draghi aired his frustrations about the political obstacles to realising his proposed €800B cash boost for European competitiveness. However, when it comes to AI, European industry and foreign investment have recently pledged the €150B EU AI Champions initiative (which the Commission topped up with another €50B), and €109B investments in French AI infrastructure by various international investors.
Meanwhile, €800B of public funds have been mobilised, but for defense. Even with the US backtracking on Ukraine military aid cuts, Europe feels the urgent, and existential need to rearm, and to do so quickly.
Markets are correcting, primarily in response to tariff uncertainty, but this could also signal a broader cooldown of a run that’s been fueled by AI hype.
OpenAI is in trouble, after GPT 4.5 didn’t deliver the kind of superintelligence long promised.
NVidia stocks crashed then rebound as expected, signaling the consequences of a highly concentrated chip market for the broader economy.
Mistral has announced a number of corporations with EU Industrial giants, further establishing the French start-up as a European national/European champion.
There has also been a German election and ongoing chaos in France.
So how did these changes impact our projections for the first 100 days?
Back in December, we expected a continued focus on resources for generative AI, promoting AI adoption through government procurement, and more concrete specifics around the new EU Cloud and AI Development Act in the first 100 days.
AI Factories vs. Trump’s Stargate PR
As promised, Virkkunen successfully launched an AI Factories initiative. On Dec. 11 (day 11), the European Commission announced it would contribute half of the planned €1.5B investment into seven European sites, with six additional sites announced on March 12.
The initiative, however, was overshadowed by Trump’s Stargate announcement of $500B for AI infrastructure. Given how frequently the number is cited in European policy circles, it is surprising how few have read the fine-print. Vague future commitments and a clever repackaging of old news suggests this was primarily a political photo op, designed to bolster domestic AI industry and its major players as “too big to fail” while signalling US “dominance” in the AI race.
AI arms race spending spree
Speaking of an AI arms race, in December, we anticipated signals that the arms race narrative will further deepen –- an anticipation that materialized beyond our expectations. The Commission has branded their €50B top up of the €150B EU AI Champions Initiative as a €200BInvest AI Initiative. The expected “CERN for AI” proposal is now baked into this initiative, which includes a new European fund of €20 billion for AI gigafactories. The objective is to enable “all our scientists and companies – not just the biggest - to develop the most advanced very large models needed to make Europe an AI continent.” The gigafactories will have around 100 000 last-generation AI chips, around four times more than the AI factories being set up right now.
The public sector as an enabler and anchor customer
The government procurement as a driver of AI adoption is increasingly becoming a key policy issue in the European Union. Beyond the public computing procurement in the AI (giga)factories, the role of the public sector as an enabler and anchor customer is increasing as a plea from European AI startup funders and founders. This is partly motivated by the pursuit of protected market opportunities that are not dependent on the ecosystems of the American hyperscalers - a trend we explored in our October report. With increasing geopolitical tensions and the capital investments behind AI startups needing markets to grow, expect a momentum towards increased AI adoption by sovereignty-anxious policymakers.
What has been cut: EU Cloud and AI Development Act nowhere in sight
Contrary to our expectations, an EU Cloud and AI Development Act did not happen in the first 100 days and we also couldn’t spot it in the Commission’s work plan for 2025. In Q1, expect the non-legislative “AI Continent Action Plan, and the “Clear Industrial Deal. Q2 is about the “EU Start-up and Scale-up Strategy” (also non-legislative), while the summer is mostly reserved for proposals on the post 2027 multiannual financial framework. Overall, new legislative initiatives seem to be on the backburner, with more focus on the implementation of the ongoing initiatives. To what extent the AI Continent Action Plan gives more guidance on the future, remains to be seen.
Issues we monitored in the first 100 days
Will chasing the specter of a global AI race deliver for the public interest?
When the Commission announced their plans in December, we argued that a key issue to monitor is whether investments in large-scale AI can in fact deliver the kind of economic growth, productivity gains, and increased efficiency that the Commission hopes for.
Our verdict: As we’ve argued before, a key issue with the EU's turn towards industrial policy on AI is its lack of public interest orientation and wholesale strategy. Why should the EU invest in AI? Where in the AI stack should those investments happen? And how exactly will member states, the environment, and people benefit from the investments in AI?
This is still unclear and under articulated, and these questions seem more urgent since the focus has shifted so drastically towards defense. As some have foreshadowed already last year “EU policy making could well redefine ‘competitiveness’ to its lowest common denominator: Europe marshaling the industrial resources to be a big power in a world hurtling toward conflict.” The renascent concerns over existential survival seem to be the political force that can break the existing gridlocks in European politics, where even fiscally conservative Germany, led by a prudent CDU chancellor, is willing to upend the taboos of German politics and unleash public investment and military spending by relaxing the constitutional debt brake, the jarring institutional anchor that has partly prevented the European political economy from reviving since the Eurocrisis.
With public funds now being directed towards defense (and less on AI industrial policy), public-private partnerships and private money is taking a more prominent lead in shaping the direction of large-scale AI investments. It’s been interesting to see how this new dynamic has shifted the focus of investments: While the Commission continued to mobilise funds to develop the capacity to train the most advanced models (Invest AI strategy, AI Factories initiative) with a renewed sense of existential urgency, industry-driven initiatives, such as the EU AI Champions Initiative, for instance, focusses on integrating applied AI into Europe’s industrial base in key sectors like manufacturing, energy, and defense, instead of competing for building the most advanced models.
Does concentrated power in the AI market factor into the Commission’s strategy?
Back in December we wondered whether the Commission would contend with the structure of the global AI market while devising its strategy— specifically Big Tech’s monopolized positions in data, talent, and the cloud. We were especially concerned about a “CERN for AI” that would have essentially redirected billions of public funds towards Nvidia. :
Our verdict: Interestingly, the US election has suddenly put Europe’s infrastructural dependencies on top of the agenda. There is much more awareness about the extent to which Europe relies on US companies that are increasingly aligning their interests with the new administration. Moreover, the Eurostack initiative has been gaining new wind in its sails, with some preliminary moves towards first concrete examples of such initiative.
This realisation, however, hasn’t yet translated into the EU’s strategy on AI, coming with two separate risks: that Europe ends up further increasing, rather than decreasing its infrastructural dependency, or that it will end up replicating the US model of concentrated power in Europe.
A good example is the part of the EU’s InvestAI strategy, akin to the idea for a “CERN for AI.” Given the initiative’s focus on developing the most advanced very large models and purchasing 100 000 last-generation AI chips, a significant share of investment will benefit NVidia, who are able to dictate prices for a quickly depreciating asset in a highly concentrated global chips market. While companies like AMD and Cerebras are making inroads to European markets. Moreover, as we argued in our analysis of Deepseek, even if training models has become more efficient, running them at scale requires a different kind of computing infrastructure that is currently still dominated by US hyperscalers, resting on unparalleled distribution networks, and scalable infrastructures. Hence the shift might imply a shift of emphasis from capital expenses to operating expenses as a key cost item for technology startups.
Mistral is now firmly established as not just a French, but also a European champion. The challenge with this approach is that national champions create dependencies for the policy-makers, whose strategic, high-political ambitions of sovereignty become entangled with the success of private companies. This deepening entanglement between state and corporate power creates new sorts of entanglements and dependencies that start shaping the dynamics in the European political economy.
What does the fine print of the Commission’s Applied AI strategy look like?
Through its Applied AI strategy, the Commission wants to increase efficiency and productivity in both industry and the public sector. We’ve previously called out this premise. Given that blanket AI adoption in government services raises concerns about infrastructural dependencies, single points of failure, and resource waste (and in sensitive social domains, also significant risks for harm) we were wondering about the strategy’s fine print.
Our verdict: We have not seen details about the strategy yet, although the push for AI adoption in the public sector has seen a renewed push. In her political roadmap, von der Leyen promised that the Apply AI Strategy would be developed with “civil society, industry, and Member States.” This remains more important than ever.
How transparently are decisions about the allocation of significant public resources made?
A key question that has become even more pressing, is how transparently decisions about the allocation of public funds are being made. Lessons from the European Chips Act, which attached few conditionalities to public subsidies, and ended up mostly benefitting US company Intel, show that there is significant room for improvement.
Our verdict: The flood of private sector money into building AI infrastructure means that many initiatives are now driven by private sector priorities and not conditioned by public interest priorities–whatever those might be. It also means it’s less likely that any public contributions will be attached to conditionalities, with public money serving as an enabling and derisking role in facilitating investments.
Will the push towards “simplifying” laws lead to deregulation?
The new Commission already marked a significant paradigm shift, away from regulation towards investments. In the past few weeks, this emphasis has further shifted towards what the Commission calls "simplification;” reducing the administrative burden for companies. Reducing unnecessary bureaucracy is a laudable goal, but there’s a fine line between simplification and deregulation that undermines people’s rights, and risks watering down necessary environmental and social protection.
Early in the year Musk’s DOGE campaign has been quite warmly received in Europe. “In general, yes. We need something like this,” EPP’s Axel Voss said in January, in response to the question whether Europe needs a DOGE-style reform. The mood on DOGE seems to be shifting, but its allure as an initiative that will drive efficiency and cut red-tape using AI remains tempting for many policy-makers.
Similar to Europe’s knee-jerk response to Stargate, few seem to be watching more closely. Asking millions of federal workers to describe their week and feeding their responses into an LLM to decide who should get fired, is neither using advanced AI, nor responsible and sustainable integration of AI into the public sector. It’s using the allure of AI to essentially dismantle parts of the US federal government and for removing obstacles standing in the way remaking the American economy according to the interests and ideologies of the Silicon Valley techno-libertarians.
Will the EU’s stricter stance on migration sabotage its AI ambitions?
As the EU shifts further to the right, migration also becomes a more central issue in AI industrial policy. On labour and migration, the Commission wants to fill the skills shortage, reduce “illegal migration,” but also boost the wide-spread adoption of AI across the EU economies, which may replace labor. These goals that may not necessarily align. The promise of AI productivity gains dominates the mainstream agenda on AI and is premised on worker devaluation and displacement with most solutions focusing on reskilling and adaptation.
Other news
Competition and market concentration
The UK Competition and Market Authority (CMA) drops its competition review of Microsoft’s partnership with OpenAI as “available evidence” of the partnership doesn’t qualify for a merger investigation.
The UK watchdog’s inquiry also finds that the market is dominated by Apple and Google’s mobile browsers. A wide probe into the company’s mobile ecosystems has been launched.
EU tech chief counters the US administration’s claim that the Digital Market Act targets Big Tech.
A prolific US start-up accelerator, Y Combinator expresses support of the Digital Market Act urging the Trump administration to revisit their opposition towards the DMA.
(De)Regulation
To support European defense tech capabilities, the Commission intends to propose regulatory simplification as well as a European preference in public procurement. As part of the Omnibus initiative, the EU Justice Minister McGrath has also announced that GDPR will undergo a simplification process.
The push for simplifying AI rules is set on the agenda for the upcoming European AI Board meeting. EU countries and the AI Office will discuss how compliance with the AI Act can be simplified as well as how to target key enforcement issues.
Sifted reports on European splintered tech entrepreneurial groups that call for large roll back on EU regulations and push for European AI whilst mimicking Trump’s pro tech agenda.
The EU Parliament launches reports to assess the impact of AI on the financial sector and copyright to create more clarity on how the AI Act applies to respective industries.
US warns of retaliation if Poland proceeds with tax on big tech.
Defense
OpenAI strikes deal with US government to use its AI for nuclear weapon security. The more widespread use of Large Language Models is creating a new cybersecurity challenge for companies.
The FT asks: Can the US switch off Europe’s weapons? Long hooked on American defence exports, allies feel buyers’ remorse over hardware dependent on Washington support
The European Commission wants to propose regulatory simplification and a European preference in public procurement to boost the EU’s tech defense capabilities.
AI meets the lawyers
French Publisher sues Meta for copyright infringement accusing the tech company over using their books to train its AI systems without authorization.
European start-ups
UK autonomous driving start-up Wayve expands internationally to Germany and the US after it raised $1bn from Softbank, Microsoft and Nvidia last year.
Italian AI start-up Synapsia finalizes a $2.5 billion contract with UAE’s Bold Technologies to offer AI-based city management services for Abu Dhabi.
One of Europe's fastest growing AI start-ups, Loveable, a Swedish company, has raised $15 million.
Board Members of the German AI Association, a lobby group, argue the US administration’s agenda is a gift for European AI in an Op-Ed in Sifted.
More than 120 European tech founders collaborate on ‘Project Europe’ to invest and mentor young entrepreneurs “to change the ‘doom loop’ in Europe.”
AI datacenters
The Polish presidency of the EU Council aims to facilitate setting up AI data centers across the bloc by fast-tracking data center permits to increase AI adoption.
At the Mobile World Congress energy and AI leaders urge that tech must speed up in decarbonising AI production by moving away from powering data centers with fossil fuels.
Trade
The European Commission and South Korea sign a digital trade deal whilst dispute with US administration over the EU’s tech rule book continues.
Digital Public Infrastructure and EuroStack
“Digital Public Infrastructure at a Turning Point.” A new report by Open Future’s fellow Mila T Samdub examines the evolution of the concept of “digital public infrastructure”, tracing a history from an initial phase of openness through the phase of global popularization to a current phase of localization.
If you missed this in February: EuroStack - A European Alternative for Digital Sovereignty. The report outlines the “EuroStack” vision and intoduces a new initiative for Europe's digital future. Commissioned by the Bertelsmann Stiftung, the report was prepared by a team of authors led by Francesca Bria.