Europe is finally playing chess with its new AI Continent Plan.
While the US and China have pursued rapid, largely unchecked AI innovation, Europe has applied regulatory brakes. Now, a new plan from the EU suggests that it’s done playing checkers.
What’s happening? In case you missed it amid all the tariff drama: The European Commission unveiled its new “AI Continent Action Plan” today. It stakes out key actions the commission will take to become “a leading AI Continent”.
Given the breakdown in trust, trade and security relations with America, this is the European Union making its own bid for leadership in advanced technology.
Europe’s Liberation Day. Following a push for sovereignty that began with the Draghi Report, deepened with the removal of US security guarantees over the conflict in Ukraine, and crescendoed on “Liberation Day” with punitive tariffs on European goods, the bloc is looking to step to the challenge of building “AI solutions that benefit society and the economy” all on its own.
What’s the aim? Self-sufficiency in terms of infrastructure and funding, as well as independence from foreign providers.
A recent communique to the European Parliament put it thus; “maintain a leading global position in this critical sector, safeguard strategic assets, interests, autonomy and security, and avoid a situation of strategic dependency on non-EU sources.”
How will it work? The plan focuses on five things; compute, data, models, talent and the EU AI Act — the EU’s comprehensive legislative toolkit applying to artificial intelligence technologies — which entered into stage enforcement last year, and will be fully applicable by next year.
What’s new? The plan puts forward several policy proposals, some of which are novel, and some of which build on parallel initiatives like the Digital Networks Act, and the EU Quantum Strategy.
Here’s some of what matters:
Makers, not takers. Similar to the UK’s AI Opportunities Action Plan released a few months ago, the EU plan aims to “stimulate the further development of AI algorithms” in the bloc. The subtext is that relying on imports of cutting-edge models engineered and trained by Chinese or American companies is not a desirable long-term solution.
In short, the EU wants Europe to be a place where foundational innovation in model development can take place, independent of US or Chinese influence.
Gigafactories. Per the plan, the EU intends to finance “large-scale facilities that develop and train complex AI models at an unprecedented scale” furnished with more than 100,000 advanced AI processors. It is betting that the scaling game isn’t dead, that more advanced GPUs will yield capability benefits, and that centralised pre-training of models will offer a big dividend in performance at commercial and scientific tasks.
Distributed supercomputing through “AI Factory Antennas”. The plan includes commitments to set up “AI Factories” that can be accessed by member states that don’t have dedicated supercomputing infrastructure locally. In this way, member states can use powerful supercomputing tools at the startup, scaleup and SME level; meaning that innovative businesses with good ideas in less developed commercial ecosystems can still get the juice they need for experimentation.
Simplified compliance with the EU AI Act. This is the headline grabber. The bloc’s sweeping AI law is viewed as too onerous for European companies to compete with America and China. It has met with widespread criticism for creating a compliance burden on companies building technology that would otherwise be able to move fast and bring experimental products to market.
Help desks. The plan includes launching an “AI Act Service Desk” to troubleshoot compliance and provide companies with self-assessment tools. It also commits to using the AI Board to develop more specific, sectoral guidance for applying the AI Act in different industries where it interacts with existing controls.
Infrastructure is not being ignored. The plan includes adoption of a “Cloud and AI Development Act”, which will seek consultation on a roadmap for building the networks that support artificial intelligence as it becomes more sophisticated. As inferencing by advanced AI models becomes more network intensive, and as usage increases, communications infrastructure will be strained.
Consequently the plan looks to scale up “cloud capacity and sustainable data centres” to support bigger workloads. This is with the aim of ensuring “highly-secure, EU-based” capacity.
Nor is data. A new “Data Union Strategy”, akin to the customs union that standardizes and provides certainty for the movement of goods and services around the bloc, will be established for data sharing; by “enhancing interoperability” and making more, high-quality data available for European companies training artificial intelligence through Data Labs, and pre-existing Common European Data Spaces.
This massive undertaking, and the homegrown cloud software, Simpl, for managing access will be a key distinguishing feature of the plan, and a key asset for the bloc in terms of its competitiveness.
People as well as products. More PhDs, bachelors and masters degrees in key technologies including artificial intelligence will be provided for, per the plan, and an “AI Skills Academy” will spin up fellowship schemes to attract both EU and non-EU candidates.
So, how far behind is Europe in the race? Adoption and use of artificial intelligence has been slower in Europe. This is partly due to constraints and uncertainty created by the bloc’s suite of specific regulations, but it also reflects deeper structural and cultural barriers to establishing the same kind of freewheeling technology ecosystem that is driving the development and application of AI in the US and China.
The stakes are high. Last year, McKinsey suggested that European businesses lagged behind US counterparts in terms of adoption of artificial intelligence by some 45 to 70 percent. Europeans are using AI less. Although the year-over-year increase in usage stands at 23 percent, lagging only China and the US.
European companies account for just 5 percent of global market share in related sectors; raw materials, AI semiconductor design, AI semiconductor manufacturing, and cloud infrastructure and supercomputers.
European investment has also been less intense. Partly due to the undersized venture capital market in Europe. Western Europe’s advanced technology and software sector is a fifth the size of America’s; and its business spend on artificial intelligence is much smaller too, approximately $8.7 billion in the US, versus $2.6 billion in Europe.
European labs have not been as productive. As American or Chinese counterparts in developing cutting-edge models; the most recent Stanford AI Index reports that American institutions produced 40 “notable AI models”, China produced 15; while Europe produced just three.
The investment gap, when it comes to funding development of foundational models, is also wide. American investment exceeded the totals in China, the EU and the UK combined by $25.4 billion last year.
The AI Continent Action Plan says many of the right things. The scale of the European ecosystem, and the underlying strengths of relatively high income per capita, sophisticated scientific and commercial sectors, rule of law and relatively stable immigration policy, as well as an executive function that believes in fairness, decency and justice could well enable it to lead in “its own distinctive approach to AI” aligned with European values.
The upshot? The European Union is signalling that it is prepared to make concessions on its legislative approach, and stump up a lot of funding, in order to accelerate innovation and use; for fear of missing out to competitors in America and China.
The fact that the EU is now taking the need for AI innovation more seriously shouldn’t stop it from tackling other challenges long-highlighted by the business community, including the need to foster deeper, EU-wide markets for growth and risk capital, rethink taxes, and enable more flexible compensation structures for startup companies and entrepreneurs — but it’s a start.
Europe is finally playing chess; the question now — as leading labs experiment with post-transformer and “natural” computing-based AI — who has already moved on to Go?