The European Laboratory for Learning and Intelligent Systems announced in December 2019 that it had selected 17 cities for a network of AI research units. Located across ten European countries and Israel, the centres represented a planned commitment of approximately €220 million.
ELLIS was created in response to a persistent European concern: the continent produced excellent researchers but struggled to retain them when large technology companies and better-funded institutions elsewhere offered higher salaries, more computing resources and concentrated teams.
The Network Chose a Distributed Model
Rather than place every researcher in one new laboratory, ELLIS established connected units across existing academic and innovation centres. Each unit was expected to begin with roughly six researchers and receive at least €1.7 million annually for five years.
The network model allowed cities to build on local strengths while giving scientists access to a broader European community. Collaboration, exchange programmes and shared standards could create some of the benefits of scale without requiring researchers to leave their institutions.
Distribution also carried risks. Small units could remain fragmented if funding, compute and mobility were insufficient. A recognizable network needed more than a shared name.
Global Demand Was Pulling Researchers Away
Machine-learning expertise was scarce, and American technology companies could offer compensation difficult for universities to match. Industry laboratories also provided large datasets, specialised hardware and the chance to deploy work at global scale.
The issue was not that researchers should never join companies or move abroad. Scientific exchange benefits from mobility. The policy concern was a one-way flow that weakened teaching, public research and Europe’s ability to shape the technology.
When senior academics leave, the effect extends beyond their individual publications. Students lose mentors, universities struggle to build teams and future startups have a smaller local talent pool.
Research Environment Mattered as Much as Salary
Compensation was only one factor in retention. Leading scientists needed reliable long-term funding, freedom to pursue ambitious questions and access to modern computing.
Short grants and administrative complexity can make an academic career less attractive even where salaries are reasonable. ELLIS sought to create an elite research environment with international visibility, allowing scientists to remain in Europe without accepting a narrower professional opportunity.
Open publication was part of the appeal. Yoshua Bengio linked high-quality research in open societies with constructive debate about ethical AI, arguing that humanist values required strong scientific capability.
Canada Offered an Inspiration
ELLIS drew inspiration from the Canadian Institute for Advanced Research and its success in supporting machine-learning communities. European researchers had discussed the initiative at the NeurIPS conference, and ELLIS later signed a letter of intent with CIFAR’s Learning in Machines and Brains programme.
Canada demonstrated how sustained investment in people and networks could generate influence beyond the size of an individual institution. It also showed that early public support could attract private laboratories and startups.
Copying a structure was not enough. Europe needed funding stability, immigration routes and career opportunities suited to its own multinational system.
Industry Support Broadened the Coalition
Companies including Audi, Bayer, Bosch, DeepMind, Porsche and Siemens supported the initiative, alongside international technology groups such as Amazon, Google, NVIDIA and Qualcomm.
Industry participation could supply resources and practical problems, while academic institutions protected independent research and education. Managing that relationship required transparency around funding, intellectual property and publication.
If corporate sponsors gained excessive control over agendas, the network could reproduce the dependence it was intended to counter. If collaboration was too difficult, researchers might still leave to obtain access to industrial-scale systems.
Talent Policy Was Technology Policy
The ELLIS announcement recognised that AI leadership depends on communities, not isolated grants. A country can purchase hardware or launch a strategy quickly, but developing mentors, research culture and trusted collaboration takes years.
Retaining talent also supports responsible governance. Regulators and public institutions need access to independent experts who can evaluate claims made by powerful vendors. Universities provide a place for research whose social value may be greater than its immediate commercial return.
The €220 million commitment was therefore an investment in European capacity and autonomy. Its success would depend on whether the units could offer careers, resources and intellectual freedom comparable with global alternatives.
The larger lesson remains relevant: a region cannot outsource all frontier expertise and still expect to shape how AI develops. Strong open research institutions are part of economic competitiveness, education and democratic oversight at the same time.
