The EU’s AI Strategy Is a Serious Blunder
and what it should focus on instead.
In a recent newsletter post, I asked if the American AI industry could collapse.
To summarize, I think it can and I wouldn’t be surprised if it will.
The EU is seeking to copy the ‘American approach’ by making a huge bet on scaling large language models through infrastructure investments, but without a detailed, coherent strategy.
In an apparent effort to rival OpenAI’s $500 billion Stargate project, Ursula von der Leyen announced last year that the EU would mobilize up to €200 billion for investments in AI. A new European fund will invest €20 billion in compute superclusters called ‘AI Gigafactories’, financed through an initiative called InvestAI. Additionally, the EU wants to champion ‘trustworthy and responsible AI” which complies with the AI Act, privacy laws, copyright laws and respects the individual’s human rights.
At first sight, the initiative looks like a welcome effort to protect the EU’s sovereignty, boost innovation, and geopolitical leverage. Upon closer scrutiny the project raises uncomfortable questions. Will the AI infrastructure really promote EU sovereignty if it depends on GPUs from NVIDIA? Who will use it, when the only major foundation model provider we have in Europe is Mistral, which is investing heavily in infrastructure of its own? Europe’s AI infrastructure is still trailing significantly behind American and Chinese AI infrastructure; should it invest even more?
Regardless of the InvestAI initiative, AI companies in Europe are facing an uphill battle to compete head-to-head with OpenAI and Anthropic. The EU lacks:
Hyperscalers.
Unified venture capital markets with deep pockets.
Strong political leadership that favors business interest over the general public and democratic norms and rules.
A market-driven regulatory landscape which favors innovation over the rights of individuals.
A culture which embraces a high risk tolerance and accepts potential business failures without stigma.
One single market without cultural and linguistic differences and without national regulatory barriers to scale.
Putting all of those factors together, the uphill battle looks more like a wall. At least if the goal is to create a European answer to OpenAI or Anthropic. The EU Commission’s AI strategy, as laid out in the AI Continent Action Plan, appears naïvely optimistic about (1) Europe’s ability to compete with the leading foundation model labs in the US and (2) the significant financial risks involved in pursuing that goal as a strategy. I am certainly in support of developing “trustworthy and human-centric AI”, but if almost all people continue to use American AI products instead, it won’t matter much. And if the AI bubble should pop for real and leave behind a global financial disaster, all the nice words and good intentions, won’t matter much either.
The core of the American AI strategy is the belief that AI ‘superintelligence’ or ‘AGI’ is within reach, and we will get there with enough computing power, data, and algorithms. It follows from this belief that superintelligence will replace most knowledge work and that AI will make groundbreaking discoveries very frequently and very soon. Such narratives have been pushed by Sam Altman, Elon Musk, and other American tech CEOs for years. What is the core belief guiding Europe’s AI strategy? Well, not anything, really. Apart from blind techno-optimism and common platitudes, the AI Continent Action Plan lacks an independent vision of what the EU wants to accomplish with the unprecedented levels of investment in its AI infrastructure. To be frank, it’s a serious blunder.
Instead of copying the American AI strategy with a European twist (namely the focus on responsibility, ethics and governance in the AI Continent Action Plan), the EU could more explicitly capitalize on the “digital sovereignty” movement. If there is one topic that is trending more among European business leaders than AI, it’s how to replace American tech with national or European alternatives. The mainstream media is failing to pick up on the scale, magnitude, and importance of this movement. The vast majority of European organizations, big or small, public or private, are drafting exit plans and currently discussing how they can replace Silicon Valley tech with alternatives, if they haven’t done so already. These conversations are being had all around the globe, of course, but I know they are prevalent in Europe.
What if, instead of co-opting the modern American dream of superintelligence, the EU decided to prioritize its actual needs? There is a need for European cloud services, operating systems, smartphones, social media and streaming platforms, GPUs, semiconductors, cybersecurity, quantum tech and the list goes on and on. What could happen if the EU mobilized €200 billion to directly support its sovereign digital sector in domains such as these? Investment in the broader digital infrastructure would lead to job creation as well, contrary to underutilized datacenters and job-replacing chatbots. It would seem like a much better place to start.
