Europe Wrote the AI Rules, and the Compute Was Elsewhere
The lesson for South Africa is that sovereignty is built in datacentres, not only in policy documents.
Sign up for my Substack daily AI newsletter here.
See my AI Training course portfolio for corporate Business Leaders here.
Follow me on LinkedIn: https://www.linkedin.com/in/johanosteyn/
Europe has just discovered that writing the world’s strictest artificial intelligence rulebook was not the same as having a seat at the table where it matters. While Dario Amodei of Anthropic, Sam Altman of OpenAI and Elon Musk publicly urge a slowdown of the frontier, and Mark Zuckerberg, the American administration and Beijing dismiss the idea, Europe has been almost entirely absent from the conversation. The European Central Bank has warned bluntly that the continent must choose between economic dependence on American and Chinese models or an accelerated build of its own. The diagnosis is stark: regulation without compute is not sovereignty. It is a rulebook you can only apply to businesses you do not own. For South African leaders watching from a country with even less capacity, the temptation is to treat this as a European story. That would be a mistake. Europe’s dilemma is a preview of our own.
CONTEXT AND BACKGROUND
The picture painted this month is unflattering to a continent used to shaping global standards. Christine Lagarde, president of the European Central Bank, framed the choice starkly: shun the technology and lose growth, or embrace it and become dependent on tools built in the United States and China. Ursula von der Leyen announced she would invite frontier laboratories to a discussion about pacing the technology, and her remarks were quickly drowned out.
A senior expert council chaired by Monika Schnitzer, and including Margrethe Vestager, Philippe Aghion and Yoshua Bengio, published A Transformative AI Strategy for Europe the same week, warning that Europeans are already bearing the economic and social consequences of decisions taken elsewhere.
The infrastructure numbers explain the awkward silence. Europe hosts roughly five per cent of the world’s AI compute capacity, against seventy-eight per cent held by the United States, and even on optimistic projections that share barely rises by the early 2030s. The regulatory machinery is real, with general-purpose AI obligations in force under the EU AI Act and Commission enforcement powers active since August. It rests, though, on hardware someone else owns.
INSIGHT AND ANALYSIS
The uncomfortable insight is that sovereignty is not first a legal category. It is chips, energy and buildings. A jurisdiction can pass any law it likes about how frontier AI should be governed, but if the models, the compute they run on, and the labs that train them all sit offshore, the rulebook is a bill of demands presented to firms that answer to a different flag. Europe has learned this the hard way, and the lesson generalises. Any economy that consumes advanced AI without owning any of its production sits in the same seat, only lower down the table.
I have previously written about this, arguing that model choice is now a geopolitical decision rather than a purely technical one, and that boards must treat AI procurement as a matter of strategic exposure and not simply of cost or capability.
IMPLICATIONS
South Africa faces the same choice as Europe, with a smaller hand of cards. We do not have the industrial base to build frontier models, nor the compute footprint to host them at scale, but we do have policy levers that shape whether we adopt AI on our own terms or simply consume whatever is sent. The response is not to imitate Europe’s rulebook, which even its designers concede is not a global benchmark. It is to invest deliberately in the layers of the stack our economy can actually influence.
Power, not policy, is often the binding constraint here, as it is in Europe. A stable grid, clean wheeling and faster municipal approvals will do more for local datacentre buildouts than another act of Parliament. Beyond the grid, sensible sovereign or trusted cloud for sensitive workloads, contracts that assume disruption and secure data residency, and open-weight alternatives that can be self-hosted if a foreign supplier is throttled or sanctioned form the rest of the answer. Shared regional weight through the African Union, BRICS and the G20 gives that answer a voice.
This is cognitive governance applied to national strategy. Humans, and countries, must remain net contributors to the systems that shape their futures, and accountability for that future cannot be handed to firms we cannot audit or vendors we cannot replace.
CLOSING TAKEAWAY
Europe’s warning to the rest of the world is quiet but unmistakable: you cannot regulate your way to sovereignty if the compute lives elsewhere. Rules matter, and so does the resolve to write and enforce them, but they are the second half of a strategy whose first half is infrastructure, capability and choice. South Africa should read the story clearly. The countries that will govern AI on their own terms are the ones that can, in some meaningful part, build it. The rest of us will be governed by the terms of the firms we license it from, whether we sign an act or not.
Author Bio: Johan Steyn is a prominent AI thought leader, speaker, and author with a deep understanding of artificial intelligence’s impact on business and society. He is passionate about ethical AI development and its role in shaping a better future. Find out more about Johan’s work at https://www.aiforbusiness.net
Comments