Europe’s AI Question Isn’t Access. It’s Where the Servers Sit.
July 24, 2026 · DiscreteStack AD
Inside Bloomberg TV Bulgaria’s “Business Start” interview with DiscreteStack CEO Hristo Todorov – and why sovereign AI has stopped being a policy slogan and become a buildable decision
Building sovereign European artificial intelligence with local control is possible. That was the claim DiscreteStack founder Hristo Todorov made on Bloomberg TV Bulgaria’s “Business Start”, in an interview later covered by Investor.bg.
Europe, Todorov said, faces a real risk of becoming a “digital colony” if it stays dependent on technology it does not control. The encouraging part, as he framed it, is that Europe already recognizes this risk – and has spent recent years responding to it with serious investment in both infrastructure and talent.
That reframes the starting point. The question stops being whether Europe can act, and becomes what it is already building.
What Sovereignty Actually Means
The assumption most boardrooms still carry is that technological sovereignty means closing borders, or that Europe must out-race the US and China on their terms to matter at all.
Todorov’s definition is narrower and more useful. Sovereignty, in his framing, is not isolation. It is the ability of states and companies to control the critical technologies their operations are built on. As he put it, the goal is to know exactly what that technology costs – and, within that pricing, to have full certainty about what happens to your data.
It’s Not the Model. It’s Where It Runs.
Todorov drew the distinction that matters most for anyone actually deploying AI inside a business:
“Artificial intelligence isn’t just the model. The real question is how you access it – from an American company’s servers in San Francisco, or from servers sitting in your own office.”
That single line collapses the sovereignty debate into an infrastructure decision. The model a company uses is only half the picture. The other half is jurisdiction: whose terms of service govern the data, whose outage takes down the workflow, whose policy change resets the roadmap overnight.
The Access Risk Is No Longer Hypothetical
Todorov did not treat vendor dependency as an abstract risk. He pointed to a live example: recent restrictions placed on Anthropic’s newest models – evidence, in his view, that access to frontier AI can no longer be treated as guaranteed.
“Dependence on the large American AI companies is becoming an increasingly live issue against the backdrop of current geopolitical dynamics – and restricting access to AI models is no longer science fiction.”
That is the evidence layer under the sovereignty argument. It is not a forecast. It is a data point about how quickly access terms can change when the infrastructure sits outside the customer’s control.
Business Is Already Moving
Todorov said Europe already has alternatives to American AI models, and that business – ahead of the general public – is already acting on them.
“Europe is already prepared. The average consumer may not yet be fully aware what the alternatives are, but business is already turning to them – because of control and because of price.”
Even if access to American AI solutions were cut off, he added, companies would not stop functioning. The European Union, he noted, is already investing actively in building its own infrastructure – a foundation for scientific work, new products, and competitive companies that is not limited to large language models alone.
A Different Model of Being First
Todorov was clear that Europe’s approach is not a slower version of the US or Chinese playbook. It is a different one, built on different priorities.
“The truth is, Europe doesn’t want to catch up – because it’s not about being first at any cost. We come from values that have to address both the problems of society and the problems of business, while in the US and China, being technologically first is the priority at any cost.”
That extends to regulation. The European rules already in place, Todorov said, protect consumers and economies from the merging of technology corporations and states – a boundary that is becoming increasingly blurred in the US and China.
“The goals of technology companies influence state policy – and vice versa – which starts to intertwine things that shouldn’t be intertwined in a democratic society.”
That is not a side comment about ethics. It is a structural argument for why the “own vs. rent” question matters at the level of the state, not just the balance sheet.
The Future Is Choice, Not a Single Model
Todorov was clear that European sovereign AI should not be one model, but a competitive environment with multiple options. Asked why that matters, he explained why the largest providers resist it:
“Choice is something the biggest providers like Anthropic and OpenAI don’t want to give – and it’s very clear why. They’re valued at 20 to 30 times their revenue, which means they have to guarantee their future revenue. That’s exactly the opposite of what business wants. Business wants predictability.”
Where DiscreteStack Fits
Todorov’s “servers in your own office” line is not a hypothetical for DiscreteStack – it is the company’s operating premise. DiscreteStack runs frontier-grade open-weight models on infrastructure the customer controls, with flat-rate licensing per execution node: no data leaving the company’s perimeter, no dependency on a foreign provider’s access terms, no token metering that turns usage into a variable the finance team cannot forecast.
We don’t build models. We make them run.
The Strategic Choice Ahead
Todorov’s argument, in the end, was a simple one: sovereign European AI with local control is not a future ambition. It is already possible – and for the enterprises weighing it, the choice comes down to renting predictability from someone else’s balance sheet, or owning it.
This article draws on ideas and quotes from Hristo Todorov’s interview on Bloomberg TV Bulgaria’s “Business Start”(Volen Chilov, June 19, 2026), also covered by Investor.bg. The original interview and articles are in Bulgarian; quotes have been translated into English for this piece.
