AI’s future: Plotting a path to competitiveness and digital sovereignty

The debate taking place today about digital sovereignty is no longer theoretical. Geopolitical tension, supply-chain shocks and growing cyber risks have persuaded governments and businesses alike that they must be able to control the technologies on which their prosperity depends.

Yet the tools used so far to protect digital sovereignty have often come at a price. Compliance regimes add cost and complexity, “buy-local” rules can deprive innovators of world-class platforms, and generous subsidies have struggled to narrow the gap with a handful of global technology leaders.

In the European cloud market, for example, local providers’ combined share fell from 29% to 15% between 2017 and 2024, while three US-based hyperscalers now account for about 70% of demand. Similar patterns are visible in artificial intelligence (AI), where the largest investments still concentrate in a few very large data centres owned by a handful of corporations headquartered in the United States or China.

The central question, then, is whether sovereignty and competitiveness must always pull in opposite directions – or whether the next phase of the AI revolution can align the two. A growing body of evidence suggests alignment is possible.

The public launch of large language models like OpenAI’s ChatGPT in late 2022 triggered an unprecedented wave of experimentation. Organizations rushed to test generative applications that draft marketing copy, summarise research or write software code. Investment exploded: analysts estimate that more than two-thirds of global AI spending in the past four years was channelled into training ever-larger foundation models, most of them in centralised hyperscale facilities.

Results have been mixed. A recent MIT survey found that barely one in twenty AI pilot projects produces measurable business value. The main reason is structural: a foundation model, however sophisticated, is only one component of a much wider system. The data that really differentiates a bank, a manufacturer or a public health authority is proprietary, highly distributed and often subject to strict confidentiality rules. To unlock that value, AI needs to move closer to the data, adapt to local context and act autonomously within tightly-defined guardrails.

Technologists describe this next stage as “agentic AI”. Instead of a single, monolithic model answering text prompts in the cloud, a network of specialized agents will collaborate, learn from one another and, where necessary, act in real time at the “edge” – on a factory floor, inside a vehicle or within a national security perimeter. In short, AI is becoming a hybrid, multi-tiered application in which centralized training and local inference continuously reinforce one another.

A distributed architecture is not only a technical inevitability; it also reshuffles the geopolitical and economic calculus. Three characteristics are decisive:

First, many tasks that create the greatest added value rely on local, often sensitive data. Fine-tuning and inference therefore happen in environments the data owner controls – an enterprise data centre, a hospital campus or an on-premises micro-data-centre in a rural factory. This arrangement reduces unilateral dependencies and allows organizations to apply their own security, privacy and compliance standards.

Second, latency matters. Autonomous robots, smart grids or algorithmic trading engines cannot wait for a distant data centre to respond. Processing at the edge improves performance and resilience, creating demand for regional infrastructure and specialized connectivity.

Third, energy consumption and sustainability considerations are forcing a rethink of the “bigger is better” philosophy. High-density AI clusters require enormous amounts of electricity and produce significant waste heat. Locating smaller facilities where low-carbon energy is abundant – and where heat can be reused – can reduce both emissions and operating costs.

The result is an AI value chain in which different actors, regions and even individual cities play distinct but interoperable roles:

Because contribution is possible at every layer, countries and companies can specialise according to their comparative advantages: abundant renewable power, advanced manufacturing data, a strong healthcare system or robust regulatory frameworks for sensitive information. Competitiveness and sovereignty, in other words, become complementary rather than conflicting objectives.

For business leaders, the message is clear: the coming wave of AI-enabled value will depend less on controlling a monolith and more on orchestrating a constellation. Competitive advantage will accrue to firms that identify their unique data assets, deploy or partner on local infrastructure where it matters, and integrate seamlessly with global model providers where it does not.

For governments, the priority is to foster an ecosystem in which domestic players can contribute essential nodes – whether in energy-efficient data centres, domain-specific model tuning or edge-AI deployment – without attempting to replicate every component of the stack. Such an approach increases resilience, supports local industry and aligns with climate commitments.

The AI revolution is only beginning its second act. By embracing a distributed, hybrid architecture, societies can capture far greater economic value while simultaneously reducing strategic dependency. In doing so, we move closer to a world in which digital sovereignty and global competitiveness are no longer trade-offs, but two facets of the same, shared objective.