What if the real instability of the past four years has not only disrupted tech markets and geopolitics, but also invalidated our usual way of making long-term forecasts? Every few months, events reshape the landscape in ways that would have seemed unlikely just before.
Still, it is hard to resist the intellectual exercise of interpreting the convergence of signals from the past few days.
A recent article offered a particularly sharp perspective on Anthropic's disruptive role. The core argument is architectural: AI coding agents and orchestration capabilities may start to pierce long-standing captive software layers—from legacy stacks to proprietary SaaS ecosystems, e.g. the IBM/COBOL monopoly—and reopen the possibility of more composable, evolutive, and sovereign software architectures. Paradoxically, frontier AI models, often perceived as vectors of dependence, might also become instruments for regaining strategic control over digital systems.
At the same time, the public standoff between Anthropic and the Pentagon—and its latest chapter just days ago—offers a striking illustration of this shift. Among frontier AI actors, Anthropic has arguably invested more than most in a vision of interoperable ecosystems and composable architectures. Even if its models are not open, its architectural philosophy largely is: AI capabilities—models, agents, tools, skills, interoperability standards—are treated as assets that can be orchestrated modularly, usable for some purposes, deliberately constrained for others (e.g. the Pentagon story). This "composable" view of AI deployment remains unusual, and increasingly consequential.
The article's author, Sylvain Rutten, also sketches a geopolitically bold hypothesis: strategic equity participation by a set of European actors in Anthropic. That thought-provoking idea might have sounded far-fetched a few years ago, but recent geopolitical developments have reshuffled many assumptions. The merit of the article is to anchor this hypothesis in concrete orders of magnitude rather than abstract speculation.
Perhaps the deeper lesson is that digital sovereignty does not necessarily mean painstakingly rebuilding every link of the value chain domestically. It may instead consist in enabling strategic rebalancing and architectural composability—conditions under which systems remain substitutable, evolutive, and ultimately controllable.
This echoes a long-standing conviction of mine in enterprise architecture, shaped over more than a decade promoting composable, open-source assembly at Michelin. What has guided us at the application level may turn out to be relevant at a much broader scale: composability may be not only a wise platform architectural principle, but also a strategic principle for global companies and even states to navigate the geopolitical realities of the digital age.
