The Internet in 2035: Four Possible Trajectories
A European user no longer has access to quite the same Internet as a Chinese or Iranian user. Certain platforms are available in some countries but not in others, for commercial, geopolitical, or regulatory reasons. But does this mean we can speak of a “splinternet”—a balkanization of networks taken to the extreme?
Julien Rossi, a lecturer in information sciences at Paris 8, does not share this view. The researcher co-coordinates the working group on Internet governance within the CNRS’s GDR Internet, AI, and Society. “We can still speak of a single Internet,” he reassures us. The network has always functioned as a network of networks, capable of understanding one another through common protocols. “What allows the Internet to create a unified transnational public space is the agreement on protocols that enable different machines to communicate across borders. ”
For how long? Julien Rossi, Lucien Castex, Francesca Musiani, and Clément Perarnaud have worked on this topic for the European Parliament and devised four scenarios for the future of the Internet. Four trajectories to trace the threads of current tensions and envision the global network by 2035.
Scenario 1: A global Internet, but one that is increasingly fragmented
The first scenario: that of a contradictory status quo. In 2035, the Internet would remain global and interconnected, but differences in access, regulation, and usage would continue to multiply. “It’s a scenario rife with all kinds of contradictions,” summarizes Julien Rossi.
This fragmentation is already noticeable from the user’s perspective. A French person on WhatsApp, a Chinese person on WeChat, a loyal user of the Apple ecosystem: they all use TCP/IP. Yet their online experiences have nothing in common. “Devices are capable of communicating, but the friction occurs at the application layer: from a sociological perspective, this still creates practical fragmentation,” explains Julien Rossi.
“Walled gardens” have thus shifted the digital frontier: they no longer necessarily separate two countries, but rather different commercial ecosystems. By 2035, AI could further exacerbate this trend. Since the emergence of LLMs, AI agents and bots have been crawling the web on a massive scale, driving up infrastructure costs for the sites they visit.
Content producers can fight back by erecting new barriers against these bots. But the models then risk having access to less information or information that is less up-to-date.
Lucien Castex, advisor to the Director General of Afnic and co-coordinator of the working group on Internet governance and the working group on Internet governance within the CNRS’s GDR Internet, AI, and Society, also highlights the challenges of cultural and linguistic diversity and the discoverability of local content.
In this first scenario, the Internet would therefore remain technically unified, while being crisscrossed by an ever-increasing number of invisible boundaries: depending on one’s country, device, platform, or AI agent used, each person could access a significantly different Web.
Scenario 2: Sovereignty Takes Precedence and Fragmentation Sets In
Second trajectory: “centrifugal trends” accelerate to the point of creating lasting incompatibilities. The pursuit of sovereignty then takes precedence over maintaining a shared global digital space.
North Korea is already taking this logic particularly far, and since 2010, Iran has built a national network that complicates access to certain resources from abroad. Even in this scenario, online payments continue to flow. “This limits most states’ ability to organize a true ‘splinternet.’ The economic cost is too high,” explains Julien Rossi.
But sovereignty issues are also taking hold in Western democracies. In April 2026, the European Commission awarded four sovereign cloud contracts to European providers. This approach also extends to physical infrastructure: submarine cables have become a strategic issue, and since 2025, the DNS4EU resolver has positioned itself as a sovereign alternative to major non-European resolvers.
For Lucien Castex, this quest for autonomy encompasses different realities: “There is a link between control and independence. There is a varying degree of effort to gain control over data, infrastructure, and technologies by managing one’s dependencies.” Sovereignty, therefore, does not necessarily mean fragmentation. However, the accumulation of localization rules and distinct technological choices fuels these “centrifugal trends.”
The Internet still holds together because different actors accept the same technical rules, developed notably by the IETF or the W3C. Governments logically seek to influence these standard-setting bodies. Lucien Castex specifically mentions the debates surrounding Chinese proposals such as “IPv6+,” which allow for the inclusion of more information about data packets and their routes.
The Internet’s fundamental architecture remains intact. “No one has seen fit to simply change the architecture,” notes Lucien Castex. The real turning point would occur if, by 2035, standards or addressing systems became incompatible with one another from one region to another.
Scenario 3: Interoperability Becomes the Priority
A third future takes the opposite path. Faced with the risk of fragmentation, preserving the universality of the Internet becomes a priority. Governments, companies, and standards organizations prioritize interoperability, even if it means giving up certain ambitions for control.
This trajectory would uphold one of the Internet’s fundamental principles: different infrastructures and systems can coexist as long as they continue to speak a common language. Some current policies are already moving in this direction. In Europe, the Digital Markets Act seeks, in particular, to open up certain digital ecosystems and strengthen their interoperability.
But defending a universal Internet also involves trade-offs. Objectives deemed legitimate place demands on the network’s architecture: protecting minors, combating illegal content, or enhancing security. In April 2026, the European Commission thus called on member states to accelerate the rollout of its age-verification app.
“Once you’ve done that, you realize it’s possible to circumvent these measures, for example by using a VPN,” notes Julien Rossi. Public authorities may then be tempted to regulate VPNs or crack down on encryption. “This logic eventually leads, little by little, to stifling individuals’ ability to evade national control measures,” he believes.
Looking ahead to 2035, this scenario therefore raises a question: to what extent should we accept limiting national control capabilities to ensure that the network remains open and interoperable on a global scale?
Scenario 4: Preserving the Unity of the Internet Without Giving Up on Regulating It
The fourth scenario seeks precisely to overcome this dichotomy. Advocated by researchers, it involves preserving the unity of the Internet without abandoning all forms of regulation.
Under this approach, each restriction would be assessed based on its necessity and proportionality, particularly with regard to freedom of expression. “In Europe, this framework could help build a more coherent policy,” says Julien Rossi.
The question, then, is no longer whether to choose between a borderless Internet and fully sovereign digital spaces, but rather to determine which control measures can be implemented without compromising the network’s interoperability.
AI could serve as a prime example. Faced with a proliferation of agents, content producers will need to find ways to protect their infrastructure and data. But an accumulation of barriers could restrict the flow of information, impair models’ access to up-to-date content, and reduce the discoverability of certain languages or cultures.
The same trade-off applies to sovereignty. Localizing certain data, prioritizing European infrastructure, or strengthening control over strategic technologies does not automatically lead to the fragmentation of the Internet. It is when these decisions prevent different networks and systems from communicating that the very nature of the Internet may change.
By 2035, the dividing line may therefore lie less between sovereignty and openness than between measures compatible with a global network and those that ultimately undermine its interoperability.
This text has been translated by an artificial intelligence.
