Économie politique de l'IA : approches post-marxistes

Voici un résumé en anglais de la première séance du séminaire « Critiquer l’IA » 2026.

Yesterday, we held the first session of the “AI critique” seminar hosted by the CIS, which I co-organize with my colleagues Marie Garin and Olivier Alexandre. We had the distinct privilege of welcoming two scholars whose work has profoundly shaped critical debates on the political economy of artificial intelligence and its repercussions in international relations: Cecilia Rikap (Associate Professor at University College London.) et Nick Srnicek (Senior Lecturer in Digital Economy at King’s College London).

Cecilia’s presentation built on her two (!) forthcoming books examining corporate power in the digital age, arguing that tech giants like Google, Amazon, and Microsoft dominate not through ownership alone, but by controlling infrastructures (especially the cloud), investments, and research. In her view, AI acts as a “Trojan horse” reinforcing this centralized cloud power, creating dependencies for states and corporations. She outlined a “trilemma” for peripheral regions (including Europe) caught between the US-China arms race and objectives of growth, ecological transition, and digital sovereignty. This dynamic enables “intangible extractivism” and “epistemic totalitarianism,” where corporate priorities shape technology and labor. For her, the state’s role has shifted from that of a steerer of technological innovation to a dependent player subject to corporate planning (a view presented in this recent paper that will be of interest to critical security studies scholars).

Nick, in turn, presented parts of his latest book, “Silicon Empires”. He framed Generative AI as a general-purpose technology , highlighting the strategic challenges of diffusion and value capture. In that regard, he outlined current corporate strategies: the cloud-based approach (Amazon, Microsoft), the frontier-model focus (OpenAI, Anthropic), the conglomerate model (Google, Huawei, both trying to build industry-specific tools), and the open-source hub strategy (Meta, DeepSeek). Geopolitically, he contrasted the US focus on innovation through concentration and profit (exemplified by the centralized “Stargate” project for data centers) with China’s emphasis on widespread diffusion and productivity (e.g. decentralized compute clusters). International strategies also differ: while the US pushes deregulation and total-stack deals (see his recent piece on this particular issue), China engages in multilateral standard-setting. He also made clear that, in the age of AI, there is a growing dissensus in the field of power, both in China and the US.

One note to try to take a step back and reflect on the main theme of the seminar, that is the “forms and modes of circulation of AI critiques.” Our two speakers teach in London and know each other quite well (I noticed that Cecilia was featured in the “acknowledgment” section of Nick’s book, and Nick is often cited in Cecilia’s work). Still, each of their post-Marxist critiques comes from different vantage points. Nick’s intellectual background draws more on post-structuralist (Deleuzian) and post-operaist theory. As for Cecilia’s work, it is deeply informed by Latin American dependency theory and structuralist economics —traditions that emerged from the experience of peripheral capitalism.

That, among other things, might explain some nuance—or even disagreements—that emerged in the discussion. Among those that we talked about was on the view of the state in its relation to the tech industry (Cecilia stressing the growing dependence of US state actors on Big Tech where Nick tends to see more of an alliance, as Nikos Smyrnaios summed in up in the Q&As —an issue that is not only descriptive but also theoretical, and which brought to mind Tim Mitchell’s great paper on The Limits of the State).

Another one related to their normative stance: Nick still endorses the project laid out in the 2013 “Accelerationist Manifesto” he co-authored in 2013, framing his latest books as ways to identify “roadblocks” hampering the advent of a post-capitalist, post-work society; Cecilia, on the other hand, made clear that she was “not an accelerationist”, that work should be emancipatory rather than merely surpassed, overall showing a greater concern for the materiality of advanced AI and its impact on the planet.