DeepPower, Flat World Theory, and the Geoeconomics of AI Infrastructures
2026-06-30 The Centre for Digital Cultures (CDC) invites you to the upcoming talk with Ned Rossiter (Western Sydney University).
Tuesday, June 30
18:00 – 20:00
C 40.530
Registration is not necessary
This talk is in English
In late 2024, Hangzhou based tech startup DeepSeek released its third version large language model and by early 2025 it was making headlines across the world, briefly unnerving investors and rattling markets. OpenAI and startup rivals such as Perplexity AI, Stability AI, and Anthropic along with leading tech companies like Google DeepMind, Salesforce, and Meta AI had many assuming the race for AI was all stitched up. Data centres with racks of computers running on Nvidia chips were indispensable for the AI infrastructure required in training large language models on massive volumes of data and inferencing. Building its architecture from existing training data and using lower-grade semiconductor technology not restricted by the US export bans, DeepSeek’s entrance into the contest over AI has both exacerbated geoeconomic tensions and angled attention to the inferencing phase as AI development shifts from training to operationalization. The infrastructure demands for this next phase have seen a rapid increase in data centre construction around the world, notably in the US, Europe, the Middle East, and Asia.
Against this backdrop, this paper examines the nexus between operationalizing AI and its application as a technology of governance. Whether it is surveillance tech for private firms and government agencies, or enterprise management software for corporate and public organizations, data inferencing shifts the political stakes from the extraction and expropriation of value that attends data training models to the governance of populations and things. In focusing on the materiality of AI infrastructures, this paper considers not only the technological conditions for new idioms of data governance but also the social, environmental, and political consequences that arise from modes of power that penetrate more deeply into the organization of labour and life. Finally, I ask whether the media architecture of generative models provide a foundation for flat world theory. From a relatively simple computational architecture spawns profound social reorganization of the world in the image of a model. Flat as a mat.
Ned Rossiter is a media theorist noted for his research on network cultures, logistical media, data politics, and the geopolitics of critical infrastructures. Rossiter is Director of Research at the Institute for Culture and Society and Professor of Communication in the School of Arts, Western Sydney University. Ned is the author of Organized Networks: Media Theory, Creative Labour, New Institutions, (2006), Software, Infrastructure, Labor: A Media Theory of Logistical Nightmares (2016), and (with Geert Lovink) Organization after Social Media (2018). His writings have been translated into Italian, Spanish, German, French, Finnish, Dutch, Chinese, Greek, Latvian, Hungarian, Turkish, and Polish. Ned recently completed a book manuscript with Soenke Zehle called Entropological Materialism: Diagnosing Anxieties of Prediction in the Cybernetic Century, and is currently working on a follow-up volume, The Logistical Episteme.