From Lock-in to Leverage – How RISC-V Diffusion Could Rebalance AI Hardware Geopolitics

From Lock-in to Leverage – How RISC-V Diffusion Could Rebalance AI Hardware Geopolitics Lina Braun September 2025 Contents 1 Abstract 2 2 Introduction 2 3 Theoretical Framework 2 3.1 Techno-EconomicParadigm ................................. 2 3.2 Neorealism .......................................... 3 3.3 LiberalInstitutionalism.................................... 3 3.4 Constructivism ........................................ 4 4 Status Quo: NVIDIA’s dominance 4 5 Emerging Challengers and RISC-V 5 6 Geopolitical Implications 5 6.1 U.S.leverageanderosion................................... 5 6.2 China’sposition........................................ 6 6.3 EuropeandMiddlePowers.................................. 6 6.4 Global system: risks of fragmentation versus pluralism . . . . . . . . . . . . . . . . . . . 7 7 Conclusion 7 8 Declaration of Originality 8 1 1 Abstract The potential of emerging chip design challengers to redistribute technological and geopolitical power is a significant shift that the audience should be aware of. This paper empirically explains how the diffusion of RISC-V could lead to a techno-economic paradigm shift away from CUDA and GPUs, particularly in terms of the distributional effects it will have on U.S. control, China’s adoption, and European technological sovereignty. Furthermore, an open ISA will lead from a neorealistic and liberal institutionalist perspective to portability, resulting in lower switching costs, a broader ecosystem, and a declining capacity for weaponized interdependence, as well as shifting leverage away from proprietary nodes based on standards. 2 Introduction The question of whether emerging challengers will genuinely redistribute technological and geopolitical power is not just a technical debate, but about how control over AI chips shapes power relations. The potential of emerging challengers to redistribute technological and geopolitical power is a significant shift that the audience should be aware of. The control over the "brains" of Artificial Intelligence will shape tomorrow’s world order, as the battle over chips increasingly reflects opposing values: openness versus closed spheres, inclusivity versus monopolization, and sovereignty versus dependency. It is not only about ChatGPT or autonomous vehicles, but chips are a strategic resource tied to defense systems, health innovations, and surveillance infrastructures. NVIDIA’s market capitalization surpassing $2 trillion in 2024, the U.S. export controls of advanced chips to China, as well as Europe’s drive for technological sovereignty, show the critical importance of understanding and shaping the future of AI chip control. Although these developments may alter the way we live tomorrow, political science has not yet fully explored the intersection of emerging chip design challengers and open-source architectures through the lens of international relations theories. Through the application of theories such as the techno-economic paradigm, neorealism, liberal institutionalism, and constructivism, the paper empirically explains the phenomenon of emerging challengers, with the AI hardware startup Tenstorrent serving as a primary example. This paper presents a theory-guided explanation with an x-centered focus on how RISC-V diffusion (x) affects CUDA-based leverage (y), as well as its geopolitical influences on U.S. control, China’s adoption, and the European technological sovereignty. 3 Theoretical Framework 3.1 Techno-Economic Paradigm A technological revolution typically manifests as a confluence of innovative technologies, products, and industries, which can catalyze a fundamental transformation of the economy as a whole, propelling sustained economic growth. Moreover, even the dynamics at the micro level within companies can precipitate significant shifts in the global political economy. This is reflected in a techno-economic paradigm, which is a best practice model comprising a set of universal, generic technological and organizational principles. The paradigm further explains why not only companies but also institutions and states reorganize around certain 2 principles, and how lock-in and institutional adaptation work (Perez 8, 16) Moreover, the paradigm offers a new common sense for companies and policymakers, influencing investment patterns and competitive advantages across various industries. Dominant designs, such as CUDA/ NVIDIA, emerge when one technological standard or architecture becomes the industry benchmark, structuring ecosystems of suppliers, developers, and users (Freeman and Louçã 142–149, 310–325). The potential of open-source architectures, such as RISC-V, to challenge CUDA’s dominance and reshape the techno-economic paradigm of AI hardware, sparking a new era of technological competition and innovation, is a fascinating prospect that could shape the future of the industry. 3.2 Neorealism The question of who controls AI chips is central to the neorealist principle of power, which is defined by the distribution of capabilities among states, rather than by simple control or outcomes. Modern tech and research-intensive resources raise barriers to entry, making only great powers capable of sustaining a high level of technological strength. From a neorealist perspective, chips can be framed as strategic resources, as only states with immense resources can sustain competitiveness. Although states face similar survival threats, they differ in their access to basic materials and high-tech capabilities. In today’s anarchic global system, security still takes precedence over economic gain, meaning that states avoid dependence and seek control over what they rely on. This logic explains why states securitize the supply chains of chips and do not want to depend on rivals for a resource critical to military and economic survival. State power is about maintaining autonomy in this self-help system, with the interdependence on chips creating vulnerability. Great powers try to control or internalize strategic industries for their own interests, but frame it as the common good. For instance, nuclear rivalry demonstrated that technology could perpetuate bipolar stability, as well as that technological advances intensify competition and widen the gaps. To sum up, systematic rivalry emerges when states measure each other’s capabilities in relative terms. Chips fit into this as strategic resources that define the structure of competition, and those who lead in semiconductors set constraints for all others (Waltz 93–97, 105–113, 174–176, 180–184, 191–196) 3.3 Liberal Institutionalism Liberal institutionalism is a theory arguing that under conditions of growing interdependence, international institutions structure interactions, reduce transaction costs, provide information, establish standards, and thereby enable cooperation among states, even in anarchic systems. The role of international institutions in fostering cooperation among states, even in the face of growing interdependence, is a hopeful sign for the potential for global collaboration. Especially, non-state actors interact across borders without governmental authorities overseeing their actions (Keohane and J. Nye, J. S. 11–15). Asymmetric interdependence serves as a negotiating power source, becoming increasingly relevant as cooperative consortia, such as RISC-V, reduce their dependency by pooling resources on proprietary architectures, like CUDA. This shift redistributes power dynamics and fosters new forms of influence. Regime theory is an intermediate approach to the governance of anarchic systems, attempting to explain the evolution of European Union partnerships and the factors that contribute to the success or failure of open consortia. While states often negotiate 3 rules behind closed doors, open consortia like RISC-V operate as inclusive networks rather than exclusive clubs, thereby creating broader legitimacy and inclusivity, despite facing coordination problems. To sum up, when RISC-V consortia reduce transaction costs for compatibility, it allows medium powers to increase their rule-setting capacity relative to considerable powers (Keohane and J. S. Nye 2–11, 33–53). 3.4 Constructivism Constructivism posits that state behavior is shaped by socially constructed meanings, identities, and expectations rather than being solely determined by anarchy. Interactions determine identities and interests, explaining why states react differently towards friends and enemies based on shared narratives. The democratization of AI hardware through open-source RISC-V frame technology as a collective good, while the security imperative portrays it as a threat through competitive, individualistic, and cooperative approaches. The competitive approach views technological gains as zero-sum, while the individualistic approach focuses on absolute gains. In contrast, the cooperation approach involves positive identification and a shared sense of community security. The democratizing hardware narrative aims to shift toward a cooperative system, redefining chips as collective infrastructures rather than mere tools. The potential framing of RISC-V chips within a specific narrative influences whether RISC-V is seen as empowering global development or exposing geostrategic vulnerabilities. Export control regimes, such as U.S. restrictions on advanced chips, are socially constructed practices that can change if a cooperative identity prevails. Ultimately, the governance struggle over AI hardware hinges on competing narratives, with the outcome determining whether chips are viewed as commons of innovations or weapons of rivalry (Wendt 394–402). 4 Status Quo: NVIDIA’s dominance In 2006, NVIDIA released CUDA, a software that enables GPUs to be programmed in a standard programming language, without any reference to graphics at all. GPUs, or Graphics Processing Units, are chips capable of parallel processing, and their primary use case is artificial intelligence (Miller 15, 211). CUDA ties developers to NVIDIA’s GPUs, creating a powerful lock-in effect that builds a developer ecosystem supporting industries ranging from AI to robotics. By 2023, NVIDIA held a dominant share of AI-training accelerators (Yoffie and von Bargen 4–6), with its software CUDA being both proprietary and industry standard (Kong 4–5). Concurrently, the U.S. weaponized chokepoints through export controls on advanced AI chips for China, thereby directly limiting revenue and market access. This illustrates the state power that the U.S. possesses (Kalera et al. 3–4). Moreover, NVIDIA’s technological dominance lowers transaction costs for U.S. allies while increasing switching costs for rivals, as well as legitimizing U.S. rule-making, which illustrates the complex interplay of power and cooperation (Keohane 44–46, 137). A possible paradigm shift could occur when a technological constellation challenges an incumbent dominant design, such as CUDA, during a period of structural crisis (Freeman and Louçã 147–148, 320-325). 4 5 Emerging Challengers and RISC-V Emerging challengers are new or underrepresented suppliers that introduce alternative compute architectures and system designs to compete on performance, efficiency, and cost along the AI value chain (Kachris 1). Their relevance is not merely technical, as in a highly concentrated supply stack where one vendor controls the bulk of training accelerators. Market structure converts into power asymmetries via switching costs, ecosystem lock-ins, and control over complementary assets (OECD 21, 30–35). These frictions make entry politically consequential because they condition who can set standards, exercise leverage over others, and weaponize interdependence. Leverage denotes enforceable bargaining power that arises either from control of proprietary network nodes or from ruleand standards-based governance in interdependent systems (Keohane and J. S. Nye 297) (Farrell and Newman 33–34, 45–46). The emerging challengers also define the counter-conditions – open toolchains and portability, competitive TCO, robust consortia and standards, and chiplet and dataflow complementarities – that must be mitigated for open architectures to matter. At the firm level, innovation is not inherently disruptive, but its impact depends on strategic positioning and the response of the incumbents. Entrants with fewer resources can target overlooked use-cases with "good-enough" performance at lower cost and then move up-market as capabilities improve. At the same time, incumbents bias towards sustaining investments demanded by their core customers (Christensen 8–10, 30). The AI hardware company Tenstorrent, led by the well-known semiconductor designer Jim Keller (Chip War Miller 350) exemplifies such an emerging challenger. It competes on architecture, stack integration, and business flexibility. The startup targets niche, CUDA-focused markets by offering open-source tooling and RISC-V customization that aim to lower switching costs and enable differentiated deployments (Tenstorrent). Each tensix core’s control is based on open-source RISC-V cores. RISC-V is a free, open instruction-set architecture defined as a small base integer ISA that every implementation must include (RISC-V International 10–12); (Thüning). Another example is Cerebras Systems, which created the largest single chip to accelerate deep learning training and inference. Under the current techno-economic paradigm, CUDA-centric GPUs constitute the dominant design, while RISC-V paired with dataflow and chiplet architectures is emerging as a paradigm variant that could, during deployment realignments, open a window for entrants like Tenstorrent (Perez p. 185–187, 199–201). This sets up the case for Tenstorrent’s open RISC-V strategy as a democratizing, potentially disruptive pathway (Christensen 23–24, 88–90, 166). 6 Geopolitical Implications 6.1 U.S. leverage and erosion From a neorealist perspective, the dynamics of U.S. leverage through weaponized interdependence reveal significant challenges that arise from over-reliance on its dominant position. Excessive use of this influence risks alienating allies, potentially undermining the network hegemony the U.S. has established. This gradual erosion of the ability to weaponize interdependence could limit the U.S.’s capacity for global power projection, resonating with neorealist theories that argue alliances may shift in response to the misuse of power. Additionally, the hegemonic stability perspective highlights that a hegemon’s authority becomes 5 precarious when it disrupts cooperative efforts among states. Despite these challenges, the U.S. is presented with opportunities for diversification, primarily due to its control over key nodes in the financial and communications networks. This dominance is reinforced by domestic institutions like the Office of Foreign Assets Control (OFAC) and the National Security Agency (NSA), which facilitate the creation of coercive strategies that extend beyond conventional military and economic means, aligning with neorealist views of structural power. Furthermore, the interconnectedness of theoretical frameworks is significant. Neorealism posits that network structures shape state behavior in a manner akin to systemic structures, illustrating how network configurations can impact international relations. Conversely, the hegemonic stability theory suggests that while U.S. network hegemony provides a stabilizing order, any abuse of this power could jeopardize its legitimacy and stability, particularly from the perspective of its allies (Farrell and Newman 50, 75, 78–79). 6.2 China’s position China’s response to U.S. export controls has evolved into a complex and strategic approach. DeepSeek’s open-source model is a tactical response to U.S. restrictions. Its openness serves not only as a strategic maneuver but as an essential escape route for China, driven by Xi Jinping’s push for innovation and self-reliance (Williams 153–159). RISC-V can help China navigate technological challenges, positioning RISC-V as geopolitically appealing due to its non-U.S.-controlled status. This segmentation of global tech ecosystems into U.S. and Chinese spheres highlights the concept of tech supremacy as both an economic and geopolitical asset, especially in the semiconductor sector, which is deeply intertwined with the rivalry between these great powers (Rahman 250–254). Although semiconductors are a top priority for China, the nation still relies heavily on foreign, particularly U.S., technological chokepoints. In this environment, RISC-V and other challengers emerged as a significant option, offering an open intellectual property framework that could reduce dependence on restricted U.S. technologies (Kennedy 1–3, 6–7). 6.3 Europe and Middle Powers The European Commission balances complete independence with friendshoring and partnerships, which reflects the liberal institutional logic that sovereignty is strengthened not only by internal capacity but by alliances, rules, and shared institutions that lower vulnerability (Donnelly 135–136). Middle powers can work directly with companies, labs, and peer agencies to build coalitions. By creating or accepting rules, norms, and procedures, governments shape the interdependence of their citizens. Regimes sit between raw power and daily bargaining, precisely where middle powers can punch above their weight by rule design (Keohane and J. S. Nye 18–22). Complex interdependence also explains why Europe relies on institutions and partnerships rather than isolation to secure its sovereignty. Middle powers use niche sovereignty through specialized companies (Donnelly 134–136). Tenstorrent has leveraged these middle powers, for instance, with expansion plans in Cyprus to tap regional talent and incentives frameworks as part of the European build-out (Nicolaou). These partnerships are a tool for sovereignty and resilience. The EU’s Economic Security Strategy highlights the ’promote, protect, partner’ approach, with a focus on investing domestically, securing critical infrastructure, and partnering with allies. In regime complexes, middle pow6 ers exercise niche sovereignty by rule design and public procurement, scaling ecosystems without autarky (Keohane and J. S. Nye 18); (Donnelly 134–136). 6.4 Global system: risks of fragmentation versus pluralism The concept of fragmentation in relation to techno-nationalism and bifurcation is drawing increasing attention. For instance, the EU Chips and Science Act represents a transition from liberalism towards a more nationalist approach, creating a zero-sum environment that threatens the foundations of the open, rulesbased international order, particularly in the context of the U.S.-China semiconductor conflict (Luo and Assche 2–5). This conflict could lead to a divided global technology ecosystem and disrupt essential multilateral forums, resulting in bifurcated supply chains (Woods 1–3). Additionally, geopolitical risks arise from the concentration of technology firms in East Asia, as exemplified by TSMC, which exposes them to vulnerabilities during crises (Peters 1643–1644). From a constructivist viewpoint, narratives surrounding techno-sovereignty and securitization tend to deepen divides and create incompatible geopolitical blocks. For instance, the EU’s emphasis on value-based AI governance showcases how these elements can shape and guide policy decisions. Meanwhile, liberal institutionalists express concern that such fragmentation undermines critical institutions designed to foster international collaboration. Conversely, there are opportunities for institutional diversity, where striking a balance between efficiency and respect for national differences is vital for effective governance. Global rule-making often appears as a regime complex, featuring overlapping, non-hierarchical arrangements that allow both state and non-state actors to collaborate without a centralized authority, thereby promoting innovation and cooperation (Stiglitz and Rodrik 27–31). 7 Conclusion This essay and its research question examined the conditions under which the diffusion of RISC-V could lead to a techno-economic paradigm shift away from CUDA and GPUs, particularly the distributional effects it will have on the U.S., China, and the EU. A paradigm shift is only plausible if all four conditions are met simultaneously, if only some hold, rebalancing not replacement, occurs. The conditions are open toolchains and portability, competitive total cost of ownership, viable consortia/standards, and complementarities through chiplets and data flow. The U.S. would lose proprietary leverage potential through lock-in and chokepoints, but retain residual power as a hub for cloud, talent, and standards. The result would be rather erosion than collapse. China’s position mitigates the pain of sanctions through open ISA and ecosystem investments, but does not automatically gain dominance and keep continued dependencies at key chokepoints. The EU and other middle powers can reduce transition costs and scale ecosystems through procurement, standard-setting, and targeted promotion, thereby creating niche sovereignty. Furthermore, an open ISA will lead from a neorealistic and liberal institutionalist perspective to portability, resulting in lower switching costs, a broader ecosystem, and a declining capacity for weaponized interdependence, as well as shifting leverage away from proprietary nodes based on standards. Boundary conditions could be re-lock-in via software stacks, fragmentation risks, and tooling maturity, which would lead to CUDA staying the dominant design and only partially rebalancing. Finally, the EU and middle powers could shift their policy to pre-commercial procurement and standardization to improve portability. The U.S. could focus on 7 its residual power over standards and talent rather than overextending its coercive instruments. China could also utilize open ISA for risk mitigation, but chokepoints will still create bottlenecks. 8 Declaration of Originality I hereby declare that I have written this seminar paper on my own, have not used any sources or aids other than those specified, and have marked the passages taken from them as such. All aids used are listed in the table below. Regarding the use of Large Language Models (LLM, e.g. Chat-GPT), I have archived the entire LLM interaction history with reference to the submitted work and can present it upon request. This seminar paper has not been presented in this or a similar form in any other course. Name of Aid/ Tool Purpose DeepL Translate Translation Grammarly Grammar and spelling check Scholar GPT Research assistant Consensus Research assistant Munich, 22 September 2025 Works Cited Christensen, Clayton M. The Innovator’s Dilemma: When New Technologies Cause Great Firms to Fail. Harvard Business School P, 1997. Donnelly, Shawn. “Semiconductor and ICT Industrial Policy in the US and EU: Geopolitical Threat Responses”. Politics and Governance, vol. 11, no. 4, 2023, pp. 129–39. https://doi.org/10.17645/pag. v11i4.7031. Farrell, Henry, and Abraham L. Newman. “Weaponized Interdependence: How Global Economic Networks Shape State Coercion”. International Security, vol. 44, no. 1, 2019, pp. 42–79. https://doi.org/10.1162/ isec_a_00351. Freeman, Christopher, and Francisco Louçã. As Time Goes By: From the Industrial Revolutions to the Information Revolution. Oxford UP, 2001. Kachris, Christoforos. “A Survey on Hardware Accelerators for Large Language Models”. arXiv, 2024. arXiv, arxiv.org/abs/2401.09890. Accessed 22 Sept. 2025. Kalera, Garv, et al. NVIDIA and the Future of AI Infrastructure: A Financial and Strategic Analysis of a High-Growth Tech Stock. IB Diploma research paper; 3155 words, Doha British School (IB Diploma Programme), June 2025, https://doi.org/10.2139/ssrn.5327765. Kennedy, Scott, editor. China’s Uneven High-Tech Drive: Implications for the United States. Center for Strategic / International Studies (CSIS), Feb. 2020, Kennedy%20CSIS.pdf. China Innovation Policy Series. 8 Keohane, Robert O. After Hegemony: Cooperation and Discord in the World Political Economy. Princeton UP, 1984. Keohane, Robert O., and Joseph S. Nye. Power and Interdependence: World Politics in Transition. Little, Brown, 1977. Keohane, Robert O., and Jr. Nye, Joseph S., editors. Transnational Relations and World Politics. Harvard UP, 1972. Kong, Matthew. Maintaining NVIDIA’s Leadership in the AI Accelerator Market. IB Business Management Extended Essay, unpublished manuscript. Luo, Yadong, and Ari Van Assche. “The Rise of Techno-Geopolitical Uncertainty: Implications of the United States CHIPS and Science Act”. Journal of International Business Studies, 2023. https://doi.org/10. 1057/s41267-023-00620-3. Miller, Chris. Chip War: The Fight for the World’s Most Critical Technology. Scribner, 2022. Nicolaou, Kyriacos. “US Tech Company Tenstorrent Takes First Step toward Cyprus Expansion”. Cyprus Mail, 13 May 2025, Startups and Innovation, cyprus - mail . com / 2025 / 05 / 13 / us - tech - company - tenstorrent-takes-first-step-toward-cyprus-expansion. Accessed 22 Sept. 2025. OECD. Artificial Intelligence, Data and Competition. OECD Publishing, 2024, www.oecd.org/content/ dam/oecd/en/publications/reports/2024/05/artificial-intelligence-data-and-competition_9d0ac766/ e7e88884-en.pdf. Accessed 22 Sept. 2025. Perez, Carlota. Technological Revolutions and Financial Capital: The Dynamics of Bubbles and Golden Ages. Edward Elgar, 2002. Peters, Michael A. “Semiconductors, Geopolitics and Technological Rivalry: The US CHIPS & Science Act, 2022”. Educational Philosophy and Theory, vol. 55, no. 14, 2023, pp. 1642–46. https://doi.org/10. 1080/00131857.2022.2124914. Rahman, Mizanur. “Tech Supremacy and Economic Warfare: The U.S.–China Trade Battle in the 21st Century”. Multitech Journal of Applied Sciences, vol. 2, no. 4, 2025, pp. 249–74. https://doi.org/10. 59890/k690qt50. RISC-V International. The RISC-V Instruction Set Manual, Volume I: Unprivileged Architecture. Document version 20240411, version 20240411, RISC-V International, 11 Apr. 2024. Stiglitz, Joseph E., and Dani Rodrik. “Rethinking Global Governance: Cooperation in a World of Power”. Cooperation in a World of Power, Open-access chapter; widely cited version superseding early 2021 drafts, Routledge, 2024, https://doi.org/10.4324/9781003571384-3. Tenstorrent. Introduction to Tenstorrent. 2024, Company presentation deck. Tenstorrent.pdf. Thüning, Moritz. “Attention in SRAM on Tenstorrent Grayskull”. arXiv, 2024. arXiv, Moritz.pdf. Waltz, Kenneth N. Theory of International Politics. McGraw-Hill, 1979. Wendt, Alexander. “Anarchy Is What States Make of It: The Social Construction of Power Politics”. International Organization, vol. 46, no. 2, 1992, pp. 391–425. https://doi.org/10.1017/S0020818300027764. Williams, Brandon Kirk. “Winning the Defining Contest: The US–China Artificial Intelligence Race”. The Washington Quarterly, vol. 48, no. 2, 2025. https://doi.org/10.1080/0163660X.2025.2517501. Woods, Dwayne. “When the Chips Are Down: Shifting Policy Signals in a Globalized Technology Regime”. Global Public Policy and Governance, 2023. https://doi.org/10.1007/s43508-023-00074-1. 9