Page 34 - The Annual AI Governance Report 2025 Steering the Future of AI

The Annual AI Governance Report 2025: Steering the Future of AI

Bifurcation and Localization of Chip Supply Chain: The semiconductor supply chain is
undergoing two major transformations: bifurcation and localization. Bifurcation refers to the
gradual division of the supply chain into two competing ecosystems—one led by the U.S. and
its allies, and the other by China—as both powers seek to reduce mutual dependencies due to Infrastructure Theme 5: AI
national security concerns . In parallel, localization is gaining momentum, with regions like East
127
Asia, Southeast Asia, and Europe investing in domestic semiconductor capabilities to enhance
resilience and reduce external risks.
128
Innovations in Computing Architectures: AI is making chip production faster, cheaper, and more
accurate through automation and predictive analytics, and chips and computing architectures
are being designed for the purpose of AI. DeepSeek’s example shows us how computing
129
architectures might change to adapt to the use case of AI. Furthermore, generative AI presents
130
new possibilities for automating and enhancing different phases of chip design by streamlining
tasks such as architectural exploration, circuit tuning, and layout creation.
131

5.4 Cloud Infrastructure and Access

Public Compute: Public compute initiatives—publicly funded programs that provide access to
computational resources—offer significant benefits in democratizing access to AI infrastructure,
especially as AI becomes central to research, innovation, and national competitiveness. These
programs help democratize access to high-performance computing, allowing researchers,
startups, and public agencies to pursue AI development without relying solely on expensive
private infrastructure. For example, the USA’s National AI Research Resource (NAIRR) pilot
132
and the UK’s AI Research Resource (AIRR) aim to provide equitable access to compute for
133
academic and scientific communities. Similarly, India’s Open Compute Cloud and GPU subsidy
program support domestic companies by reducing costs and offering more flexible, market-
driven access to computing resources. These initiatives not only fuel innovation but also attempt
to bridge the “compute divide” that leaves smaller institutions and less wealthy nations behind. 134

Despite their promise, public compute efforts face several challenges. One is the risk of value
capture, where large private tech firms disproportionately benefit from public infrastructure,
potentially undermining the intended public good. Another issue is coordination: overlapping
efforts within jurisdictions, such as multiple city-levels or state-levels can dilute effectiveness
without cohesive policy frameworks. Additionally, long-term planning is complicated by the
need for flexibility in a rapidly evolving AI landscape; the UK’s cancelled £900 million exascale
supercomputer project illustrates how political uncertainty can destabilize these efforts. 135

127 Center for Security Studies. “The Semiconductor Race: Global Technopolitics and Supply Chain Resilience.”
CSS Analyses in Security Policy, no. 345, ETH Zürich, March 2024. https:// css .ethz .ch/ content/ dam/ ethz/
special -interest/ gess/ cis/ center -for -securities -studies/ pdfs/ CSSAnalyse345 -EN .pdf.
128 Merle, Q. (2024). Chips Supply Chain: bifurcation and localization. ETH Zürich.
129 Ranaweera, J. (2025). How will artificial intelligence reshape the semiconductor industry?: Artificial Intelligence
meets Silicon: The next era of chip Design and manufacturing. IEEE Electron Devices Magazine, 3(1), 12–14.
130 Zhao, C., Deng, C., Ruan, C., Dai, D., Gao, H., Li, J., Zhang, L., Huang, P., Zhou, S., Ma, S., Liang, W., He, Y.,
Wang, Y., Liu, Y., & Wei, Y. X. (2025, May 14). Insights into DeepSeek-V3: Scaling Challenges and Reflections
on Hardware for AI Architectures.
131 Raghuweanshi, P. (2024). REVOLUTIONIZING SEMICONDUCTOR DESIGN AND MANUFACTURING WITH
AI. Journal of Knowledge Learning and Science Technology ISSN 2959-6386 (Online), 3(3), 272–277.
132 National Artificial Intelligence Research Resource Pilot. (n.d.). NSF - National Science Foundation.
133 Shearer, E., Davies, M., Lawrence, M. (2024). The role of public compute. Ada Lovelace Institute Blog.
134 Davies, M., & Vipra, J. (2024). Policy briefing Mapping global approaches to public compute Understanding
the options available to policymakers. Ada Lovelace Institute.
135 Davies, M., & Vipra, J. (2024). Policy briefing Mapping global approaches to public compute Understanding
the options available to policymakers. Ada Lovelace Institute.

25