Interview
How AI could impact geopolitics
- Global hyperscalers project Capital Expenditure of approximately $1.1 trillion to meet rising generative AI demand.
- The US faces a critical capacity gap, with only 3,000 of 8,000 global data centers located domestically and vacancy rates below 3%, necessitating an additional 35+ gigawatts of baseload power.
- Existing US data centers are largely unsuitable for AI workloads due to prohibitive retrofitting costs and an inability to accommodate high-density engineering requirements.
- US infrastructure cannot maintain global leadership exclusively within national borders and must secure supplemental capacity internationally to meet demand.
- Critical decisions regarding AI infrastructure location must be made within the next 12 to 18 months, with all challenges requiring resolution by the end of the decade.
- Geographic expansion options include the democratic world (Canada, Nordic countries, Australia, France, France), the Global South (Indonesia, Malaysia), or the Middle East.
- Building within the democratic world carries risks of similar domestic power constraints and a lack of proven track records for executing large-scale, rapid infrastructure projects.
- Expansion into the Global South presents a risk of capacity reallocation to China rather than serving US interests.
- The Middle East is identified as the optimal short-term solution due to low energy costs, abundant land, rapid construction speeds, coastal proximity for cooling, sovereign ambitions, and flexible capital.
- Many Middle Eastern nations have demonstrated a willingness to exclude Chinese hardware to secure US chip allocations.
- Long-term geopolitical uncertainty exists regarding Middle Eastern allies (Saudi Arabia, Qatar, UAE) potentially shifting alignment despite their current US-centric AI stance.
- Future AI infrastructure placement will likely be a hybrid of the democratic world and a strategic bias toward the Middle East, potentially intensified by a new administration.
- Chip allocations to the Middle East may be linked to geopolitical concessions, including the normalization of relations with Israel and regional re-architecture.
- US companies are currently in the early stages of establishing Middle Eastern operations, a process expected to be lengthy.
- Protectionist policies toward China, including tariffs on critical minerals for wafer production, are expected to intensify under the upcoming administration and remain a bipartisan priority.
- Immediate policy shifts, such as announcements on Day One, are anticipated to trigger second and third-order effects on the AI supply chain.
- China is striving for an efficiency renaissance in large language models but currently lags due to GPU compute shortages, internet firewall restrictions, and model governance limits.
- China is investing $6.1 billion to create global data center hubs and accounts for roughly one-third of global clean energy investments, including significant nuclear power capacity.
- The US and China compete to influence Middle Eastern nations, which face a binary choice between leaning into US GPU-dependent strategies or Chinese initiatives.
- Current trends indicate all three major Gulf Cooperation Council nations are leaning toward the US regarding AI infrastructure development.