Beyond the U.S.-China Binary: The New Landscape of AI Partnerships in the Global South

Center for Strategic and International Studies | Sabeen Dhanani

Developing nations across the Global South face a severe technology divide, with 2.2 billion people remaining unconnected and 96 percent residing in low- and middle-income countries. This infrastructure deficit limits access to cloud services, reliable power, and advanced compute capacity essential for operating artificial intelligence systems. Constructing resilient technical supply chains requires substantial capital and energy production.

China dominates this landscape by pairing hardware financing with open-weight models like Qwen, which accounted for 41 percent of Hugging Face downloads by early 2026, while offering 5,000 training slots to Global South partners. Compute logistics are failing elsewhere. Gulf states aim to boost data center capacity from 1 gigawatt in 2025 to 3.3 gigawatts by 2030, but a $1 billion data center project in Kenya collapsed in 2026 when the domestic grid failed. Meanwhile, India exports open-source architectures, and middle powers utilize multilateral bodies to establish regional governance frameworks.

Comment

Exporting high-density compute capacity requires physical energy logistics that sovereign capital alone cannot bypass. The failure of the $1 billion G42 and Microsoft data center project in Kenya demonstrates that compute availability is fundamentally constrained by primary power generation and electrical grid capacity. High-performance AI workloads demand uninterruptible megawatt-scale power supply, making data center deployment a problem of baseline utility endurance rather than hardware procurement. When regional electrical grids lack sufficient baseload stability, external capital investments in server architecture remain unviable.

This infrastructure bottleneck shifts strategic advantage toward actors capable of supplying integrated power generation alongside digital hardware. External providers that pair telecommunication equipment with localized energy infrastructure can establish persistent operational footprints where sovereign cloud initiatives collapse. Consequently, technical dominance across East Africa will hinge on sovereign capacity to construct heavy electrical grid infrastructure rather than exporting unbacked software models.

Strategic Question for Discussion

If primary energy grid deficiencies continue to stall mega-projects like the G42 and Microsoft data center in Kenya, will Global South states prioritize sovereign power generation capabilities over cloud software partnerships, or will they accept foreign ownership of utility infrastructure to secure compute access?

The trajectory indicates that developing states will increasingly demand integrated infrastructure packages that combine power generation with compute deployment rather than accepting unbacked software access. While surrendering utility ownership carries significant sovereign risk, the immediate imperative to run advanced foundation models is forcing governments to accept bundled foreign energy and telecom investments. My assessment is that energy-secure actors capable of financing both power plants and server farms will establish long-term technological leverage across East Africa.

Share your assessment in the comments below.

Read the original article