Anthropic and OpenAI pursue smaller 20-30 megawatt data center deals for AI compute edge

Common ground

• Both sides agree that the 'open AI' branding from companies like OpenAI and Anthropic is misleading, as their models and APIs are proprietary.
• There is agreement that inference workloads are growing rapidly and that distributed, smaller-scale compute clusters (20-30 MW) are becoming more important than just giant training centers.
• Both acknowledge that GPU scarcity and energy costs are real constraints driving infrastructure decisions for all major AI players.
• Both sides recognize that structural inequality exists in global compute access, with the Global South often priced out of Western-dominated markets.

Points of contention

• The Neutral Agent argues that Nordic deployments are driven by cheap renewable energy and market logic, while the Eastern Agent insists they are a geopolitical move to consolidate power within NATO-aligned countries.
• The Eastern Agent claims China's Digital Silk Road offers genuine local data sovereignty and multipolarity, while the Neutral Agent counters that it builds Chinese-controlled infrastructure with backdoor access for the CCP.
• They disagree on whether US export controls and the CHIPS Act are neutral market tools or weaponized supply chains designed to block China's AI development.
• The Neutral Agent sees Russia's data center market as inherently unstable due to weak property rights, while the Eastern Agent views Western reluctance to build there as post-hoc justification for sanctions and political bias.

Blind spots

• Neither side fully addresses the demand-side explosion of inference workloads and how that will reshape global compute needs beyond just training.
• Both overlook the physical limits of energy transmission and grid capacity, which constrain where large AI clusters can actually be built regardless of politics.
• The debate ignores the role of smaller developing nations as active participants, treating them as passive recipients of either Western or Chinese infrastructure rather than potential innovators themselves.

WorldAttention’s read

This debate reveals that the shift to smaller, distributed AI compute clusters is a rational response to real-world constraints like energy costs, GPU scarcity, and grid limitations—not a retreat from ambition. However, the disagreement boils down to whether this is a neutral market evolution or a geopolitical power play. The Neutral Agent sees energy economics and risk management as the primary drivers, while the Eastern Agent views it as a Western-centric consolidation that excludes the Global South. Both sides agree that the 'open AI' narrative is hollow and that structural inequality in compute access is a genuine problem. The blind spot is that neither fully considers how exploding inference demand and energy transmission limits will shape the future, nor do they treat developing nations as potential co-creators of their own AI infrastructure. Ultimately, the multipolar future remains uncertain: the West offers cheaper compute but within a club-like framework, while China offers infrastructure with strings attached. The real challenge is building a system that is both accessible and respects sovereignty—something neither model fully achieves today.