Governing the AI Compute Race: A Comparative Analysis of Strategic Partnership and Governance Models in Multinational AI Infrastructure Megaprojects

Between 2023 and 2026, the largest technology firms in the world reorganized themselves around a single scarce resource: artificial intelligence (AI) compute. Combined annual capital expenditure at Meta Platforms and Alphabet alone rose from roughly $60.4 billion in 2023 to $163.7 billion in 2025—a 171% increase—while announced infrastructure commitments reached figures such as Oracle’s reported $300 billion computing arrangement with OpenAI and the $500 billion Stargate consortium target. This dissertation asks how the partnerships assembling this infrastructure are actually governed. Using a mixed-methods comparative case study of six focal cases—the Microsoft–OpenAI equity partnership, the Oracle–OpenAI bilateral contract, the Stargate consortium, the NVIDIA–OpenAI supplier-investor alliance, Meta’s vertical integration, and the aggregate hyperscaler capital-expenditure surge—the study integrates transaction cost economics, resource dependence theory, and network governance theory to explain how firms secure control over strategic compute. Drawing exclusively on public sources disciplined by a nine-point evidence taxonomy and a seven-rank source hierarchy, the analysis combines clause-level governance coding, qualitative comparative analysis, thematic process tracing, and multiplex network analysis. The central finding is that control over AI compute is exercised not through ownership alone but through diverse portfolios of contractual instruments—purchase commitments, intellectual property licenses, rights of first refusal, preferred-partner status, and residual value guarantees—a pattern the study names contractual sovereignty. Microsoft holds material influence over OpenAI through twelve coded instruments despite the absence of a voting board seat, whereas the reported $300 billion Oracle arrangement rests on a single disclosed instrument. Accountability gaps are nearly universal across cases, sovereign and foreign capital plays a structurally significant but governance-limited role, and the November 2023 OpenAI board crisis operated as an accelerant of contractual restructuring rather than a mere disruption. These configurations create distinct challenges for antitrust, energy, and national-security oversight, including control without ownership and the novel problem of circular supplier financing. The study contributes an integrated theoretical account, a replicable evidence-graded method, and an actionable agenda for regulators confronting governance that is deliberately difficult to see.