Economists on AI Platform Market Power & Risks for Africa

Everyone talks about AI disrupting industries. Almost nobody talks about whocontrols the disruption itself. I just published a new research paper that asks an uncomfortable question: What happens when the entire global economy depends on AI services - and those services flow through a bottleneck of 3 to 5 companies? Here's what the data reveals: →One company(NVIDIA) controls 80–95% of AI accelerator hardware. HHI exceeds6,400- nearly 3× the "highly concentrated" threshold. →Three providers(OpenAI, Anthropic, Google) capture ~88% of enterprise LLM spending. Switching costs reach 80–120 engineering hours for deep integrations. →Three hyperscalers(AWS, Azure, GCP) mediate ~70% of all cloud AI traffic - creating a secondary chokepoint most people don't even see. But here's the part that doesn't get discussed enough: These firms don't justsellAI. They alsobuythe inputs - GPUs, talent, training data - as a concentrated oligopsony. They depress input prices through sheer purchasing scale while elevating output prices through ecosystem lock-in. I call this the"AI Hourglass"- a wide base of upstream suppliers funnels through a narrow bottleneck of dominant platforms, then expands back out to billions of downstream users. Value extraction happens at the waist. My paper formalizes this through an extended Cournot oligopsony model with endogenous barriers, and introduces aCross-Layer Concentration Index (CLCI)that captures compounding across the value chain. The result:85.7%of the effective market space is controlled by concentrated structures at one or more layers. Why this matters for Africa and the developing world: Every African enterprise accessing AI today passes through a triple dependency - U.S. hardware monopoly, U.S. foundation models, U.S. cloud infrastructure - all priced in dollars. Zero bargaining power. Zero influence on the technological roadmap. This is the new digital colonialism, and it's happening in real time. The paper proposes five policy interventions: mandatory API interoperability, oligopsony-aware merger review, public compute infrastructure, pricing transparency, and sovereign AI development programs. The AI revolution is transformative. But transformation without competition is just dependency by another name. The full paper is available on my website. I welcome rigorous engagement, disagreement, and dialogue. 📄 Full paper:https://lnkd.in/gmuqiKFA 💬 What's your take?Is the current concentration a temporary growing pain or a structural risk? I'd genuinely like to hear from economists, policymakers, and practitioners. #AI #ArtificialIntelligence #Oligopsony #MarketConcentration #Antitrust #AIGovernance #DataScience #Economics #DigitalEconomy #AfricaTech #AIaaS #CompetitionPolicy #Innovation #Research #OpenSource #SovereignAI #NVIDIA #FoundationModels #CloudComputing #IndustrialOrganization