NVIDIA Becomes the Largest "Landlord" in AI Sector with $5.4 Trillion, Big Model Giants Reduced to "Computing Power Tenant Farmers" - 36 Kr

This article delves into the profound transformation in the current AI industry landscape, highlighting NVIDIA's dominant position as the largest 'landlord' in the AI era, controlling the entire indus... Read original →

NVIDIA’s synergy between GPU architecture and 3nm EUV manufacturing has turned AI infrastructure into a capital-intensive, rentable asset class, forcing large-model firms into high-cost tenancy. Technologically, this accelerates adoption of chiplet and optical I/O to reduce single-supplier risk. Regulatory shifts—U.S. export controls and the EU AI Act—are inflating geopolitical costs, especially for non-U.S. players like Anthropic struggling to secure stable H100 access. In response, AMD and Taiwan, China’s TSMC are co-building an MI300X ecosystem targeting a viable alternative stack by 2027. Over the next 18 months, if OpenAI or Anthropic fail to justify IPO valuations against annual compute bills exceeding $10B, expect MaaS model collapses or consolidation. The real winners will be vertically integrated players controlling advanced packaging capacity and power-chip co-planning.