# Nvidia May Face Chip Competition in 2028 as HBM Risks Extend Beyond 2028 By John Kim Canonical URL: https://www.tokenpost.com/news/business/40457 Published: 2026-10-25T09:50:16.000Z Updated: 2026-10-25T09:50:16.000Z Section: Business > Some emerging cloud providers face unsecured debt yields near 15%. HBM capacity growth after 2028 could pressure prices and margins if AI demand slows. AI infrastructure expansion faces tighter funding and component constraints, while chip competition and memory capacity growth after 2028 could reshape the buildout's economics. The U.S. 10-year Treasury yield has topped 5%, while yields on some unsecured debt from emerging cloud providers are nearing 15%. Non-hyperscale cloud companies account for nearly half of AI construction demand but often lack the balance sheets to support it. Projects without long-term purchase commitments, or offtake agreements, are finding financing harder to secure. The financing pressure adds to [broader concerns about AI infrastructure costs](). Substrates, high-bandwidth memory (HBM), optical interconnect components and cleanroom capacity are supply constraints. Producing the same memory capacity with HBM requires about three to four times as many wafers as DDR5, adding to competition for manufacturing capacity. New cleanrooms are expected to begin coming online in 2027. Electricity shortages can be partly eased with on-site gas turbines and fuel cells, while the other supply bottlenecks may be harder to resolve in the near term. Nvidia could face a turning point in 2028. Google's TPU chips, Amazon's Trainium and custom-chip efforts at OpenAI could give large customers alternatives and strengthen their bargaining power. Performance gains continue to support Nvidia's high prices, but its revenue could grow without its market share or profit margins rising at the same pace. Current HBM gross margins stand at 88% to 90%. Capacity additions after 2028 could outpace demand, especially if AI demand growth slows, putting current prices and margins at risk of reversal. Enterprise demand will depend in part on how widely companies use AI inference services. Most enterprises are expected to access those services through cloud providers rather than build their own computing capacity, leaving infrastructure demand tied to inference usage. ## Links in this article - [broader concerns about AI infrastructure costs](https://www.tokenpost.com/news/technology/28095)