[SemiAnalysis: Nvidia and Memory High Profits Face 2028 Upheaval, AI Infrastructure Constrained by Financing and Supply Chain] BlockBeats reported on October 25th that SemiAnalysis founder Dylan Patel and Bernstein analyst Stacy Rasgon recently pointed out that while there is still room for investment in AI infrastructure, it is facing the dual challenges of rising financing costs and supply chain constraints. The yield on 10-year US Treasury bonds has exceeded 5%, and the yield on unsecured bonds issued by some emerging cloud service providers is approaching 15%. Meanwhile, non-hyperscale cloud vendors, which bear nearly half of the construction demand, often lack sufficient balance sheet support. Projects lacking long-term offtake agreements are finding it increasingly difficult to obtain funding, and financing capability is becoming a significant watershed for AI expansion. At the same time, substrate, HBM, optical interconnect components, and cleanroom capacity all face bottlenecks that are difficult to resolve in the short term. The number of wafers required to produce the same capacity HBM is approximately 3 to 4 times that of DDR5, intensifying capacity competition; new cleanrooms will not be operational until 2027. In contrast, power shortages can be partially alleviated through on-site gas turbines and fuel cells. Regarding the core supply chain, the two analysts warned of a potential turning point in 2028. Nvidia currently supports high prices through continuous performance improvements, but the self-developed chip strategies of companies like Google TPU, Amazon Trainium, and OpenAI are providing alternatives for large customers and enhancing their bargaining power. Even if Nvidia's revenue continues to grow, its market share and profit margins may not increase in tandem. Regarding memory, current HBM gross margins are as high as 88% to 90%, but the market has not given them extremely high valuations. The core concern lies in the concentrated release of new production capacity after 2028. If the expansion rate exceeds the increase in demand, coupled with a slowdown in AI demand growth, the current high prices and high profits will face reversal. On the enterprise side, most companies will prefer to purchase tokens through cloud services rather than build their own computing power, meaning that future demand for AI infrastructure will be highly dependent on the usage of inference services. [Original Link]