SemiAnalysis: Nvidia and memory high margins face 2028 a changing landscape,AI with infrastructure constrained by financing and supply chains

AT A GLANCE

Analysis indicates that financing and supply-chain bottlenecks could put AI infrastructure and the high margins of Nvidia and HBM under pressure in 2028.

Article

BlockBeats reports that October 25, SemiAnalysis founder Dylan Patel and Bernstein analyst Stacy Rasgon recently noted that AI infrastructure investment still has room to grow, but faces the dual challenges of rising financing costs and constrained supply chains. U.S. 10-year Treasury yields surpassed 5%, while yields on unsecured bonds issued by some emerging cloud service providers approached 15%. Non-hyperscale cloud providers, which account for nearly half of construction demand, often lack sufficient balance-sheet support. Projects without long-term purchase commitments (offtake agreement) are finding it increasingly difficult to secure funding, making financing capacity a key dividing line in AI expansion. Meanwhile, capacity for substrates,HBM, optical interconnect components, and clean rooms all face bottlenecks that are difficult to resolve in the short term. Producing the same capacity of HBM requires approximately DDR5 3 to 4 times as many wafers; new clean rooms will gradually come online only in 2027. By contrast, power shortages can be partially mitigated through on-site gas turbines, fuel cells, and other means.

Regarding the core industrial chain, the two analysts warned of a potential turning point in 2028. Nvidia currently continues to support high prices through sustained performance improvements, but Google TPU, Amazon Trainium, and OpenAI’s in-house chip initiatives are providing large customers with alternatives and strengthening their bargaining power. Even if Nvidia’s revenue continues to grow, its market share and profit margins may not rise in tandem. In the memory industry, current HBM gross margins are as high as 88%to 90%, but the market has not assigned an extremely high valuation, with the core concern being the concentrated release of new capacity after 2028. If capacity expansion outpaces incremental demand at that time, compounded by slowing AI demand growth, current high prices and high profits will face a reversal. On the enterprise side, most companies will prefer to purchase Token through cloud services rather than build their own computing capacity, meaning that future demand for AI infrastructure will depend heavily on the utilization of inference services.

Original link https://m.theblockbeats.info/flash/373593

Key points

SemiAnalysis founder Dylan Patel and Bernstein analyst Stacy Rasgon noted that AI infrastructure investment faces rising financing costs and constrained supply chains.
U.S. 10-year Treasury yields surpassed 5%, while yields on unsecured bonds issued by some emerging cloud service providers approached 15%。
Substrates,HBM, optical interconnect components, and clean-room capacity face short-term bottlenecks; new clean rooms will gradually come online only in 2027.
Producing the same capacity of HBM requires approximately DDR5 3 to 4 times as many wafers.
The report cites analysis indicating that Nvidia faces alternatives from Google TPU, Amazon Trainium, and OpenAI’s in-house chips.
The report cites analysis indicating that HBM’s current gross margin is 88%to 90%，2028; the concentrated release of new capacity after
Most enterprises prefer to purchase Token through cloud services rather than build their own computing capacity.

AI-assisted interpretation

The following is analysis, separate from reported facts. Verify important claims independently.

This report relays analysts’ views:AI infrastructure is constrained not only by chip, memory, and clean-room supply, but also by financing costs. The analysts further predict that 2028 if HBM expands capacity faster than demand grows, or if in-house chips strengthen customers’ bargaining power, the related high margins may decline.

Why it matters to readers

AI Whether infrastructure construction can continue expanding depends on project financing, long-term purchase commitments, supply-chain expansion, and demand for inference services. Revenue growth at Nvidia and HBM does not necessarily equate to simultaneous growth in market share or profit margins.

It can be understood as follows:AI demand growth depends not only on how many chips are sold, but also on who can secure financing for construction, whether supply chains can deliver, and whether future demand can keep pace with new capacity.

Risks and unknowns

2028 Forecasts regarding supply and demand changes in
The report relays analysts’ views and does not provide the original interview or the full report.
Google TPU, Amazon Trainium, and OpenAI’s in-house chips were not described in the evidence.
HBM The actual commissioning progress of new capacity and AI growth in inference demand remain uncertain.

SemiAnalysis: Nvidia and memory high margins face 2028 a changing landscape,AI with infrastructure constrained by financing and supply chains

Article

AI-assisted interpretation

Why it matters to readers

Risks and unknowns

Related Developments

Related concepts

Keep following

Security Center