The focus of AI infrastructure is shifting from “whether demand exists” to two more specific questions: whether equipment vendors can maintain selling prices, and whether the supply of key components has already been locked in by customers in advance.
The allocation to risk assets has approached a point where marginal changes warrant more caution. Fund managers’ cash holdings have fallen to 3.5% of assets, the sixth lowest level since 1998.
The bottleneck in AI infrastructure is extending from procuring chips to whether power can be secured and actually used. UBS expects hyperscale cloud providers’ combined capital expenditures on AI infrastructure from 2026 to 2028 to total approximately USD 4.
The risk from disruptions in the Strait of Hormuz is not limited to oil prices. What is more worth tracking is that energy and shipping costs may gradually pass through to food prices via two channels: agricultural inputs and grain transportation.
The first channel is logistics.
The key issue Walmart now faces is not whether its AI tools are appealing, but whether they can sufficiently offset the slowdown in core retail growth. U.S. same-store sales rose only 2.6%, below expectations of 3.8% and down to a six-year low.
Moderna’s Phase III melanoma data has prompted the market to reassess the value of mRNA: it is no longer tied solely to COVID vaccine revenue and may also become a technological path for personalized cancer treatment.
The constraints on AI infrastructure buildout are extending from equity markets’ expectations for growth to bond investors’ assessments of financing costs and debt-servicing capacity.
The competitiveness of TSMC’s advanced processes is beginning to be validated on both fronts: technological progress and customer adoption. A16’s backside power delivery technology has reportedly completed development and validation, and has been described as the industry’s first angstrom-class C…
Trading in semiconductors cooled noticeably that day, but for now it looks more like a positioning adjustment within the technology sector rather than a broad withdrawal of capital from growth assets.
Nvidia is evaluating lower HBM configurations for Rubin Ultra. What is worth watching is not whether AI memory demand has peaked, but rather that memory supply and certification progress may already be influencing the product specifications of next-generation chips in reverse.