The commercial threshold for robotaxis may not necessarily be reducing the cost per mile to USD 0.25. The more direct comparison is the all-in cost for users to own a private car. According to relevant estimates, this cost is about USD 0.
The supply response in the AI chip supply chain is extending from HBM to advanced packaging substrates. Micron plans to double its HBM capacity year over year, while ZDT plans to expand advanced packaging substrate capacity with TWD 56.
The supply constraints for AI chips appear to extend not only to advanced process manufacturing, but also to testing. TSMC is simultaneously building nearly 20 fabs domestically and overseas at a much faster pace than before, while purchasing manufacturing equipment in advance; on the other end,…
CrowdStrike is attempting to extend Falcon from a traditional endpoint security platform into the security and execution layer for AI agents. This is worth watching because if enterprises use agents for triage, threat hunting, and response, they will gain not only automation efficiency but also n…
New demand for AI servers may be shifting from the training side to the inference side. Dell’s view is that inference demand has already surpassed training, becoming the industry’s pure incremental demand; it expects inference token volume to grow 87-fold by 2030.
MediaTek’s USD 3.9 billion convertible bond financing has pushed NVIDIA and Google further along the partnership chain to become direct investors. NVIDIA subscribed to USD 3.5 billion, and Google subscribed to USD 0.4 billion.
Saudi Arabia’s AI infrastructure buildout is shifting from single-accelerator procurement toward long-term capacity deployment involving cloud, chip, and network vendors.
Memory supply tightness may no longer be limited to the high-bandwidth memory required for AI. The market’s focus is on whether capacity allocation for high-end products is squeezing broader DRAM supply and passing cost pressure on to traditional end markets such as PCs.
If the value of Rubin platform servers rises significantly, HBM4 will no longer be just a single component, but may become a delivery constraint on whether complete system revenue can be realized.
NVIDIA’s single-quarter supply commitments reportedly rose from USD 119 billion to USD 279 billion, with memory procurement considered an important component.