NVIDIA and AMD Expand AI Platform Deployments
NVIDIA and Oracle are expanding AI infrastructure cooperation, planning to scale Vera Rubin system deployment to several GW and officially bringing Vera CPU to Oracle Cloud. The partnership helps NVIDIA extend from GPU to CPU and a full AI computing platform, further expanding product demand; shares have been volatile recently, and institutions suggest buying on dips in batches. AMD and Meta are deepening AI cooperation, with the Helios rack-scale AI system already operating at Meta, and EPYC server CPUs also supporting Muse AI agent services under Meta; real deployment by large customers helps validate its AI platform competitiveness and expand CPU and GPU business opportunities.
Cloud and Memory Demand Strong, TSMC Expands Capacity
Alphabet's last-quarter revenue and profit came in below expectations, but Google Cloud revenue grew 103% YoY to USD 307億, above market expectations; shares pulled back to intra-year lows in the short term, though institutions see cloud business still growing rapidly with medium- to long-term allocation value. On memory, SKHynix reported revenue breaking above KRW 100兆 for the first time and said more than half of 2027 to 2028 capacity has been locked in by long-term contracts, indicating high-end memory supply-demand remains tight, supportive of Micron product pricing. TSMC announced a NT$388億 purchase of AUO's Phase 2 fab and related facilities in Taichung, with the market watching whether it will be used for advanced packaging; AI demand is driving capacity expansion needs, though benefits from the new plant will take time to materialize.
Autonomous-Driving Chips, Servers and Satellites Each Advance
Tesla's Austin Robotaxi fleet is about 415 vehicles according to third-party tracking, with plans to enter Dallas in November; management expects related business to contribute meaningful revenue starting 2027 Q2 at the earliest, while profitability remains to be proven, so a wait-and-see stance is advised. Apple launched new MacBook Pro models with M6 chips, with AI computing speed up to 25% faster than M5, alongside new iMac models with AI performance 4 times faster than M4 models, helping drive Mac replacement demand. Broadcom and Anthropic secured a USD 630億 financing arrangement for procurement of Broadcom AI chips and networking equipment, expected to deliver starting 2027, though part of the financing requires residual asset support. Intel's last-quarter revenue grew 21% YoY to USD 165億, with data center and AI revenue up 78% YoY to USD 73億, though fourth-quarter revenue is guided down to USD 150億 to 160億. The US Federal Communications Commission passed a proposed rule to expand spectrum for satellite direct-to-phone services; if implemented it would favor SpaceX in expanding Starlink services, but spectrum has not yet been allocated, so a wait-and-see stance is advised.