① Apple and Broadcom are collaborating on the "Baltra" data center processor, expected to launch by 2027, featuring proprietary AI accelerators and HBM memory for Apple Intelligence services.

② The processor may scale up to 8x the CPU/GPU cores of the M3 Ultra, though AI workloads typically require specialized tensor units, suggesting a hybrid CPU-NPU design.

③ Apple’s roadmap includes M5 updates for 2025 iPads/MacBooks, M6 (Komodo), M7 (Borneo), and a high-end Sotra chip, alongside custom processors for wearables like AR glasses.

➀ Nvidia may avoid recent tariffs on its AI servers due to their production in Mexico, which accounts for 60% of their AI datacenter servers;

➁ The tariffs could potentially affect Nvidia's datacenter shipments to U.S. hyperscalers;

➂ Nvidia's DGX and HGX servers are sorted under U.S. Harmonized Tariff Schedule (HTS) codes that are exempt from U.S.-Mexico tariffs due to the USMCA trade agreement.

➀ The ASRock Rack 4U8G-TURIN2 presents a new approach to the classic AI server design, utilizing AMD EPYC CPUs and featuring unique server engineering aspects.

➁ The server is equipped with 2.5" drive bays, NVMe SSDs, and offers flexibility with its various connectivity options and expansion slots.

➂ The review highlights the server's design focus on PCIe expansion cards, power supplies, and rear I/O connectivity, emphasizing its capabilities for high-end card support and redundancy.

AMD is in talks to sell its AI server assembly plants for up to $4 billion, with potential buyers including Taiwan-based companies. The facilities were acquired from ZT Systems, a leading AI server supplier, and the sale is expected to be a significant transaction in the AI server industry.

The US locations of these plants could provide advantages in avoiding tariffs and import restrictions, making them attractive to potential buyers.

AMD's decision to sell the plants is part of its strategy to focus on design expertise and compete with companies like Nvidia in the AI server market.

➀ Nvidia has removed supercapacitor shelves from its GB300 NVL72 design due to technical and mass-production challenges.

➁ The decision may influence CSPs' future adoption of supercapacitors.

➂ Nvidia is exploring alternative energy storage capacitors for power shelves.

➃ Musashi's goal to reach 6.5 million units annually by 2026 is being monitored.

➀ TSMC reports a strong demand for 2nm nodes over 3nm, with A16 attracting AI server clients.

➁ TSMC's 3nm shipments accounted for 20% of total wafer revenue in Q3 2024.

➃ TSMC's 2nm is expected to enter volume production in 2025.

➄ TSMC plans to expand its 2nm capacity to meet strong demand.

➅ TSMC's A16 process is highly attractive for AI server applications.