Zhongtai: Emphasize the investment opportunities brought by Google's AI computing industry chain and OCS technology innovation.

Zhongtai releases research report, stating that Alphabet Inc. Class C (GOOGL.US) has fully deployed chips (TPU)-networks (OCS)-models (Gemini)-applications (cloud computing/search/advertising, etc.). The self-developed TPU chips have achieved significant progress, and the Gemini model capabilities are globally leading. The company is also exploring the application of OCS, considering the performance advantages of OCS, the future prospects in data center applications are promising. The report suggests that attention should be paid to investment opportunities brought by the AI computing industry chain of Alphabet Inc. Class C and OCS technological innovation. Zhongtai's main points are as follows: Alphabet Inc. Class C's full-stack advantage builds a moat, with the potential to increase capital expenditures. The deployment of chips (TPU)-networks (OCS)-models (Gemini)-applications (cloud computing/search/advertising, etc.) has realized a significant development in self-developed TPU chips, with Gemini model capabilities leading globally. The high growth demand for Alphabet Inc. Class C's computing power and investment opportunities brought by AI hardware innovation should be emphasized. Alphabet Inc. Class C's 2025 Capex has been raised to 910-930 billion, with the potential to increase capex-to-revenue ratio and operational cash flow ratio compared to other CSPs. It is expected that capital investment will continue to increase in 2026. Gemini has over 650 million monthly active users, with the total monthly Token processed increasing by more than 20 times in a year. Alphabet Inc. Class C's cloud revenue and share continue to rise, with strong demand driven by AI, and sufficient orders in hand. Why should AI data centers use OCS? Optical switching technology first appeared in the telecommunications market and was used in traditional telecommunications networks represented by WSS. The special optical switch in the telecom field is OCS (optical circuit switch), which directly transmits data using optical signals, avoiding the signal conversion process, reducing latency and energy consumption, especially showing great advantages in cross-regional expansion and large-scale AI computing demand. Traditional optical-electronic architecture faces challenges in dynamic performance, power consumption, and cost. OCS is a key solution to meet dynamic computing requirements - core dynamic interconnection technology for server decoupling. When ScaleOut scale expands exponentially, traditional Clos architecture faces significant challenges in power dissipation, wiring, and corresponding capital expenditures; while the traditional data center architecture of "CPU-memory-GPU-storage" exacerbates the above problems. According to Drut, the core technology of server decoupling is to use OCS dynamic photon fabric to achieve long-distance, low-latency transmission of PCIe signals, significantly improving efficiency without the need for multiple hops through switches. Alphabet Inc. Class C's TPU shipment expectations have been raised, and OCS optical switches have broad prospects. In the early days, Alphabet Inc. Class C data centers used Clos topology architecture and introduced OCS to replace the spine switch in the self-developed TPU cluster Jupiter/Apollo architecture. According to SemiAnalysis, Alphabet Inc. Class C's customized OCS network increased the throughput of the entire network by 30%, reduced power consumption by 40%, shortened data flow completion time by 10%, reduced network downtime by 50 times, and reduced capital expenditure by 30%. Currently, Alphabet Inc. Class C's TPU V7 Ironwood is about to be fully launched, with the introduction of 1.6T optical modules. OCS mainly uses MEMS and liquid crystal solutions, and suppliers of core components such as MEMS arrays, fiber arrays, transmit/receive modules, optical devices, yttrium vanadate crystals, optical modules, ring resonators, light sources, etc. are expected to benefit. OCS mainly includes four solutions: MEMS, liquid crystal, piezoelectric, and silicon optical waveguide, among which MEMS and liquid crystal solutions are progressing rapidly, and other solutions are also being technologically optimized. NVIDIA Corporation is also exploring OCS applications, and considering the performance advantages of OCS, the future prospects in data center applications are promising. Risk Warning: AI progress falls short of expectations, global data center capital expenditures fall short of expectations, downstream demand falls short of expectations, OCS industrial technological progress falls short of expectations, risks of domestic and overseas macroeconomic downturns, risks of outdated research reports, etc.