Zhongtai Securities released a research report emphasizing Alphabet Inc.'s (GOOGL.US) comprehensive AI ecosystem spanning chips (TPU), networking (OCS), models (Gemini), and applications (cloud computing/search/ads). The company’s self-developed TPU chips have achieved leapfrog advancements, while its Gemini model leads globally in capability.
The report highlights Google’s exploration of OCS (Optical Circuit Switching) applications, noting its performance advantages and promising prospects in data centers as the technology matures. Investors are advised to focus on opportunities arising from Google’s AI computing demand and OCS-driven hardware innovation.
Key insights from Zhongtai Securities include: 1. **Google’s Full-Stack AI Edge**: The company’s integrated approach—TPU chips, OCS networking, Gemini models, and cloud applications—strengthens its moat. Capex is projected to rise to $91–93B in 2025, with further increases expected in 2026. 2. **Gemini’s Rapid Adoption**: Monthly active users exceed 650M, with token processing volume growing 20x YoY. Google Cloud revenue continues to climb, supported by robust AI-driven demand and a solid order backlog. 3. **OCS’s Role in AI Data Centers**: Optical switching, initially telecom-focused, offers latency and energy savings by bypassing signal conversion. OCS is critical for dynamic computing needs, enabling efficient PCIe signal transmission without multi-hop switches. 4. **TPU and OCS Synergy**: Google’s OCS adoption in its Jupiter/Apollo TPU clusters has boosted throughput by 30%, cut power use by 40%, and reduced downtime 50x. The upcoming TPU V7 Ironwood, paired with 1.6T optical modules, will further drive demand for MEMS arrays, fiber arrays, and related components. 5. **Industry Momentum**: NVIDIA and others are also exploring OCS, signaling broad data center potential as the technology evolves.
**Risks**: Slower-than-expected AI progress, subdued data center capex, weak downstream demand, OCS tech delays, macroeconomic downturns, and outdated research data.