🔥Stalecollected in 4m

Shenzhen Pushes 1.6T Optical Modules

Shenzhen Pushes 1.6T Optical Modules
PostLinkedIn
🔥Read original on 36氪

💡Shenzhen accelerates 1.6T optics for AI servers—key infra upgrade

⚡ 30-Second TL;DR

What Changed

Upgrade optical modules: 800G to 1.6T/3.2T for AI servers.

Why It Matters

Enhances China's AI data center capabilities, reducing reliance on imports for high-speed interconnects critical for scaling LLMs.

What To Do Next

Monitor Shenzhen subsidies for 1.6T module prototypes in your AI infra builds.

Who should care:Enterprise & Security Teams

🧠 Deep Insight

AI-generated analysis for this event.

🔑 Enhanced Key Takeaways

  • Shenzhen's initiative aligns with the 'Action Plan for High-Quality Development of the AI Industry (2026-2028)', which specifically targets the integration of optical interconnects into the city's local computing infrastructure to reduce latency in large-scale model training.
  • The policy includes direct financial subsidies for local manufacturers achieving yield rates above 85% for 1.6T optical modules, aiming to lower the cost-per-bit to compete with international suppliers by late 2026.
  • The focus on thin-film lithium niobate (TFLN) is a strategic move to bypass supply chain bottlenecks in traditional indium phosphide (InP) substrates, leveraging Shenzhen's existing semiconductor manufacturing ecosystem to localize the production of high-speed electro-optic modulators.

🛠️ Technical Deep Dive

• 1.6T Module Architecture: Utilizes 8x200G PAM4 DSPs to achieve higher throughput compared to the 8x100G architecture of 800G modules. • CPO (Co-Packaged Optics) Integration: Moves the optical engine closer to the ASIC/GPU to reduce electrical trace length, targeting a power consumption reduction of approximately 30% compared to pluggable modules. • LPO (Linear Drive Pluggable Optics) Implementation: Eliminates the retimer/DSP within the module to reduce latency and power, requiring high-performance SerDes on the host switch/server chip. • Material Advancements: TFLN modulators offer higher bandwidth and lower insertion loss compared to traditional silicon-based modulators, critical for the signal integrity required at 200G per lane speeds.

🔮 Future ImplicationsAI analysis grounded in cited sources

Shenzhen-based optical module manufacturers will capture over 40% of the global 1.6T market share by 2027.
Aggressive local government subsidies combined with rapid scaling of TFLN and CPO production lines provide a significant cost and volume advantage over international competitors.
The transition to 1.6T modules will force a redesign of standard AI server rack power delivery systems.
The increased power density required for 1.6T and CPO-enabled switches exceeds the thermal and power capacity of current standard rack architectures.

Timeline

2024-05
Shenzhen releases initial guidelines for AI infrastructure upgrades.
2025-02
Local pilot programs for 800G optical module mass production achieve stable yields.
2025-11
First successful prototype of a 1.6T module using TFLN technology is demonstrated in Shenzhen.
📰

Weekly AI Recap

Read this week's curated digest of top AI events →

👉Related Updates

AI-curated news aggregator. All content rights belong to original publishers.
Original source: 36氪