🔥Stalecollected in 14m

ChiNext Surges 2% on Compute Hardware Lead

ChiNext Surges 2% on Compute Hardware Lead
PostLinkedIn
🔥Read original on 36氪

💡Compute hardware rally hints at AI infra boom—watch for supply plays

⚡ 30-Second TL;DR

What Changed

ChiNext up 2%

Why It Matters

Signals strength in AI compute hardware supply chain amid market rally.

What To Do Next

Track compute hardware stocks via 36Kr for AI infra investment signals.

Who should care:Founders & Product Leaders

🧠 Deep Insight

Web-grounded analysis with 10 cited sources.

🔑 Enhanced Key Takeaways

  • Co-packaged optics (CPO) technology is experiencing accelerated adoption in 2026 data center infrastructure, with Lumentum forecasting approximately $50 million in CPO revenue by Q4 2026 and explosive growth expected in H1 2027, driven by AI workload scaling requirements[1][3]
  • CPO and LPO technologies are critical for addressing power consumption challenges in hyperscale data centers, reducing power consumption by 30-50% while enabling higher bandwidth and better total cost of ownership as AI data center workloads are forecast to triple or quadruple annually between 2026-2030[3]
  • Chinese semiconductor and computing hardware stocks, particularly CPO manufacturers like TFC Communication and Juguang Technology, have surged to all-time highs in early 2026, with multiple stocks gaining over 10% as CITIC Securities research indicates accelerating CPO adoption[1][2]
  • The broader Chinese stock market has demonstrated strong performance year-to-date in 2026, with the MSCI China Index posting over 30% gains supported by policy pivots and emergence of new growth drivers in AI infrastructure sectors[10]
  • Memory supply constraints continue to drive semiconductor valuations in 2026, with advanced HBM production capabilities concentrated among Samsung Electronics, SK hynix, and Micron, while Chinese suppliers like CXMT and YMTC remain limited to lower-end segments[8]

🛠️ Technical Deep Dive

  • CPO integration in data center switches enables higher aggregate bandwidth per rack with lower Ethernet/InfiniBand switch footprint
  • Chiplet-based architectures with HBM integration on silicon interposers or 3D stacks allow data movement between GPUs/NPUs and memory at multiple terabytes per second with improved energy efficiency
  • CPO and LPO technologies shorten electrical paths in data center networks scaling to 51.2 terabits per second and above switching capacities, addressing limitations of traditional copper pluggables
  • Advanced HBM production remains concentrated at three manufacturers (Samsung, SK hynix, Micron) with Chinese suppliers accounting for only 5-10% of market share in lower-end segments

🔮 Future ImplicationsAI analysis grounded in cited sources

CPO technology will become standard infrastructure in hyperscale AI data centers by 2027
Lumentum's $50M Q4 2026 CPO revenue forecast with expected explosive H1 2027 growth, combined with Deloitte's prediction of increased HBM-logic integration, indicates rapid mainstream adoption.
Memory pricing will remain elevated through 2027-2028 despite gradual supply easing
Sustained AI-driven demand and rising capital intensity in advanced HBM production will support pricing even as supply constraints gradually resolve.
Chinese semiconductor suppliers will face continued competitive disadvantages in advanced AI infrastructure
Chinese firms lack advanced HBM capabilities required for leading AI systems, limiting their participation in the highest-margin segments of the CPO and memory supply chain.

Timeline

2023-04
Offshore renminbi last appreciated beyond 6.90 mark against USD before March 2026 surge
2025-01
Micron stock peaked at $455, representing 340% gain from start of 2025
2026-02
ChiNext Index surged over 1% with CPO sector strength; Zhipu announced GLM-5 AI model launch
2026-03-09
ChiNext Index plummeted over 3% with computing power hardware and semiconductor chips leading declines
📰

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