Source36氪•Stalecollected in 2h
VeriSilicon Secures 6.4B RMB in AI-Focused Orders
Massive AI chip order growth indicates a surge in demand for custom hardware for AI compute.
30-Second TL;DR
What Changed
New orders of 6.413 billion RMB in less than 3 months.
Why It Matters
The massive influx of AI-related orders confirms the strong demand for custom silicon solutions in the AI hardware ecosystem.
What To Do Next
Analyze VeriSilicon's custom silicon capabilities for your next hardware-accelerated AI project.
Who should care:Developers & AI Engineers
Key Points
- •New orders of 6.413 billion RMB in less than 3 months.
- •Over 90% of orders are for AI compute and data processing.
- •Total new orders for 2026 reached 146.53 billion RMB.
Deep Insight
AI-generated analysis for this event — not the original article.
Enhanced Key Takeaways
- •VeriSilicon's recent order surge is largely driven by the rapid adoption of its 'Chiplet' interconnect technology and NPU (Neural Processing Unit) IP cores in domestic Chinese data center projects.
- •The company has shifted its business model significantly toward 'Silicon Platform as a Service' (SiPaaS), allowing clients to integrate AI-specific compute modules faster than traditional ASIC design cycles.
- •A substantial portion of the 6.413 billion RMB order volume is attributed to the automotive intelligence sector, specifically for autonomous driving SoC (System-on-Chip) development.
- •VeriSilicon has expanded its strategic partnerships with major Chinese foundries to ensure supply chain resilience for these AI-focused orders amid ongoing global semiconductor export restrictions.
- •The company's financial performance in 2026 reflects a pivot toward high-margin AI IP licensing, which has improved its gross profit margins compared to its legacy video processing business.
Competitor Analysis
Cadence Design Systems
- Primary Focus
- IP & EDA Tools
- Competitive Advantage
- Comprehensive ecosystem & global standard
- AI Benchmarking
- High (Industry benchmark)
Synopsys
- Primary Focus
- IP & EDA Tools
- Competitive Advantage
- Market dominance in interface IP
- AI Benchmarking
- High (Industry benchmark)
Imagination Technologies
- Primary Focus
- GPU/NPU IP
- Competitive Advantage
- Strong mobile/embedded AI performance
- AI Benchmarking
- Moderate (Niche focus)
ARM
- Primary Focus
- CPU/NPU IP
- Competitive Advantage
- Ubiquitous architecture & software support
- AI Benchmarking
- High (General purpose)
| Competitor | Primary Focus | Competitive Advantage | AI Benchmarking |
|---|---|---|---|
| Cadence Design Systems | IP & EDA Tools | Comprehensive ecosystem & global standard | High (Industry benchmark) |
| Synopsys | IP & EDA Tools | Market dominance in interface IP | High (Industry benchmark) |
| Imagination Technologies | GPU/NPU IP | Strong mobile/embedded AI performance | Moderate (Niche focus) |
| ARM | CPU/NPU IP | Ubiquitous architecture & software support | High (General purpose) |
Technical Deep Dive
- VeriSilicon's NPU IP (VIP series) utilizes a scalable architecture that supports INT8, INT16, and FP16 precision for diverse AI workloads.
- The platform integrates proprietary 'FlexAIBus' technology, which optimizes data movement between heterogeneous compute units to reduce latency in large-scale AI models.
- Implementation includes support for multi-die packaging (Chiplets), enabling the integration of high-bandwidth memory (HBM) with AI compute dies for data-intensive processing.
- The software stack, AIFlow, provides a unified compiler that maps neural network models (TensorFlow, PyTorch, ONNX) directly to the underlying hardware IP, minimizing manual optimization efforts.
Future ImplicationsAI analysis grounded in cited sources
VeriSilicon will achieve a 20% increase in annual revenue growth by end-of-year 2026.
The high concentration of AI-focused orders suggests a sustained demand for high-margin IP licensing that will materialize in revenue as projects reach tape-out.
The company will increase R&D spending on advanced packaging technologies by at least 15% in the next two quarters.
To maintain its competitive edge in AI compute, VeriSilicon must accelerate its Chiplet and 2.5D/3D packaging capabilities to meet the requirements of high-performance data center chips.
Timeline
2020-08
VeriSilicon completes its IPO on the Shanghai Stock Exchange STAR Market.
2023-05
Company announces the expansion of its NPU IP portfolio to support transformer-based AI models.
2025-02
VeriSilicon achieves a milestone of over 100 million units shipped containing its NPU IP cores.
2026-01
Strategic pivot announced to prioritize AI computing power and data processing as the primary growth engine.
- 2020-08VeriSilicon completes its IPO on the Shanghai Stock Exchange STAR Market.
- 2023-05Company announces the expansion of its NPU IP portfolio to support transformer-based AI models.
- 2025-02VeriSilicon achieves a milestone of over 100 million units shipped containing its NPU IP cores.
- 2026-01Strategic pivot announced to prioritize AI computing power and data processing as the primary growth engine.
Weekly AI Recap
Read this week's curated digest of top AI events →
AI-curated news aggregator. All content rights belong to original publishers.
Original source: 36氪 ↗
This is a summary, not the original. Read the source, or get the weekly briefing.
The weekly digest
One email a week. Unsubscribe anytime.