Startups & Funding

Edge AI Chip Startup Chenghengwei Raises Nearly RMB 200M Series A

Chenghengwei's nearly RMB 200M Series A, backed by CICC and CRRC Capital, funds mass production of its 64-TOPS CH37 edge AI SoC already in small-batch delivery for robots and drones.

Chenghengwei Raises Nearly RMB 200M Series A to Scale Edge AI SoC Output
Chenghengwei Raises Nearly RMB 200M Series A to Scale Edge AI SoC OutputRemazteredStudio / pixabay
By Elena Vasquez4 min read

Why it matters

  • Chenghengwei raised nearly RMB 200 million in a Series A announced November 19, 2026.
  • The CH37 SoC delivers peak AI compute of 64 TOPS (INT8) with a six-core NPU rated up to 48 TOPS.
  • First CH37 silicon powered on in December 2025; the chip formally launched in Wuxi on August 13, 2026.
  • Investors include CICC Capital, CRRC Capital, Addor Capital, Xichuang Xingwu and Qi'an Investment.
  • Chenghengwei was founded in June 2023 and is headquartered in Wuxi.

Chenghengwei has closed a Series A round of nearly RMB 200 million to push its CH37 edge AI chip from small-batch delivery into scaled mass production. The company announced the round on November 19, 2026. CICC Capital, CRRC Capital, Addor Capital, Xichuang Xingwu and Qi'an Investment all participated.

The money lands at a specific moment. The CH37 series, which packs an eight-core CPU, dual GPUs and a six-core NPU onto a single chip, came back from the fab and powered on for the first time in December 2025. It entered small-batch production in 2026. Now the Wuxi-based startup has to convert engineering momentum into commercial volume, and the round exists to fund exactly that conversion.

The stakes are straightforward. Edge AI chips compete on power efficiency and integration, and every robotics, drone and smart-terminal maker shopping for compute weighs those numbers against incumbents. A fabless startup founded in June 2023 has to prove it can deliver silicon at scale, not just demo it.

Who put in the money?

The cap table reads like a map of China's state-linked and regional capital. Chenghengwei describes CICC Capital as a large state-owned private-equity investor and CRRC Capital as a large state-owned industrial investor — CRRC being the rolling-stock giant whose venture arm signals industrial demand for embedded compute. Xichuang Xingwu represents local capital from Wuxi, Chenghengwei's home city. Qi'an Investment is private industrial capital, and Addor Capital is a market-oriented investor active in Jiangsu's semiconductor sector.

The company expects more than cash from this mix. In its announcement, Chenghengwei said the investors are expected to supply resources for production delivery, customer introductions and ecosystem development — connections meant to link the production ramp with real customer adoption and a partner base forming around CH37.

The proceeds are designated mainly for daily operations and business expansion. In practice, that means funding scaled mass production and market delivery of the company's edge AI system-on-chip products.

What's actually on the chip?

The CH37 series is a bid to collapse the traditional multi-chip stack into one piece of silicon. It combines:

  • An eight-core CPU
  • 2D and 3D GPUs
  • A six-core neural processing unit rated at up to 48 TOPS as a subsystem
  • A GPGPU for general-purpose computing, delivering up to 1 TFLOPS at FP32 with CUDA support
  • Image signal processors

Chenghengwei says this integration replaces conventional multi-chip designs, cutting power consumption and system complexity while increasing data throughput. That claim is the whole pitch for edge devices, where battery life and thermal budgets constrain how much compute a robot or drone can carry.

The chip's peak AI compute, per the CH37 product page, is 64 trillion operations per second using INT8 calculations. It also supports FP16, FP32 and FP64 mixed-precision computation. On the vision side, it carries two independent image-signal-processor channels — one for visible light, one for infrared — plus 4K video encoding and decoding at 120 frames per second.

The software stack matters as much as the silicon for adoption. Chenghengwei's AI tooling supports ONNX models, PyTorch, TensorFlow and ONNX Runtime, with an end-to-end NPU deployment toolchain. It handles CNN, DNN, RNN and Transformer model deployment — a notable inclusion given how quickly Transformer architectures have spread from language models into vision and robotics workloads.

From launch event to partner hardware

The chip got its formal launch in Wuxi on August 13, 2026. At that event, Chenghengwei signed strategic cooperation agreements with four companies: Inspur Electronic Information, Guangzhou Gaoneng Computer Technology, Zhejiang Liangchi Construction Technology and Shenzhen Xinchuang Technology. The agreements cover AI edge computing, smart sensing and embodied intelligence.

Guangzhou Gaoneng went further than a signature. It presented three CH37-based hardware products at the launch, including an industrial-control motherboard and an edge-computing box. Demonstrations at the event spanned security, road monitoring, industrial, commercial and home settings — a deliberate show of breadth across the markets Chenghengwei says it targets: robots, drones, edge-computing terminals and smart interactive terminals.

Those partner products show how CH37 is already being packaged into deployable systems. But hardware demos at a launch event are a different commodity from sustained orders.

Why does this round matter beyond Chenghengwei?

Chenghengwei is a young company with an aggressive timeline. Founded in June 2023 and headquartered in Wuxi, it runs branches, subsidiaries or R&D centers in Beijing, Shanghai, Changsha and Shenzhen. First silicon back in December 2025, formal launch in August 2026, small-batch delivery the same year — that pace reflects how quickly China's domestic edge AI chip sector is moving, and how much regional capital is willing to fund it.

The round also illustrates a funding pattern: state-owned investors like CICC Capital and CRRC Capital sitting alongside local Wuxi capital and private Jiangsu semiconductor money. For edge AI chipmakers, that blend of backers often comes with procurement relationships attached, which is precisely what Chenghengwei says it expects — customer introductions and ecosystem resources beyond the capital itself.

The funding announcement shifts the story's emphasis from technology to commercial execution. The next test is whether Chenghengwei can convert early deliveries and cooperation agreements into sustained deployments, with scaled production and market delivery backed by the new capital and its investors' expected customer and ecosystem reach.

Source: unite.ai

Share this article:

More from Elena Vasquez

Elena Vasquez

Show full bio

Market editor covering technology and product news in artificial intelligence and machine learning at AI In Context.

795 articles

Related articles

  1. Alibaba Trains Qwen on Its Own Zhenwu Chips as U.S. Export Controls Bite
  2. INFevo Raises Over RMB 100M to Build AI Virtual Cells and Robotic Labs
  3. Huawei and Alibaba Chip Blitz Arms Xi With Leverage Ahead of Trump Summit
  4. CXMT's Revenue Jumped Sevenfold. China's Memory Chipmakers Are Closing In
  5. China's AI Chip Market Splits From the West: RMB 800 Billion at Stake

« Previous articleNext article »