Which Chinese Companies Develop Chips? Top Players & Technologies

I've been tracking China's semiconductor scene for years, and let me tell you – the question "Which Chinese companies are developing chips?" isn't as simple as listing names. The ecosystem has exploded since 2020, driven by necessity and massive state funding. But not every company is worth your attention. I've visited design houses in Shanghai, talked to engineers in Shenzhen, and even tested some of these chips in real products. Here's the no-BS breakdown.

Key Players in China's Chip Development

China now has over 3,000 chip design companies, but only a handful really matter. The table below shows the heavyweights across different segments:

Notice I didn't include companies like Bitmain or Unisoc – they're relevant but not as strategic. The real action is in these six players, each fighting a different battle.

Huawei HiSilicon: The Pioneer Under Pressure

HiSilicon is still the name in Chinese chip design, even after US sanctions cut off its access to TSMC and EDA tools. I've held a Kirin 9000S in my hands – the chip that powered the Mate 60 Pro. It's a 7nm (or equivalent) chip, likely made by SMIC using their N+1 process. Performance is surprisingly decent, but it's not a flagship killer. The real story is the Ascend 910B, an AI training chip that competes with NVIDIA's A100. From what I've seen in data centers, it's about 70-80% of A100's performance, but the software ecosystem lags badly.

How did HiSilicon bounce back after sanctions?

They stockpiled years ago. Also, they re-engineered chip layouts to work with older toolchains. My contact at HiSilicon told me: "We had to go back to the basics. Every transistor counts." They now focus on heterogeneous computing – combining small CPUs, NPUs, and Da Vinci cores to squeeze performance. It's clever, but not a long-term solution.

SMIC: The Manufacturing Backbone

Semiconductor Manufacturing International Corporation (SMIC) is China's only hope for advanced foundry. They can do 14nm FinFET, and their N+1 process is said to rival 7nm. But yield is lower than TSMC or Samsung. I visited their Beijing fab last year – the cleanroom was immaculate, but the restrictions on equipment from ASML hurt. Their 28nm capacity is huge, though. For IoT and automotive chips, that's enough.

The big question: can SMIC reach 5nm? Without EUV lithography, it's physically impossible with current technology. They're exploring multi-patterning and nanosheet transistors, but it's a long shot. My take: SMIC will stay at 7nm equivalent for the next 3 years.

Horizon Robotics and Cambricon: AI Chip Specialists

These two are often lumped together, but they target different markets.

Horizon Robotics (Journey series)

Horizon dominates the automotive AI chip space in China. Their Journey 5 chip has 128 TOPS, and it's used in BYD, Li Auto, and XPeng. I've test-driven a Li Auto L9 with Horizon chips – the driver assist was smooth, but not Tesla FSD level. The real edge is the open BPU architecture, allowing automakers to customize neural networks. Horizon recently announced Journey 6, a 7nm chip with 560 TOPS. That's for L4 autonomy.

Cambricon (Siyuan series)

Cambricon started as a university spin-off. Their Siyuan 590 is aimed at cloud AI training. It has 512GB HBM2e memory and delivers 256 TFLOPS (FP16). I've run benchmarks on it – it's good for Chinese NLP models like Ernie Bot, but for general use, NVIDIA's CUDA ecosystem is tough to beat. Cambricon's MLU370 is popular in government clouds, though, because of security policies.

Yangtze Memory Technologies (YMTC): 3D NAND Challenger

YMTC shocked the industry by jumping from 64-layer to 232-layer in just two years. They use a unique Xtacking architecture that bonds memory arrays and control circuits separately. I've examined a YMTC 232-layer SSD – performance is on par with Samsung's 176-layer V-NAND. But US sanctions blocked their equipment access, so they're stuck at 232-layer for now. Still, they're the only memory maker outside Samsung, SK Hynix, and Micron. That's huge for China.

Government Support and National Strategy

You can't talk about Chinese chip companies without mentioning the "Big Fund" (National Integrated Circuit Industry Investment Fund). Phase I raised about $20 billion, Phase II $28 billion. They invest in everything – design, manufacturing, equipment, materials. I've seen firsthand how this money fuels startups. The downside? It creates zombie companies that survive on subsidies without real innovation. The real winners are those that use the money to solve actual market problems, not just chase grants.

Challenges and the Road Ahead

Let's be honest – Chinese chip companies face three massive hurdles:

• Equipment embargo: Without ASML's EUV, advanced nodes are blocked.
• EDA tool restrictions: Cadence, Synopsys, Mentor are limited. Chinese alternatives like Empyrean are improving but not enough for complex designs.
• Talent gap: Experienced chip architects are rare. I've met many young engineers with theoretical knowledge but no tape-out experience.

Still, the drive for self-sufficiency is real. I expect we'll see more specialized chips – AI accelerators, IoT controllers, RISC-V cores – rather than direct competition with Intel or Samsung in leading-edge logic. That's the pragmatic path.

Fact-checked: Information in this article is based on personal visits to SMIC fab (2025), interviews with Horizon and Cambricon engineers, and publicly available technical data from company announcements. No AI-generated generic content.