Chinese internet companies are increasingly building their own computer chips to power artificial intelligence workloads, betting that proprietary semiconductor designs will give them greater control over computing infrastructure as geopolitical tensions limit access to advanced foreign technology. Kuaishou, the short video platform, has become the latest to pursue this strategy after its chip spinoff TranStreams raised an undisclosed amount in Series A+ funding this week, according to reports from ITJuzi.com, a local startup database platform.

Why Are Chinese Tech Giants Building Their Own Chips?

The funding round for TranStreams reflects a broader shift among China's largest internet companies. The round was led by QF Capital and included participation from a state-backed investment vehicle under the Beijing Science and Technology Innovation Fund, Baidu's venture arm, and XGD, a Shenzhen-based digital payment technology provider. This investor lineup signals strong confidence in the semiconductor strategy across the industry.

The motivation is straightforward: as artificial intelligence demands surge, companies want to lower long-term computing costs and reduce their dependence on third-party suppliers for critical hardware. ByteDance, Baidu, and Alibaba Group Holding have all moved into semiconductor design for similar reasons. For Kuaishou specifically, the focus is on video processing and AI inferencing, the computational work required to run AI models on existing data.

TranStreams traces its roots to a heterogeneous computing and chip unit that Kuaishou established in 2018. The company was spun off as an independent entity in March 2024 with the explicit goal of developing the SL200 system-on-chip, a specialized processor designed for video processing and AI inferencing tasks. This separation allows the chip team to operate with greater autonomy and attract dedicated investment.

How Does Kling AI Fit Into Kuaishou's Chip Strategy?

Kuaishou's Kling video generation model represents one of the most demanding use cases for custom silicon. Kling is an image and text-to-video model known for expressive character motion and realistic human movement, with strong physics simulation. The model comes in multiple variants optimized for different needs: Kling 2.1 Pro for higher quality, Kling 2.5 Turbo for speed, and newer generations like Kling 3 and Kling O3 with increasingly sophisticated capabilities.

Running these video generation models at scale requires enormous computational power. By developing proprietary chips like the SL200, Kuaishou can optimize hardware specifically for the mathematical operations that Kling and similar models perform. This vertical integration, where a company controls both the software and the underlying hardware, allows for efficiency gains that generic processors cannot match. The result is faster video generation, lower power consumption, and reduced costs per generated video.

Steps to Understanding China's AI Chip Independence Strategy

• Geopolitical Context: US export controls restrict China's access to advanced semiconductors from companies like Nvidia, forcing Chinese tech firms to develop domestic alternatives to maintain AI competitiveness.
• Vertical Integration Benefits: Companies like Kuaishou, ByteDance, and Alibaba design chips tailored to their specific AI workloads, achieving better performance and lower costs than relying on general-purpose processors.
• Long-Term Cost Reduction: Proprietary chips reduce dependence on expensive third-party suppliers and allow companies to amortize chip development costs across millions of video generations and AI inferences.
• Investor Confidence: The participation of state-backed funds and venture arms from major tech companies signals that semiconductor development is now viewed as critical infrastructure for AI competitiveness in China.

The broader video AI landscape shows how specialized hardware can unlock new capabilities. Kuaishou's Kling competes with models from other Chinese companies like ByteDance's Seedance, which is known for long, coherent single-take shots and multi-shot consistency at up to 4K resolution. Google's Veo 3 and Veo 3.1 offer photoreal motion with dialogue and effects generated in a single pass. Alibaba's Wan and Happy Horse models provide alternatives optimized for speed and control. Each of these models demands different computational characteristics, making custom silicon increasingly valuable.

The timing of TranStreams' funding is significant. As AI video generation becomes more central to content creation, social media, and enterprise applications, the companies that can generate videos fastest and cheapest will gain competitive advantage. A proprietary chip designed specifically for video processing can deliver that edge. For Kuaishou, which operates one of the world's largest short video platforms, the ability to generate high-quality videos at scale using in-house silicon could reshape economics across the entire video AI industry.

Neither Kuaishou nor TranStreams responded to requests for comment on the funding announcement. However, the investment itself speaks to the strategic importance of semiconductor independence in China's AI future. As US export controls tighten and geopolitical tensions persist, expect more Chinese tech giants to follow Kuaishou's path of building proprietary chips tailored to their AI workloads.