🧔♂️ A friendly human may check it before it goes live. More news here
China’s Moore Threads expects threefold revenue growth
Chinese AI chipmaker Moore Threads expects to report a threefold increase in revenue for 2025, according to a stock exchange filing.
The company projects revenue of about 1.5 billion yuan (US$215.4 million), up more than 230% from 2024.
It also expects its net loss to narrow to between 950 million (US$136.4 million) and 1.1 billion yuan (US$158 million), compared with a 1.6 billion yuan (US$229.8 million) loss last year.
The growth is driven mainly by adoption of its MTT S5000 chip, based on the fourth-generation Pinghu GPU architecture, which entered mass production in 2025.
The chip supports AI model training with trillion-level parameters and offers efficiency comparable to foreign GPU clusters.
Moore Threads listed on Shanghai’s Star Market last month, as Chinese chipmakers push to build domestic AI technology and cut reliance on foreign suppliers.
🔗 Source: South China Morning Post
🧠 Food for thought
Implications, context, and why it matters.
Specific evidence is needed to assess Moore Threads’ competitive claims
- The stock exchange filing claim that MTT S5000 clusters deliver computational efficiency “comparable to GPU clusters of the same size by foreign companies” should be backed with clear metrics, and it should be framed as Moore Threads’ claim rather than an independently verified result.
- Independent, third-party benchmarks should be reviewed for performance on industry-standard AI workloads so an objective comparison can be made.
- The customers behind the company’s described “wide adoption” of the MTT S5000 chip should be named so market traction can be checked beyond reported revenue growth.
The rise of domestic GPUs could create a software and services opportunity
- Software firms and system integrators can build a business around support services as domestic chips like the MTT S5000 gain ground.
- Teams can create migration tools and offer consulting to move AI workloads from Nvidia’s CUDA (a widely used software platform for programming Nvidia GPUs) to domestic stacks such as Moore Threads’ MUSA (the company’s software platform for its GPUs), including code-porting toolchains described in publicly available materials.
- Cloud providers can sell compute instances built on domestic GPUs to reach Chinese enterprises and state-backed entities that prefer technological self-sufficiency.
