Moore Threads, the Beijing-based GPU maker listed on Shanghai's STAR Market, swung to a net profit in the third quarter as revenue rose 1,773.92% year over year to 1.55 billion yuan, according to its unaudited Q3 2026 report published Wednesday, October 28.
Net profit attributable to shareholders came to 49.26 million yuan for the quarter, with total profit of 62.25 million yuan and earnings of 0.10 yuan per share. Stripping out non-recurring items, Q3 net profit was 20.07 million yuan. On that same measure, the company had posted a loss of about 150.8 million yuan in the first half, according to an analysis by China AI Dispatch.
For the first nine months of 2026, Moore Threads reported revenue of 3.29 billion yuan, up 319.15% from 784.6 million yuan a year earlier. Net profit attributable to shareholders was 37.7 million yuan, compared with a loss of 723.5 million yuan in the same period of 2025.
Key facts
- 01Moore Threads' Q3 2026 revenue rose 1,773.92% year over year to 1.55 billion yuan.
- 02Q3 net profit attributable to shareholders was 49.26 million yuan, a swing from a loss.
- 03Nine-month revenue reached 3.29 billion yuan, up 319.15%, while R&D spending hit 1.34 billion yuan.
- 04Operating cash outflow was 3.31 billion yuan over nine months as inventory rose to 3.36 billion yuan.
- 05Long-term borrowings climbed to 3.65 billion yuan from 1.49 billion at the end of 2025.
Non-recurring items contributed 168.4 million yuan to the nine-month result, including 91.0 million yuan in government subsidies. Excluding those items, Moore Threads recorded a nine-month loss of 130.7 million yuan.
Research spending kept climbing. Q3 R&D investment was 571.1 million yuan, up 87.89%, and nine-month R&D reached 1.34 billion yuan, up 55.72% and equal to 40.75% of revenue. China AI Dispatch calculated a nine-month gross margin of 53.71%, down from 61.87% in the first nine months of 2025.
The company sells MTT S5000 intelligent computing products and is developing HuaGang, its next-generation full-function GPU architecture. It organizes its business around three AI factories for model training, token production and agent production, and it recently open-sourced Megatron-MUSA-Titan, a training stack for trillion-parameter mixture-of-experts models on 10,000-card clusters.
On September 9, JD Cloud announced plans for a 100,000-GPU full-function cluster built on Moore Threads GPUs for large-model training, inference and embodied intelligence. Cao Peng, president of JD Cloud, called the partnership "an important step in JD Cloud's AI strategy."
Moore Threads used 3.31 billion yuan in operating cash over the nine months, against 3.29 billion yuan of revenue and up from a 1.50 billion yuan outflow a year earlier. The company attributed the gap to expanded production and strategic inventory stocking. Inventory rose to 3.36 billion yuan from 1.33 billion at the end of 2025, and supplier prepayments rose to 2.60 billion yuan from 1.78 billion.
Cash and cash equivalents fell to 5.39 billion yuan from 9.18 billion, with 1.87 billion yuan of that moved into short-term financial assets. Long-term borrowings rose to 3.65 billion yuan from 1.49 billion, while total assets grew 15.68% to 17.74 billion yuan.
China AI Dispatch noted that Moore Threads, Cambricon and Unitree have all reported revenue and inventory rising together, and argued that the inventory build shifts risk onto the company while being partly financed with borrowed money.
Moore Threads' board approved a plan on August 7 to list H shares in Hong Kong. Its stock closed at 381.50 yuan on Wednesday, about 59% below its peak close of 941.08 yuan on December 11, 2025. Founder, chairman and CEO Zhang Jianzhong holds 9.41% of the company directly.
Working on something we should cover, or seeing a story we missed? Send leads, documents, or feedback to hello@aichatdaily.com. For sensitive tips, see our secure tips page for Signal and PGP options.
Spotted an error? Email hello@aichatdaily.com with the URL and the issue, or read our full corrections policy.
