On April 26, domestic GPU leader Moore Threads Technology Co.,Ltd. disclosed its 2025 annual report and first-quarter 2026 results. According to the announcement, the company achieved annual revenue of 1.505 billion yuan in 2025, representing a 243.37% year-over-year increase. Gross profit reached 987 million yuan, up 218.43% from the previous year. In Q1 2026, Moore Threads reported revenue of 738 million yuan, growing 155.35% compared to the same period last year.
The company demonstrated robust performance growth and continuous optimization of operational efficiency. As commercialization accelerated, 2025 business maintained rapid expansion, setting new historical highs. From 2022 to 2024, Moore Threads achieved a compound annual growth rate exceeding 200% in operating revenue. Full-year 2025 revenue reached 1.505 billion yuan, surging 243.37% year-over-year. Leveraging comprehensive technological advantages and strong customer recognition, gross profit scaled to 987 million yuan, with overall gross margin reaching 65.57%, positioning the company at an industry-leading level. Net loss attributable to shareholders and adjusted net loss narrowed by 38.16% and 33.38% respectively year-over-year. Excluding share-based payment impacts, the 2025 net loss was 648 million yuan, improving by 847 million yuan from the prior year—a 56.65% reduction—reflecting sustained operational efficiency gains.
First-quarter performance was equally impressive. Moore Threads posted Q1 revenue of 738 million yuan, up 155.35% year-over-year. Net profit attributable to shareholders reached 29 million yuan, an increase of 142 million yuan from the previous year. Adjusted net loss narrowed to 54 million yuan, shrinking 60.10% compared to the same quarter last year.
In March 2026, Moore Threads secured a major contract valued at 660 million yuan for its Kuae intelligent computing cluster, underscoring its delivery capabilities and market competitiveness in large-scale intelligent computing solutions.
The GPU industry is technology-intensive, and sustained R&D investment is crucial for maintaining competitiveness. Since its inception, Moore Threads has focused on innovation, maintaining high R&D intensity. In 2025, R&D expenses totaled 1.305 billion yuan, accounting for 86.68% of operating revenue, placing the company among industry leaders in R&D commitment.
Long-term, substantial R&D investment has fortified Moore Threads' innovation capacity, building a strong technological moat. By December 31, 2025, the company had filed 2,014 intellectual property applications, including 1,743 invention patents, and secured 806 IP rights, with 590 invention patents granted, maintaining a leading industry position. This has enabled full-stack independent control from chip architecture and instruction sets to software stacks, with five chip generations developed over five years, consistently setting industry benchmarks.
Product-wise, Moore Threads has established a comprehensive "cloud-edge-device" matrix covering diverse sectors such as internet, embodied AI, smart education, intelligent transportation, industrial manufacturing, and cloud services. The flagship MTT S5000 AI training and inference card delivers up to 1000 TFLOPS dense computing power, supporting full-precision, general-purpose computing. It demonstrates performance and stability comparable to international flagship products, whether deployed in 10,000-card training clusters or high-concurrency, low-latency online inference services.
The Kuae 10,000-card training cluster, built on MTT S5000, overcomes significant engineering challenges in hardware optimization, high-speed interconnect, and system-level fault tolerance. It enables end-to-end training of trillion-parameter models and has achieved commercial deployment, with key metrics meeting international standards. To address future large-scale computing needs, Moore Threads is advancing 100,000-card ultra-large cluster development and core technologies like super-nodes based on its next-generation Huagang architecture.
Addressing AI development demands in the agent era, Moore Threads launched the MTT AIBOOK, powered by its self-developed Yangtze SoC. Pre-installed with native Linux, OpenClaw, and 12 curated Skills, it supports mainstream AI agent frameworks, eliminates complex setup, reduces performance overhead, and ensures 24/7 stable multi-agent operation, combining security with efficiency for out-of-the-box AI experiences.
Ecosystem development is vital for domestic GPU sustainability. Moore Threads has prioritized this as a core strategy, enhancing ecosystem layout through foundational technology support and developer empowerment.
Technically, the MUSA architecture and software stack offer full compatibility with CUDA ecosystems. MUSA natively supports mainstream AI frameworks like PyTorch, Megatron-LM, vLLM, and SGLang. The company has open-sourced software libraries ported to domestic GPUs and TileLang-MUSA to fully support the TileLang programming language, with deep Triton-MUSA backend compatibility. Since 2025, Moore Threads has achieved Day-0 adaptation for leading models including DeepSeek, GLM, MiniMax, Kimi, and Qwen, establishing routine support mechanisms.
In developer empowerment, Moore Threads' Moore Academy has built an industry-education integration system, attracting over 450,000 developers and learners. The program introduces cutting-edge technology and practical experience to more than 200 universities, laying a talent foundation for long-term ecosystem growth.
As AI adoption expands, demand for high-performance computing across industries surges. Aligning with trends toward computational autonomy, Moore Threads continues to iterate full-function GPU architectures and enhance product and ecosystem strategies, empowering various sectors with robust intelligent computing capabilities.
Comments