MetaX's revenue in the first three quarters was 2.824 billion yuan, more than doubling year-on-year, and it turned a profit

On the evening of October 30, MetaX released its 2026 third-quarter report. Data shows that the company's revenue for the first three quarters was 2.824 billion yuan, a year-on-year increase of 128.45%; net profit attributable to shareholders was 639 million yuan, and non-GAAP net profit attributable to shareholders was 259 million yuan, turning profitable year-on-year. Looking at the third quarter alone, revenue reached 1.5 billion yuan, exceeding the total for the first half of the year, and non-GAAP net profit attributable to shareholders was 308 million yuan, marking two consecutive quarters of non-GAAP profitability, a clear signal of commercialization and profitability improvement. The strong performance is underpinned by the continued conversion of GPU product orders. In May this year, the Xiyun C600 went into mass production and launched, completing the high-performance GPU product matrix and, relying on a fully domestic supply chain for stable delivery, supporting market expansion. At the industry level, the rapid adoption of AI applications such as agents is driving up demand for computing power. Data from the China Academy of Information and Communications Technology shows that as of June 2026, domestic daily average Token calls exceeded 500 trillion, and the expansion of computing power demand is opening up space for domestic GPUs. Currently, the investment logic for AI computing power is shifting, with market focus moving from the scale of hardware investment to return on investment, and GPU competition shifting from single-card performance comparisons to system-level computing efficiency. MetaX is simultaneously pursuing coordinated optimization of hardware, high-speed interconnect, and software stack, and at this year's WAIC launched the Xijing S series super nodes and the Xisuo X300 series for scientific intelligence. The Xijing S600 can scale horizontally to a 10,000-card cluster, reducing multi-card communication latency; the accompanying MXMACA software stack has been iterated to version 3.9.0, and the relevant KV Cache technology code has been merged into the open-source project LMCache. The company has also completed rapid adaptation of mainstream large models such as Zhipu GLM, Qwen, and Hunyuan. In terms of deployment cases, the 10,000-card-class "Token Super Factory" jointly developed by MetaX and Moxing Intelligence has already gone into production; relying on software-hardware synergy, the cluster requires only 1 to 2 days from readiness to model launch, greatly shortening the deployment cycle for large-scale computing power. In terms of talent incentives, the company has simultaneously introduced a draft restricted stock incentive plan, setting revenue scale and the number of new invention patent applications as assessment indicators, both locking in operating targets and encouraging continuous R&D innovation. The assessment sets: revenue in 2027 of no less than 6 billion yuan, cumulative revenue from 2027 to 2028 of no less than 15 billion yuan, and cumulative revenue from 2027 to 2029 of no less than 30 billion yuan, binding the core team to long-term development. This third-quarter report confirms that domestic GPU manufacturers have achieved phased results in commercialization. As computing power construction places greater emphasis on cluster efficiency and deployment costs, MetaX's full-stack solution will also continue to be tested in large-scale production scenarios.