MetaX (688802.SH) Sees Revenue Double in First Three Quarters with Non-GAAP Turnaround to Profit as System-Level Computing Layout Accelerates

On the evening of October 30, MetaX (688802.SH) released its 2026 third-quarter report. In the first three quarters, the company achieved operating revenue of RMB2.824 billion, up 128.45% year on year; net profit attributable to shareholders was RMB639 million, and non-GAAP net profit attributable to shareholders was RMB259 million, both turning from loss to profit compared with the same period last year. By quarter, the company achieved revenue of RMB1.5 billion in the third quarter, with single-quarter revenue exceeding the total for the first half; non-GAAP net profit attributable to shareholders was RMB308 million, marking two consecutive quarters of non-GAAP profitability. The trends of revenue growth and improvement in main-business profitability have become further evident. Revenue growth accelerates, main-business profitability continues to improve From quarterly performance, the company's single-quarter operating revenue in the third quarter has exceeded the total for the first half, showing an acceleration in the company's commercialization process. During the reporting period, the company's GPU products and services continued to gain recognition and procurement from downstream customers, and product sales scale continued to expand. At the same time, with the Xiyun C600 officially entering mass production and launch in May, the company's high-performance GPU product matrix has been further improved, and product delivery capability based on a fully domestic supply chain has continued to strengthen, providing stronger support for subsequent market expansion. On the profit side, after achieving non-GAAP net profit attributable to shareholders of approximately RMB54 million in the second quarter, the company further increased it to RMB308 million in the third quarter, achieving single-quarter non-GAAP profitability for two consecutive quarters and driving cumulative non-GAAP net profit attributable to shareholders in the first three quarters to turn from loss to profit year on year. For a GPU company that continues to maintain high-intensity R&D investment, this change reflects that while expanding business scale, the company's main-business profitability has also achieved substantive improvement. While the company's performance is accelerating, demand for AI computing power also remains robust. With the continuous improvement of large-model capabilities and the accelerated adoption of applications such as agents, demand for AI computing power continues to grow. According to data from the China Academy of Information and Communications Technology, as of June 2026, China's average daily Token calls had exceeded 500 trillion, a significant increase from 100 trillion at the end of 2025. The rapid growth of AI application scale is continuously expanding demand for computing power, opening broader market space for domestic high-performance GPUs. Super nodes and software ecosystem work in concert to continuously improve system-level computing efficiency Global demand for AI applications continues to be released, driving investment in computing infrastructure to remain at a high level. According to statistics, in the second quarter of 2026, the four major U.S. technology companies (Microsoft, Alphabet, Meta, and Amazon) had combined cash capital expenditures of approximately US$165 billion, up 87% year on year. As investment continues to expand, whether computing infrastructure can generate corresponding revenue and whether cash flow can match investment intensity are becoming core issues for the market in evaluating this round of the AI investment cycle. This shift from "investment scale" to "ROI (return on investment)" is also driving competition in the GPU industry to extend from single-card performance to system-level capabilities. In line with this trend, during the reporting period MetaX continued to advance the coordinated development of chips, interconnect, and software stacks, using system-level optimization to improve computing utilization efficiency and economics. During WAIC in July, the company released the Xijing S-series super node products and the Xisuo X300 series products for AI for Science. Among them, Xijing S600 supports high-speed in-rack interconnect and strategies such as expert parallelism and tensor parallelism, and can be further scaled horizontally to 10,000-card-level clusters. By reducing multi-card communication latency and improving parallel computing efficiency, it unleashes the effective computing power of clusters. Combined with the MXMACA software stack, it can further improve system efficiency in large-scale training and inference scenarios, providing support for customers to optimize the full life-cycle cost of computing infrastructure. Focusing on the actual efficiency of computing systems, MetaX continued to advance upgrades to the MXMACA software stack. During the reporting period, the company released MXMACA version 3.9.0, improving software capabilities for areas such as super node data access, multi-card communication, and operation monitoring, providing support for large-scale training and inference. At the same time, the MXMACA build support code for MetaX GPUs has been merged into the international open-source project LMCache, providing a foundation for the application and optimization of KV Cache management technology on domestic computing platforms. During the reporting period, the company also completed Day 0 adaptation for multiple cutting-edge models, including Z.AIGLM, Alibaba Qianwen, and Tencent Hunyuan, improving software support from underlying computing power to upper-layer applications and responding to customers' comprehensive requirements for computing efficiency, deployment convenience, and usage cost. On the large-scale application side, during the reporting period, MetaX's 10,000-card-level "Token Super Factory" in cooperation with Moxing Intelligence has been put into the production environment. With the coordinated support of MetaX's domestic GPU hardware and MXMACA software stack, the first batch of large-scale computing clusters completed the entire process from cluster readiness to business model launch in only 1 to 2 days, demonstrating the important value of MetaX's software-hardware coordination capabilities in large-scale computing deployment, model adaptation, and large-scale production-grade applications. Equity incentives balance revenue and innovation, improving long-term incentive mechanisms While continuing to advance product and technology iteration, MetaX further improved its long-term incentive mechanism. The company recently disclosed the draft of its 2026 restricted stock incentive plan and included operating revenue and the number of new invention patent applications in the vesting assessment. Among them, the revenue assessment targets for the first grant portion are no less than RMB6 billion for 2027, no less than RMB15 billion cumulatively for 2027 to 2028, and no less than RMB30 billion cumulatively for 2027 to 2029. By incorporating both business growth and technological innovation into the assessment, the plan helps bind core talent with the company's long-term interests and also demonstrates the company's confidence in future development. The strong performance in the first three quarters reflects the positive progress MetaX has made in product commercialization and large-scale application. The release of Xijing super nodes and the continuous upgrading of the MXMACA software stack have further improved the company's full-stack technology system integrating chips, interconnect, and software, enhancing system-level computing capability. As AI computing infrastructure development moves toward scale and high efficiency, the company will continue to advance technological innovation and commercialization, enhancing the market competitiveness of domestic high-performance GPUs.