On June 17, the authoritative "2026 Top 30 Domestic Computing Power Chips" list, jointly selected by authoritative institutions such as the Chinese Academy of Sciences and the Chinese Academy of Social Sciences, was officially released. Huawei, Cambricon, and Hygon Information ranked in the top three. The aforementioned institutions believe that computing power has become one of the most certain tracks in the technology industry. The demand for computing power for large model iteration continues to rise, and the industry consensus on self-reliance and controllability is continuously strengthening. Driven by these dual forces, domestic computing power chips have moved from behind the scenes to the forefront. Currently, the domestic cloud computing power chip echelons are differentiating, and the tracks are becoming more segmented: general-purpose GPU and AI-specific ASIC routes are developing in parallel, training and inference scenarios are competing in misaligned ways, and top-tier, core, and innovative echelons are breaking through in layers, with a differentiated competitive landscape beginning to emerge. At the same time, domestic chip companies with high R&D investment are accelerating their embrace of the capital market. Journalists have found that many companies on the list have already gone public. Among them, companies like Zhongcheng Hualong and Xiwang focus on the pure inference chip track, achieving differentiated parallel development, and are recently accelerating their capitalization layout, becoming the focus of market attention.

Journalists have found that several companies on the list, including Cambricon, Hygon Information, and Moore Thread, have successfully gone public. In this list, two "preparatory teams" accelerating their sprint are worth noting. Both Zhongcheng Hualong and Xiwang focus on the differentiated track of inference chips and are also accelerating their IPO processes recently.

According to Securities Times reports, Zhongcheng Hualong and CITIC Securities recently formally signed a comprehensive cooperation agreement. Representatives from China Investment Fund, Beijing Jintian Gongcheng Law Firm, and Lixin Accounting Firm jointly attended and witnessed the event, marking the official launch of Zhongcheng Hualong's listing and capitalization process. Zhongcheng Hualong is one of the few domestic technological innovation enterprises with full-stack, self-reliant, and controllable capabilities in "chip + whole machine + solution." It is also one of the few domestic chip companies that have laid out a "three-core self-developed + three-computing integrated" strategy. The company, based on "Shenwei technology + RISC-V technology + quantum computing technology + trusted technology," builds a differentiated and diversified self-reliant and controllable technology system, creating a one-stop domestic service capability, and is committed to promoting the self-reliance, controllability, and scaled implementation of core chips such as GPU, CPU, and QPU. Zhongcheng Hualong's chief scientist is Academician Shen Changxiang, the "father of trusted computing." The core technical team's backbone members come from NVIDIA, Huawei HiSilicon, Alibaba DAMO Academy, ByteDance, Arm China, Texas Instruments, SMIC, and national-level research institutes. Zhongcheng Hualong focuses on the construction of full-stack computing power infrastructure such as general computing, intelligent computing, supercomputing, in-memory computing, quantum computing, and space computing power. On the "2025 Top 100 Sci-Tech Unicorns" list, Zhongcheng Hualong ranked fifth.

In 2025, Zhongcheng Hualong successfully taped out and released a fully domestic AI high-computing-power chip based on domestic manufacturing processes. According to evaluations by authoritative institutions of the Ministry of Industry and Information Technology, the measured energy consumption of the HL100 is as low as 65.33W, with an ultra-high energy efficiency ratio breaking through 3.41 TFLOPS/W. Its computing power under the same power consumption is 8 times that of a certain foreign AI chip. Under the same computing power, the total cost of ownership (TCO) of the HL100 is only 1/4 of that of a certain foreign AI chip. The Zhongcheng Hualong HL series chips, with a higher "energy efficiency ratio" as the core, create a golden triangle of performance, power consumption, and cost, revolutionizing the Token economy cost through computing power efficiency. The next-generation Zhongcheng Hualong HL200 chip, by equipping FP4 precision computing units specifically strengthened for inference scenarios and matching them with an MP8 adaptive variable precision quantization engine, pushes computing density to the extreme. The Zhongcheng Hualong HL200, HL200Pro, and HL400 AI chips will natively support FP8/FP4, with performance benchmarked against international mainstream AI chip levels, fully meeting the inference needs of next-generation generative AI and AI Agent applications.

Another company on the list, Xiwang, was formerly the large chip department of SenseTime. At the end of 2024, SenseTime promoted a "1+X" organizational restructuring, and Xiwang, as an important part of the "X" camp, officially became independent. In 2025, former SenseTime executive director Xu Bing personally took the field as the chairman of Xiwang, leading this team with an average of 15 years of industry experience. The company's Co-CEO Wang Yong is the core architect of Baidu's Kunlun Chip and has over 20 years of chip R&D experience. After independence, Xiwang has demonstrated strong fundraising ability in the capital market, accumulating approximately 4 billion yuan in financing. Xiwang has completed the tape-out of two products, S1 and S2, and released Qiwang S3 in January this year, with the tape-out expected to be completed within 2026.

Doben Consulting believes that the future survival space for domestic chip companies lies in the deep refinement of vertical scenarios. By optimizing the adaptation, optimization, and experience of a single scenario to the extreme, forming a differentiated advantage that others cannot replace, is far more competitive than creating a "large and comprehensive" but mediocre product. The two "preparatory teams" of domestic computing power chips, Zhongcheng Hualong and Xiwang, are standing at a critical juncture. As parallel runners in pure inference chips, whether they can successfully break through in differentiated tracks and promote the scaled implementation of domestic AI computing power deserves continuous market attention.