Domestic DPU vendor Yunbao Intelligent submitted its IPO materials to the Shenzhen Stock Exchange on June 30, becoming the first DPU chip company to file under the ChiNext's fourth set of listing standards, and is widely regarded in the market as "China's first DPU IPO." However, as the prospectus disclosed its operating data, two points of contention emerged in the market: three consecutive years of net losses attributable to parent company shareholders, and heavy revenue reliance on a single major customer.

The implementation of the ChiNext's fourth set of listing standards has completely broken the requirement that companies must be profitable to go public, precisely accommodating hard-tech enterprises like high-end chipmakers that require heavy upfront investment and are not yet profitable. Breaking the "chokepoint" in chips is a core national strategic task. Among the three core computing chips in data centers, domestic substitution in the CPU and GPU tracks has already achieved initial results, with domestic players such as Haiguang, Huawei Kunpeng, Cambricon, and Huawei Ascend making continuous breakthroughs, gradually alleviating supply-side gaps. DPUs, however — as the third critical track in the computing infrastructure — have long been monopolized by overseas vendors. This is a key link in the AI industry chain where domestic substitution urgently needs to be addressed. This is also the key policy consideration behind the ChiNext's fourth set of listing standards, specifically aimed at supporting hard-tech DPU companies like Yunbao Intelligent.

On one hand, the market holds high expectations for Yunbao Intelligent as the "first DPU IPO." On the other hand, there is some skepticism. At its core, this skepticism applies the profitability and customer evaluation standards of mature consumer companies and traditional manufacturing to a high-end computing chip company still in the early stages of commercialization.

Looking at the domestic chip leaders that have already gone public, most of them went through a complete growth cycle before listing — with substantial R&D losses, reliance on key customers, and a sudden surge in revenue. The noise surrounding Yunbao Intelligent today is not a sign of operational risk, but rather the typical challenges inherent in the "3U" sector — heavy investment, long cycles, and a path of validation first, volume production later.

Yunbao Intelligent was founded in August 2020 and is a leading domestic independent vendor in China's full-featured DPU market, focused on R&D of data center data processor chips and solutions. Its growth trajectory stands out: revenue grew from just RMB 172,700 in 2023 to RMB 36.36 million in 2024, and then surged to RMB 370 million in 2025 — a year-on-year increase of over 900%, significantly outpacing comparable chip companies during the same period.

JiWei Network notes that behind this rapid revenue growth is the fact that the company has been in the red for three consecutive years, which has become a key point of concern in the capital markets. In fact, this is closely tied to the capital-intensive nature and long industrialization cycles of the DPU industry, with heavy R&D spending putting pressure on short-term profits. Yunbao Intelligent has invested a total of RMB 1.758 billion in R&D over the past three years, with R&D spending alone accounting for 150.12% of revenue in 2025.

In traditional investment logic, consecutive losses are often seen as a sensitive risk signal. But does this standard still apply when applied to domestic chip companies listed on the A-share market?

Take Cambricon, a domestic GPU leader, as an example. It listed on the STAR Market in 2020 as "China's first AI chip IPO." Its prospectus showed that from 2017 to 2019, revenue grew from RMB 7.84 million to RMB 444 million — a 50-fold leap — but cumulative losses over those three years exceeded RMB 1.6 billion, largely due to sustained high R&D investment. Cambricon quickly refined its chip technology, iterated through adaptations, closed the commercialization loop, and gradually diversified its customer base. By 2025, its revenue had grown 453.2% year-on-year to RMB 6.497 billion, reaching its profitability inflection point. Similarly, domestic CPU leader Haiguang also faced profit pressure before its IPO, but after product deployment across multiple scenarios and expanding its customer base, it quickly achieved stable profitability.

In the computing ecosystem, the "3U" (CPU, GPU, DPU) each have their own roles, together forming the core hardware foundation of the modern computing industry. DPUs, as the core chip of next-generation intelligent computing centers, require massive upfront R&D investment and go through long technology iteration and product validation cycles. Early-stage losses and sustained investment are an inevitable norm in industrial development, and a necessary path for all leading chip companies.

Clearly, Yunbao Intelligent's operating profile fits the common pattern of early-stage hard-tech companies — "heavy R&D investment and periodic losses." Looking beyond short-term book figures, its growth trajectory is clearly positive.

As for the large loss in 2025 that has drawn market attention, there is a clear non-operating factor. Data shows that Yunbao Intelligent's net losses attributable to parent company shareholders for 2023–2025 were RMB -667 million, -610 million, and -1.189 billion respectively. The 2025 loss includes a one-time share-based compensation expense of RMB 643 million. Excluding non-recurring items, the recurring net loss for 2025 was only RMB 562 million — and has been narrowing year over year.

For hard-tech companies, short-term losses are the necessary cost of product development. And top-level policy support provides a solid foundation and room for long-term growth.

In October 2023, the Ministry of Industry and Information Technology and five other government departments jointly issued the Action Plan for High-Quality Development of Computing Infrastructure, which explicitly stated: "For scenarios such as intelligent computing, supercomputing, and edge computing, carry out technology upgrades and pilot applications of data processing units (DPUs), lossless networking, and other technologies, to achieve high-performance transmission in computing center networks." This policy incorporates DPUs into the core national computing infrastructure framework, providing solid policy support and clear direction for the standardized, large-scale, and accelerated development of the domestic DPU industry.

Looking back, Yunbao Intelligent's development history may be short, but its growth rate and technological achievements have set rare industry records. It completed its first angel round in 2021. In just over two years, it achieved a one-pass successful tape-out of its DPU chip, and its A0 version product went directly to mass production without a single transistor modification. With strong product performance and technical strength, Yunbao Intelligent's DPU series products passed rigorous evaluation by the Ministry of Industry and Information Technology, were selected for the "14th Five-Year Plan" National Manufacturing Achievements Exhibition, and were displayed in the "National Treasures of Technology" section at the National Museum of China — a clear testament to its technical capabilities and national-level industrial value.

DPUs, as core acceleration chips for data centers and AI computing, are not general-purpose standardized products. They require deep software and hardware integration with server clusters, cloud operating systems, and computing infrastructure, with long cycles for system debugging, testing, and iteration. Looking at the development patterns of overseas DPU leaders, they typically require three generations of technology iteration and up to seven years of R&D to achieve technological maturity and commercial deployment. Yunbao Intelligent, however, has achieved rapid development in less than three years — a feat made possible by the background of its founder Dr. Sunny Siu and the team he has assembled.

Dr. Siu is one of the few founders in China's large-chip industry who combines top-tier academic credentials with a proven track record of entrepreneurial success. He received his Ph.D. in Electrical Engineering from Stanford University at the age of 24. His doctoral dissertation tackled a classic theoretical problem in AI that had remained unsolved for over thirty years. His research findings were compiled into the book Discrete Neural Computation: A Theoretical Foundation, which received a foreword endorsement from Marvin Minsky, the father of AI. His breakthrough research earned him the Young Investigator Award from the National Science Foundation. He later focused on networking and distributed computing research and served as an endowed-chair associate professor at MIT. Before founding Yunbao Intelligent, he had already gained large-chip startup experience — his previous venture, a networking processor company in Silicon Valley where he was a co-founder, was later acquired by Broadcom for USD 3.7 billion. In addition, Yunbao Intelligent's core team comes from Broadcom, Intel, ARM, HiSilicon, Alibaba, and other companies, covering networking chips, cloud computing, and system architecture.

The company has established deep partnerships with leading customers such as Tencent and China Mobile. Leveraging rich business scenarios, massive computing demands, and stringent deployment standards, it has continuously optimized its technology stack, rapidly adapting to domestic computing center application scenarios and significantly shortening the cycles for technology iteration and product deployment.

Proactively deepening strategic partnerships with leading customers is the core commercialization path for computing industry players to accelerate product iteration and expand growth opportunities. Tencent Cloud, as a leading domestic cloud service provider (CSP), offers critical support for Yunbao's DPU product optimization and iteration through its accumulated experience in general computing, AI computing, storage, security, and other core scenarios, along with product-level ecosystem collaboration. DPU solutions adapted for Tencent Cloud can meet the majority of general-purpose scenarios and functional requirements in China's cloud industry, laying a solid product foundation for Yunbao's expansion to other cloud service providers and telecom operators.

Tencent Cloud's rigorous mass-production standardization and quality control systems have enabled Yunbao to scale from small-batch gray releases to full-scale network deployment, achieving mass production at the 100,000-card level. This process has fully validated the large-scale production capabilities of Yunbao's DPU products, providing a reference for scalable deployment by customers such as China Mobile.

It is also worth noting that Tencent is an organizer and leader of industry technology summits in the cloud sector. Yunbao's DPU can participate in the development of industry standards such as ODCC super nodes and pooling solutions, as well as next-generation technology iteration, effectively enhancing product industry recognition and go-to-market advantages. Currently, powered by large-scale deployment in Tencent's AI computing clusters, Yunbao's self-developed 400Gbps full-featured DPU chip has undergone rigorous scenario testing and performance refinement.

Industry observers note that after completing validation with leading customers, Yunbao Intelligent's customer diversification has accelerated significantly, establishing a "multi-dimensional matrix" covering telecom operators, government and enterprise sectors, finance, and OEMs. Recently, Yunbao Intelligent entered into a deep strategic partnership with Super Fusion, embedding its DPU products into the full range of server equipment to deliver domestic computing solutions to government, enterprise, and industry clients nationwide. As its customer base continues to expand and downstream application scenarios broaden, Yunbao Intelligent is gradually building a well-balanced, multi-tiered, high-quality customer structure — with a clear development trajectory.

From an application perspective, the DPU sector offers strong extensibility within the three-pillar computing architecture. As the core foundation for AI computing, cloud computing, and digital infrastructure, DPUs are not limited to data center scenarios — they can also be widely applied in 5G/6G edge computing, network security encryption, industrial systems, intelligent storage, and other diverse tracks.

As previously mentioned, Yunbao Intelligent's cumulative DPU chip shipments have exceeded 100,000 units, fully validating the product's stability, advanced capabilities, and reliability. The company has already initiated R&D on next-generation 800G and 1.6Tbps DPU chips, and its chip design workflow has fully transitioned to an "All in AI" approach, with the goal of becoming China's most efficient chip design company. As its high-end products continue to iterate and reach the market, Yunbao Intelligent is well-positioned to enter more high-value niche scenarios, opening up new growth opportunities.

Within the "3U" concept, China's domestic CPU track has players such as Haiguang and Loongson, while the GPU track has Cambricon, Moore Thread, and MXMACA — most with market capitalizations in the hundreds of billions of RMB. Notably, in the DPU track, Yunbao Intelligent currently holds the leading position (ranked first among domestic independent vendors in China's full-featured DPU market by 2025 revenue). With the global DPU market tightly controlled by international giants such as Nvidia and Broadcom, and the domestic substitution rate for high-end DPUs still low in China, Yunbao's scarcity is particularly pronounced.

According to a Frost & Sullivan report, China's DPU market was approximately RMB 50 billion in 2025 and is expected to expand to around RMB 129 billion by 2030 — a trillion-yuan market is within sight. In this vast blue ocean, Yunbao Intelligent is one of the few companies in China to have achieved full-process independent R&D and large-scale commercialization of DPU chips, ranking first in sales among domestic independent full-featured DPU vendors. Backed by China's enormous computing market, and after a short period of technology accumulation and product validation, it has quickly established itself in the top tier of the domestic DPU industry.

It is worth noting that traditional listing standards, with profitability as the core evaluation metric, have often struggled to accommodate the development characteristics of frontier industries such as semiconductors and AI — "heavy upfront investment, steady mid-term growth, and high late-stage returns" — leaving many high-quality tech enterprises locked out of the capital markets. This April, the implementation of the ChiNext's fourth set of listing standards completely broke the "profitability lock," precisely aligning with the development attributes of future and emerging industries. Yunbao Intelligent, with a post-money valuation of RMB 14.273 billion, annual revenue of RMB 370 million, and cumulative R&D investment exceeding RMB 1.7 billion over three years, perfectly meets the high R&D access requirements of the fourth set of listing standards, with all core metrics significantly exceeding the listing thresholds.

With the RMB 3.035 billion raised in this IPO, Yunbao Intelligent will fully invest in next-generation full-featured DPU R&D and industrialization projects, next-generation AI DPU R&D and industrialization projects, and next-generation high-performance DPU core technology R&D projects, continuously narrowing the technology gap with international giants. As its product iterations mature, volume shipments increase, and customer structure continues to optimize, Yunbao Intelligent estimates that it will reach profitability as early as 2028 (this does not constitute a performance commitment), transitioning from high R&D investment to high returns — a trajectory similar to that of listed companies Cambricon and Haiguang.

From an industry investment perspective, evaluating hard-tech companies requires stepping away from the short-term profit mentality of traditional manufacturing — not judging success by "short-term profitability." As a scarce domestic DPU asset, Yunbao Intelligent's periodic losses and relatively high customer concentration during its IPO phase are typical challenges faced by China's high-end chip industry in its pursuit of independent breakthroughs. Driven by industry policy support, the dividends of computing industry development, and domestic substitution demand, the company's long-term growth potential deserves continued attention, and it is well-positioned to become a core benchmark for China's independent and controllable computing chips.