Tianshu Smartcore was the first to break through the "valley of death in mass production", and the hearing in Hong Kong Stock Exchange is just the beginning.

On December 19th, Tensentech went through a hearing at the Hong Kong Stock Exchange, and if nothing unexpected happens, it will officially go public in January 2026. Previously, Moore Threads and Muxi Group had successively debuted on the ChiNext board, and their stock prices soared after listing, igniting unprecedented enthusiasm in the market for domestic GPUs. Now, Biteng Technology and Tensentech have both successfully passed the hearing at the Hong Kong Stock Exchange. With this, the four leading domestic companies known as the "GPU four dragons" have all gathered on the capital market stage, ready to engage in a fierce competition for the future computing power landscape. When we temporarily strip away the high premiums brought by market hype on companies like Moore Threads, the uniqueness of Tensentech lies in its value proposition based on mass production and delivery instead of narrative concepts. As the industry-recognized "domestic general-purpose GPU mass production breakthrough", it was not the first to propose a grand technology blueprint, but one of the few players to actually send 7nm high-performance general-purpose GPU chips into data centers and achieve large-scale commercial deployment. Its Tianzai series and Zhikai series products have been deployed in key industries such as finance, energy, education, and healthcare, forming a verifiable revenue loop and customer rebuy. Today, the market's attention to Tensentech has surpassed its IPO price or short-term stock price fluctuations, focusing more on the value proposition represented by Tensentech. What lies behind this is a fundamental showdown in the GPU industry about valuation logic: in the world of hard technology investment, is it the forward-looking technological narrative that dominates pricing power, or the solid commercial deployment anchoring value? A new perspective on valuation framework: technology vs. commercialization Whether in terms of industry evolution or market perceptions, the dynamic growth logic exhibited by Tensentech has broken the traditional framework of incremental development for hardware companies, presenting a nonlinear leap driven by the conversion power of technology. In 2023, the company's revenue grew by over 60% year-on-year, with GPU shipments exceeding 50,000 units - a figure that not only far surpasses its peers but also makes it the only company in the industry willing to publicly disclose detailed shipment data. More importantly, this scale marks Tensentech's successful crossing of the "mass production death valley" threshold, becoming the first Chinese company to truly transition general-purpose GPUs from the lab to large-scale commercial deployment. This is no longer a parameter competition but a response to the ultimate proposition for hard technology companies: can technology be converted into real productivity? While some peers still focus on "chip success" or "architecture release" as valuation anchors, Tensentech has chosen to use the doubling of shipments, auditable customer lists, and continuous industry orders as its strategic moat. According to data from Frost & Sullivan, among Chinese chip design companies, Tensentech is the first company to achieve mass production of inference general-purpose GPU chips, the first to achieve mass production of training general-purpose GPU chips, and the first to achieve such milestones using advanced 7nm process technology. Data is the best evidence: Revenue jumped from 189 million yuan in 2022 to 540 million yuan in 2024, with a three-year compound annual growth rate of 68.8%; revenue in the first half of 2025 was 324 million yuan, a 64.2% year-on-year increase, indicating strong growth continuation; Annual shipments of general-purpose GPUs increased from 7.8 thousand units in 2022 to 16.8 thousand units in 2024, with a single quarter shipment of 15.7 thousand units in the first half of 2025, totaling more than 52,000 units delivered, leading the industry in scale delivery capacity. Mass production is not the end, but the beginning of value. The true barrier for Tensentech lies in the ability of its products to obtain validation from multiple industries. By the end of June 2025, the company had served over 290 enterprise customers in various fields such as finance, healthcare, transportation, education and research, and retail. This cross-industry, multi-scenario deployment capability not only diversifies market risks but also confirms the chip's "general-purpose" attributes, contrasting sharply with competitors focused on a single track. Supporting this expanded capability is its complete product matrix of "training and inference integrated": the Tianzai series focuses on training, with the Tianzai Gen1 being China's first mass-produced general-purpose training GPU, and Gen3 released in Q3 2024; the Zhikai series is designed for inference, with products entering the market rapidly in 2023 and covering all cloud-edge-end scenarios. The closed-loop coverage of training and inference allows Tensentech to provide customers with full-stack computing support from model development to application deployment, establishing a deep product moat and continuous growth momentum. User-centric thinking: The key to Tensentech's crossing of the mass production gap Why was Tensentech able to achieve large-scale delivery first? The underlying logic lies in the scarce "user-centric" mindset: focusing on "usable, easy-to-use, and willing to use." This pragmatic approach, in the transition from "having or not" to "whether it's good to use and cost-effective" in the current era of domestic substitution, is particularly valuable. Tensentech places user needs at the core, focusing on scenario applications, improving product yield through various means, optimizing the supply chain, enhancing scene adaptability, and improving the toolchain to make users truly adopt and actively use domestic general-purpose GPUs that are both effective and user-friendly. This shift to user-centric thinking, making their product the most popular domestic general-purpose GPU product among customers, is the key to their success in crossing the mass production gap. Therefore, the core narrative of Tensentech's IPO is not another groundbreaking technological prophecy, but a certainty report on large-scale commercial deployment. Its true breakthrough lies in elevating "doubling shipment volume" from a mere financial growth indicator to the company's strongest strategic moat. In the field of hard technology investment, the traditional valuation logic often focuses on the depth of research and development investment, the depth of the patent pool, and the advancement of chip design parameters. The risk of this "heavy asset" model is the potential entrapment in a "no sales, all technology" laboratory dilemma, with high R&D costs that cannot be diluted through large-scale sales, ultimately leading to valuation ambiguity. Tensentech has chosen a path of "heavy delivery", with the core focus of its valuation logic on actual shipment volume, customer retention rate, and compound revenue growth rate. Between 2022 and 2023, its revenue saw a leap in growth, particularly from cloud training and inference solutions, reflecting the direct manifestation of the strong growth in shipments. Each successful delivery and stable operation of a chip is a stringent customer validation, proving that the product not only exists on PowerPoint slides but can withstand real workloads and undergo the critical transition from "sample" to "commodity." This growth curve supported by shipment volumes has multiple strategic implications. Firstly, it signifies that product strength has gained market endorsement, creating genuine customer stickiness. Secondly, large-scale shipments have honed end-to-end capabilities from design, chip production to packaging testing, particularly accelerating the process of domestic substitution in the supply chain - Tensentech has deep collaboration with leading domestic semiconductor foundries, forming crucial supply chain resilience in a complex international environment. Lastly, the continuous revenue inflow provides valuable "blood" for the research and development of next-generation more advanced products, forming a virtuous cycle of "sales - research and development - iteration." Therefore, the significance of tens of thousands of shipments goes beyond market ranking. It signals that a company has crossed the "Darwinian Sea" from technological success to business success and entered a new phase driven by continuous evolution through market feedback. Reassessment of value in three dimensions For investors, the case of Tensentech provides a crucial observation: when evaluating a hard technology company, should one chase distant and grand ecological stories or focus more on its clear and expanding profitability path and delivery capacity? The market's final choice will profoundly affect the valuation paradigm for future hard technology investments. In fact, traditional valuation models (such as P/S or P/E) struggle to measure long-term value for ecosystem-level companies. Tensentech's valuation needs to be redefined in three dimensions: Short-term: The present value of cash flows brought by mass production certainty. With a scale of over 50,000 units and a compound growth rate of 68.8%, Tensentech is the only domestic GPU company with the ability to "self-generate revenue"; Medium-term: Continuous increase in market share for general-purpose GPUs. If Tensentech can capture 20% market share in the training + inference market, its gross profit margin could rise to 65% (from the current 58%), showcasing exponential growth Long-term: Compatibility beyond the CUDA platform, adaptation to domestic operating systems, databases, large models, application software, and other ecosystem layers will build a complete domestic ecosystem, further expanding the market space. The endorsement by top strategic investors further validates this logic. Tensentech's investors not only provide funds but also deeply collaborate in scenarios and resources. This "industry + capital" dual-driven strategy shifts Tensentech's valuation logic from "technological expectations" to the certainty of mass production + ecosystem pricing. Since Tensentech has not publicly disclosed its final offering price, the outside world cannot currently determine its market value. Yet, with examples like Moore Threads and Muxi Group - both of which saw their stock prices increase by over six times post-listing, with current market valuations around 300 billion yuan - the market is even more promising for Tensentech, which demonstrates better commercial deployment and more trust from domestic users. Conclusion As global computing power competition enters deep waters, the ultimate answer for domestic GPUs ultimately lies in commercial grounds. Tensentech, as the "mass production disruptor", has used doubled shipment volume, steady revenue growth, and cross-industry customer coverage to prove one thing: in the world of hard technology, the true barrier lies not only in technological advancement but also in the ability to convert technology into products and deliver continuously. Its commercial practices provide a landing template for the domestic GPU industry, connecting the entire process from "technology - product - market" and driving the industry as a whole from "chip design" into "commercial deployment." In an era where computing power is becoming a cornerstone of national core competitiveness, companies that can deliver reliable computing power on a large scale are becoming the backbone of China's computing power breakthrough.