Transwarp Technology passed the Hong Kong Stock Exchange main board listing hearing on August 27, with Haitong International serving as its sole sponsor, bringing the company within striking distance of achieving its A+H dual-listing goal. The company is an enterprise-grade AI and big data foundational software developer, offering clients a bottom-layer software platform for "storing data, managing data, computing data, and training models," delivering full-stack foundational software services across the entire data lifecycle from integration, storage, governance, modeling, analysis, mining, to circulation.
According to Frost & Sullivan data, by revenue, Transwarp Technology ranks as China's fifth-largest AI infrastructure software provider with a 2.7% market share in 2025, and stands as the largest independent service vendor in the market. However, this leading market scale has not translated into growth momentum, as the company's recent financial performance has been both volatile and persistently loss-making.
The prospectus reveals that from 2023 to 2025, Transwarp Technology generated revenues of approximately RMB 491 million, RMB 371 million, and RMB 447 million respectively, trending downward with fluctuations, while net losses stood at RMB 289 million, RMB 344 million, and RMB 245 million over the same period, accumulating total losses exceeding RMB 870 million over three years. In the first half of 2026, revenue reached approximately RMB 174 million, up 13.83% year-on-year, with net profit attributable to parent excluding non-recurring items at RMB -143 million, slightly narrowed from RMB -156 million in the same period of 2025, yet still reflecting substantial losses.
At this critical juncture of secondary listing, market concerns are brought to the forefront: why does Transwarp Technology continue to incur significant losses, and when can the company achieve a turnaround to profitability?
Software products and technical services revenue share climbs to 80%, with improving revenue quality. The revenue curve from 2023 to 2025 displays a clear pattern of "peak, decline, recovery," primarily driven by the impact of macroeconomic fluctuations, customer procurement cycles, and the pacing of large-scale project acceptance in the enterprise foundational software business. In 2024, revenue dropped 24.31% to RMB 371 million, mainly because macroeconomic headwinds made customers more cautious in procurement decisions, with bidding timelines delayed compared to previous years and extended acceptance processes, pushing overall revenue recognition later into the period.
As supporting evidence, in 2024, both the number of new orders and average order value per customer actually increased, with customers paying over RMB 3 million growing noticeably, and repeat purchases from existing clients contributing 78.83% of core business revenue—meaning demand itself did not collapse, but was instead "stuck" in the delivery and acceptance stages, forcing revenue recognition to be deferred. This fluctuation in "recognition timing" stems from a revenue structure heavily reliant on project acceptance, with primary clients concentrated in finance, government, and energy sectors, which typically enforce strict budget management and procurement approval systems.
Understanding this mechanism clarifies the logic behind the 2025 revenue rebound: growth of 20.6% to RMB 448 million, driven largely by previously backlogged and delayed projects passing acceptance and entering revenue recognition. In the first half of 2026, accelerated AI adoption across industries boosted procurement demand from government and state-owned enterprise clients, lifting AI and big data infrastructure software and related technical services revenue by 25.46% to RMB 139 million, which in turn drove overall revenue growth of 13.83% to RMB 174 million for the period. During the reporting period, the share of software products and technical services revenue further rose to 80.35%, with software product licensing and supporting services generating RMB 85.77 million, up 46.82% year-on-year; government sector revenue jumped from RMB 18.49 million in the prior-year period to RMB 50.09 million, emerging as a prominent incremental contributor.
The rising share of software products and technical services revenue, coupled with sustained strength in standardized software licensing growth, is steadily improving the company's revenue quality, enhancing its resilience to cyclical fluctuations. However, compared with revenue quality improvements, the profit side has shown little noticeable progress, with the core reason being the company's ongoing investment growth phase, where relatively small revenue scale combined with rigid costs that are difficult to compress results in substantial losses.
Taking 2025 as an example, although the gross margin reached a healthy 52.8%, R&D expenses, selling expenses, and administrative expenses accounted for 45.6%, 35.7%, and 24.82% of total revenue respectively, collectively totaling as much as 106.14%. The rigidity on the expense side stems from the need for continuous iteration of underlying databases, big data platforms, and AI infrastructure, as distributed architectures, multimodal capabilities, domestic IT adaptation, and generative AI data foundations all require sustained substantial capital investment. Cutting R&D spending would be akin to "killing the goose that lays the golden eggs," undermining long-term development prospects.
Additionally, given the highly specialized nature of enterprise-grade software, pre-sales solutions, implementation, and technical support all require dedicated personnel deployment, making selling expenses hard to reduce. Furthermore, as a growth-stage company, maintaining finance, legal, HR, internal controls, intellectual property, domestic certification, ecosystem partnerships, and overseas/subsidiary operations—combined with labor costs from custom implementation projects, low-margin solution businesses, and credit impairment charges—means administrative expenses also cannot contract in sync with quarterly revenue fluctuations. In the first half of 2026, although selling and administrative expenses declined nearly 10% and 15% respectively year-on-year, with ongoing optimization in organizational efficiency and cost control leading net profit attributable to parent excluding non-recurring items to narrow slightly from RMB -156 million to RMB -143 million, a clear gap to profitability remains.
Against this backdrop, rapidly scaling revenue emerges as the most direct and effective path to improving the profit side—only when revenue growth outpaces cost rigidity can economies of scale begin to manifest at the margin.
Domestic substitution anchors the base, while GPU-native databases unlock incremental upside. Looking ahead, whether Transwarp Technology can quickly scale revenue hinges on two fronts: first, whether domestic substitution can accelerate in key industries, and second, whether GPU-native cognitive databases can open new elasticity in AI inference and data services. Driven by both policy and industry tailwinds, the data asset management space is now witnessing rigid demand. On the policy front, the "AI+" initiative and national AI strategy continue to advance, with 2026 designated by the National Data Bureau as the "Year of Data Element Value Realization," focusing on public data resource registration, authorized operations, and pricing mechanisms, while advancing data infrastructure, trusted data spaces, and high-quality dataset construction per the 2026 Digital Economy Development Work Points. On the industry side, core systems in party/government, finance, and energy sectors are migrating from Oracle/DB2 to domestic databases, creating clear substitution demand.
To address growing domestic substitution needs, Transwarp Technology holds corresponding qualifications and products: its ArgoDB V6 passed joint security and reliability assessments by the China Information Security Evaluation Center and the National Secrecy Science and Technology Evaluation Center in May 2026, achieving Distributed Level I certification, serving as a critical admission and compliance reference for domestic database selection in party/government and key industries. Combined with privatized data foundations and AI platforms including TDH, TDC, TDS, and Sophon, the company possesses full-stack advantages among independent software vendors. The first-half 2026 government revenue surge from RMB 18.49 million to RMB 50.09 million, up over 170% year-on-year and representing 28.88% of total revenue, serves as early validation of government data infrastructure and digital government demand materializing. Should domestic substitution extend from peripheral to core transaction systems, and from standalone databases to integrated lakehouse-governance-AI solutions, software licensing and technical services are poised for steady volume growth, anchoring the company's fundamentals.
While domestic substitution volume growth represents a high-certainty revenue source, the GPU-native cognitive database stands as the key to Transwarp Technology's future earnings elasticity, as this product brings generational leaps and business model reconstruction. In July 2026, the company launched its cognitive database cloud service, running a full relational database on GPUs, with officially disclosed benchmark performance reaching 20 times that of dual-CPU solutions, 100 times better in TPC-DS testing on multi-GPU clusters, over 70 times faster in complete CRUD concurrency, and 20 to 200 times improvement in vector retrieval performance. This order-of-magnitude performance jump is not simple parameter optimization but directly addresses the core bottleneck of the AI inference era—large-scale model inference demands massive data throughput, where traditional CPU-based databases have become the bottleneck, while GPU-native architecture enables data to be fed "locally" to compute units.
More importantly, this product is driving Transwarp Technology to shift from "one-time License sales" to a "per-Token billing" cloud service model, charging for AI data engineering, data analytics, and intelligent agent development scenarios. Once paying customers achieve scale and usage ramps up, the revenue curve will transition from step-wise project-based patterns to compounding cloud service growth. Institutions estimate the potential market at approximately USD 21.6 billion to USD 43.2 billion based on calculations of "AI inference data volume × processing unit price × penetration rate × payment rate," which, while representing a theoretical ceiling rather than near-term revenue, outlines the height of this development path's ceiling.
This development path is currently at a critical window transitioning "from technical validation to scenario implementation." The cognitive database cloud service has already launched for trials, with clients in finance, manufacturing, transportation, retail, and biopharmaceutical industries beginning evaluations. Transwarp Technology is also collaborating with partners including AWS and NVIDIA for joint promotion, having already established overseas presence in Singapore, Canada, and Malaysia, with the company explicitly setting 2026 as the year to "establish benchmark customers" and 2027 as the stage for "scaled volume growth." As the core vehicle for high-growth elasticity expectations, the GPU-native cognitive database serves as key support for the company's valuation premium—but it cuts both ways: should commercialization of the GPU-native cognitive database fall short of expectations, the valuation premium faces significant compression pressure.