（本文作者为 Chelsea_Sun，钛媒体经授权发布）
Related articles
TMTPOST -- China's robotics startup X2 Robotics has confidentially filed an initial public offering application with the Hong Kong Stock Exchange (HKEX).
According to investment banking sources, a large number of embodied-intelligence companies are currently lining up at the HKEX, and X2 Robotics is the only robot maker positioning itself around a fully self-developed, full-stack, end-to-end embodied model.
In the embodied-intelligence space, most “in-house models” are essentially open-source models that have undergone post-training and then been repackaged. Behind X2 Robotics’ “original” label, however, its model was pre-trained entirely from scratch.
Mainstream VLA (Vision–Language–Action) models in embodied intelligence are, in essence, a patchwork of multiple independent modules. Information is passed along step by step like an assembly line—and it degrades with every handoff.
X2 Robotics, by contrast, was the first in the industry to explicitly propose a “World Unified Model Architecture” (World Unified Model, WUM). This architecture gives the model native multimodal capabilities from day one of training, eliminating the need for “message relaying” and translation between modules.
WUM is widely seen as a major leap for embodied foundation models—from a stitched-together VLA architecture to a natively fused multimodal architecture. Public information to date shows that X2 Robotics is the only company that has clearly proposed and put into practice a “World Unified Model Architecture.”
In May this year, X2 Robotics also open-sourced its embodied foundation model Wall-OSS-0.5, achieving a world-first: after pre-training, it can be deployed directly onto real robots without any task-specific fine-tuning.
Early in its founding, X2 Robotics’ founder Wang Qian chose to build China’s first large-scale robot data factory in Shenzhen. In July this year, X2 Robotics also released the QUANXTA Zero series—an embodiment-free data-collection solution. Starting from the real data requirements of model training, it reverse-designed data-collection equipment and built a complete data pipeline spanning collection, cleaning, automated labeling, and quality control.
System-level efforts such as original, innovative R&D; full-stack autonomy and controllability; and building an in-house data factory require substantial investment from X2 Robotics. At the same time, the infrastructure and capability-building have also helped make it a darling of the capital markets.
From late April 2026 to the end of June, in just a little over two months, it completed and closed four consecutive rounds of financing—B, B+, B++, and C—pushing its valuation rapidly from 10 billion to over 20 billion yuan.
Its fundraising drew lead investments from four internet giants—Meituan, Alibaba, ByteDance, and Xiaomi—as well as top-tier VCs such as Sequoia China and IDG Capital, alongside “national team” players including China Mobile and China Development Bank Capital.
This breakneck fundraising pace not only provided ample ammunition for X2 Robotics’ technology R&D, but the influx of top-tier capital also further validated its market value—paving the way for its eventual IPO.
In the view of investment bankers, Invariant Robotics has been highly sought after in the primary market, but a valuation of RMB 20 billion could mean higher difficulty and costs in raising funds. Going public is therefore an inevitable choice to open up a broader financing channel.
Even as Invariant Robotics confidentially submitted its listing application to the Hong Kong Stock Exchange and races toward the capital markets, it has not slowed down product iteration. TMTPost has learned from sources that Invariant’s third-generation robot product is set to be released in Q4. This embodied-intelligence company is pushing on two fronts—“technology + capital”—as it sprints at full speed along the industry track. (Reporting by Qian Yujuan and Editing by Yang Lin)
更多精彩内容，关注钛媒体微信号 （ID：taimeiti），或者下载钛媒体 App
