With the 2026 World Artificial Intelligence Conference (WAIC) just concluded, the "humanoid robot shows" from various exhibitors remain impressive—Unitree's humanoid robot G1 performed difficult backflips, while Zhiyuan Expedition A2 demonstrated smooth human-machine interaction. However, the scene at one company's booth was completely different.
GEEKPLUS-W (ASX: 02590) had no dancing, no walking, no conversation. On the display floor, a wheeled robot named Gino 1, an unmanned picking workstation, and several mobile robots were cooperating with high precision, quietly picking up items like potato chips, bread, and rubber bowls one by one from storage bins and accurately placing them into corresponding positions on a seeding wall. There was no superfluous "performance," only a complete, end-to-end operational process running stably—sorting, transporting, delivering—each step as fluid and silent as in an actual warehouse or factory. The movements weren't particularly dazzling—but industrial scenarios don't need dazzle; they need stability.
Among the more than 150 embodied AI exhibitors at this year's WAIC, Geek+ was the only one to simultaneously demonstrate multi-SKU generalization capability and end-to-end operational efficiency in a real work scenario. The core ambition of Geek+ is not merely to build better robots, but to become the "solution expert" for the era of embodied intelligence—leveraging over a decade of accumulated real-world experience to provide practical, replicable, and scalable embodied AI solutions for countless industries. This represents a strategic leap from "champion product" to "expert service," and is the key to understanding Geek+'s long-term value.
The Gravity Framework: An "Operating System for Embodied Intelligence"
When Amazon AWS quietly launched in 2006, no one thought it was a big deal. At the time, everyone's attention was on Google and YouTube—the dazzling applications that truly "changed the world." Looking back twenty years later, AWS's market value surpassed the combined worth of almost all the star applications of that era.
Today, the wave of embodied intelligence is sweeping in at a staggering pace. While many companies are keen to showcase the most cutting-edge humanoid robot demos, Geek+ is taking a contrarian approach, focusing on "real machines doing real work" by demonstrating the stable performance of its wheeled robot Gino 1 in complex, multi-SKU, and generalized scenarios. Behind this apparent "ordinariness" lies a profound strategic judgment: the future of embodied intelligence depends not on the flashy skills of individual robots, but on the ability to build a unified, reliable, and open "operating system."
Geek+'s key announcement at this WAIC is a unified embodied intelligence framework named Gravity. Gravity adopts a "dual-brain collaboration" architecture: the cognitive brain is responsible for understanding instructions and decomposing tasks; the action brain conducts physical "sandbox simulations" before execution, predicting outcomes to generate the optimal sequence of actions. Chen Chao, head of Geek+'s Embodied Large Model team, pinpointed the industry's core contradiction at the WAIC opening: "Models that understand language and semantics can't calculate physical motion accurately, and models that can predict physical motion lack high-level semantic understanding."
More crucial is the technical path chosen for Gravity 4D. It moves beyond merely "replicating visuals," strengthening 3D spatial and 4D physical motion representation to balance reasoning speed with cross-material generalization capability. It employs whole-body parallel coordinated control, synchronously adjusting grasping posture with movement, improving single-task efficiency by approximately 30%. Data validation shows the LIBERO-Plus benchmark zero-shot success rate increased from 73.73% to 78.62%, with the most significant improvements in scenarios involving camera viewpoint changes, sensor noise, and lighting variations—cases where "appearance changes but physics remains unchanged."
The GINO ECO Ecosystem: Builder and Evangelist of "Standards"
Technological iteration inevitably requires ecosystem support. The "One Core, Dual Engines" strategy launched by Geek+ at this juncture tightly integrates the Gravity framework (the core) with the data engine and ecosystem engine (the dual engines), outlining its platform ambitions. If OpenAI's moat is internet text data, Geek+'s moat is physical-world interaction data and an open ecosystem engine. This is not a gap that can be quickly closed by financing and hiring.
Geek+'s 11-year accumulation of a real-world business network—covering over 40 countries, more than 1,700 projects, and serving over 950 internationally renowned brands (including Walmart, Adidas, Siemens, BMW)—generates millions of real-world SKU picking data points daily. This constitutes a nearly insurmountable data moat that is expanding at an accelerating rate.
Concurrently, Geek+ launched the GINO ECO open ecosystem: opening its hardware platform to algorithm companies, collaborating with leading customers to explore innovative scenarios, and partnering with complementary hardware firms to build end-to-end workflows. Many interpret this as a "win-win posture." However, to understand its deeper logic, consider another historical case: Android. By opening Android to all phone manufacturers, Google seemingly relinquished hardware control, but in reality, it traded openness for control over the mobile internet gateway, ultimately fostering the prosperity of the entire Android ecosystem.
Geek+'s logic is identical. The more open GINO ECO becomes, and the more algorithm companies and hardware firms that join, the more unshakeable the Gravity framework's position as the "interface standard" becomes. As Geek+ founder and CEO Zheng Yong stated, "Embodied intelligence is a marathon, not a sprint. We are willing to be the super-connector for the entire industry." By linking various segments, Geek+ is constructing a comprehensive picture of embodied intelligence with "Gravity as the core, data as the fuel, and the ecosystem as the branches and leaves."
The True Pricing Opportunity Lies in the Perception Gap
Behind the strategic ambition lies a proven commercial logic. For the 2025 fiscal year, Geek+ delivered a report card that crossed the profitability inflection point: revenue reached 3.171 billion yuan, a year-on-year increase of 31.6%; adjusted net profit was 43.82 million yuan, achieving a turnaround to profitability; operating cash flow significantly turned positive. This is a company that has passed the profitability inflection point, not one still burning cash to tell a story.
Simultaneously, management has conducted consecutive share buybacks for over 15 trading days since June 25, accumulating 16.2 million shares repurchased—a signal that speaks for itself.
An explosive growth of the embodied intelligence market is a consensus within the industry. A Goldman Sachs research report predicts the global humanoid robot market will reach $38 billion by 2035; Citi is even more optimistic, estimating the global humanoid robot market space could hit $154 billion by 2035.
Benefiting from the industry's tailwinds and the company's fundamentals, all eight covering brokerages for Geek+ have issued Buy or Outperform ratings. Among them, CITIC Securities has the highest target price of HK$53, with a comprehensive average target price around HK$24.6—implying an upside potential of over 120% from the current share price.
However, more important than the price gap is the shift in the valuation framework. The market currently values Geek+ within a "warehouse AMR hardware company" framework. But Geek+ is transforming into a "physical AI operating system" company. The valuation gap between these two frameworks is not a 20% or 30% correction, but potentially a fundamental reconstruction of the pricing logic—similar to what happened when the market repriced NVIDIA from a "gaming graphics card company" to "AI computing infrastructure."
Comments