Listen to this Post
In a decisive move that signals the next wave of embodied‑AI development, Chinese startup GigaAI—also referred to as Jijia Vision in recent reporting—has secured roughly ¥100 million (around ¥1 billion CNY, roughly 20 billion usd) in its Series A round from investors including Hubble Technology Investment (tied to Huawei) and TH Capital.
36kr
Earlier this year, in late August, GigaAI had already closed a pre‑Series A (and an additional round), raising several hundred million usd.
36kr
With fresh funding secured, GigaAI is reportedly scaling up the supply of its “world model” systems—virtual, high-fidelity digital environments which serve as training grounds for AI-driven humanoid robots. The company says this marks a breakthrough in three major technical bottlenecks, making a highly available world-model system accessible.
36kr
As companies globally race to merge AI cognition with physical embodiment, GigaAI’s move comes amid surging interest in embodied intelligence and robot‑driven physical AI. In China alone, investment in humanoid robotics and embodied AI startups has exploded in 2025, with 114 deals recorded in just the first five months—already exceeding the 77 deals across all of 2024.
TMTPOST
+1
the Report
GigaAI (Jijia Vision) recently closed a substantial Series A funding round, raising roughly one hundred million usd (~20 billion JPY) from Hubble Technology Investment and TH Capital. This follows earlier funding earlier in the year during a pre-Series A and an additional round, which together injected several hundred million usd into the company. The infusion appears to be earmarked for expansion of its “world model” infrastructure. The world model refers to a virtual, simulation-based environment that can replicate realistic spatial, physical, and sensory conditions for training embodied AI—such as humanoid robots. The company claims it has broken through three major technical barriers to deliver a “highly available” world-model system, indicating readiness for broader deployment. This move aligns with a larger trend: in 2025, investment in China’s embodied-intelligence and humanoid-robotics sector has surged markedly, signaling growing confidence that world-models can overcome long-standing limitations in embodied AI, especially around data scarcity and poor generalization.
What Undercode Say: The Significance and Implications
GigaAI’s recent funding and world‑model rollout represent more than just another round of corporate financing. They cast a spotlight on a tectonic shift in how AI is developed and deployed: from text-based, disembodied LLMs to physically grounded, embodied intelligences capable of perceiving, moving, and acting in real-world environments. The emergence of a “world model system” is a critical inflection point for robotics and embodied AI — it could transform how we train robots, accelerating their learning while bypassing the old bottleneck of laborious real-world data collection.
First, consider the problem: training a humanoid robot typically requires extensive real-world interaction, costly sensors, and time-consuming trial-and-error. That makes scaling up both expensive and slow. By creating high-fidelity virtual environments, companies like GigaAI dramatically reduce the reliance on real-world robotics data. This not only accelerates development cycles but also allows training across many more scenarios than physical testing alone could realistically afford. In effect, virtual worlds become training grounds that can simulate complex, diverse, even hazardous environments safely and at scale.
Second, the fact that GigaAI attracted major investment—especially from firms tied to Huawei—demonstrates how seriously China is investing in physical AI. The recent surge in deals in the embodied-robots sector (114 deals in the first five months of 2025) suggests a robust belief that embodied intelligence will be the next frontier, not just in commercial robotics but in AI’s broader evolution.
TMTPOST
+2
36kr
+2
Third, from a technological ecosystem perspective, world‑model frameworks could spur new forms of embodied AI beyond humanoids: industrial robots, virtual agents in VR/AR settings, and even autonomous agents in simulated environments and gaming worlds. As virtual and physical data converge, we may soon see robots and AI agents capable of intuitive, adaptive behavior — not just preprogrammed tasks but dynamic decision-making in complex environments.
However, this shift raises important questions. How realistic and reliable are these virtual environments? Do they faithfully replicate the physics, variability, and unpredictability of the real world? Overreliance on simulated data could produce brittle AI — good in simulation, but flawed when transferred into reality. The quality of world models and the fidelity of simulation will matter enormously.
Moreover, as more funding flows into embodied AI, we may see increasing consolidation: only well-backed players with both capital and computational resources will advance. This might narrow the field of innovation, concentrate power, and reduce open or community-driven robotics research.
Finally, there are societal and ethical implications. As embodied AI becomes more realistic and ubiquitous, the lines between virtual training, real-world deployment, and human–robot interaction may blur. From labor displacement to privacy, from safety concerns to inequality in access — the consequences of bringing robots into daily life are wide-ranging.
In short, GigaAI’s funding and world‑model initiative mark both a technological milestone and a signal that embodied AI may mature faster than many expect.
Prediction
Expect 2026 to be a breakout year for embodied‑AI startups — driven by world‑model systems, more humanoid robot prototypes, and possibly the first robots capable of performing complex, real-world tasks trained entirely in simulation. We may also see rapid consolidation: only a few big players will dominate, while standards and best practices for simulation-to-reality transfer emerge.
Fact Checker Results
✅ The report of GigaAI (Jijia Vision) raising roughly ¥100 million from Hubble Technology Investment and TH Capital is confirmed.
36kr
✅ The surge in Chinese investment in embodied intelligence and humanoid robotics in 2025 — 114 deals in first five months — has been documented.
TMTPOST
+1
✅ The claim that world‑model systems can ease data scarcity and generalization issues in embodied AI is supported by recent academic work (e.g. “GigaWorld-0” framework).
arxiv.org
+1
🕵️📝✔️Let’s dive deep and fact‑check.
References:
Reported By: xtechnikkeicom_c5d4d930cfa36c853186961e
Extra Source Hub (Possible Sources for article):
https://www.quora.com
Wikipedia
OpenAi & Undercode AI
Image Source:
Unsplash
Undercode AI DI v2
Bing
🔐JOIN OUR CYBER WORLD [ CVE News • HackMonitor • UndercodeNews ]
📢 Follow UndercodeNews & Stay Tuned:
𝕏 formerly Twitter 🐦 | @ Threads | 🔗 Linkedin | 🦋BlueSky | 🐘Mastodon
