Manycore's Lux3D Aims to Build 3D Worlds in 20 Seconds

HANGZHOU, China – August 27, 2026 – Manycore Tech Inc. today pulled back the curtain on Lux3D, a generative AI model that can create detailed 3D assets from text or images in as little as 20 seconds. While the blistering speed is set to disrupt industries from gaming to e-commerce, the launch is more than just a product release; it's a bold declaration of the company's strategic path toward building true AI world models—not just from pixels, but from explicit three-dimensional geometry.

The announcement comes on the heels of staggering financial growth for the Hong Kong-listed tech firm. According to its H1 2026 interim results, revenue from new AI applications surged 177% year-over-year, with adjusted net profit climbing an impressive 211%. This financial momentum, coupled with a 120% increase in capital expenditure on computing infrastructure, signals that Manycore is not just experimenting with AI but is aggressively investing to become a foundational player in the next generation of spatial intelligence.

The New Economics of 3D Content Creation

For years, the creation of high-quality 3D assets has been a significant bottleneck for developers, designers, and digital marketers. The process is traditionally slow, expensive, and requires specialized expertise. Manycore's Lux3D, particularly with its "Turbo Mode," directly confronts this challenge. The ability to generate a model in under half a minute promises to fundamentally alter production workflows.

"3D Gaussian technology has been an important area of exploration for Manycore," said Tianze Long, Senior Director of the company's AI business. "Its combination of rendering fidelity and computational efficiency makes it particularly relevant to applications that require rapid previews or large-scale asset generation."

This efficiency is not just for one-off creations. Through its API and a "Harness Mode," developers can integrate Lux3D into tools like Blender and generate assets in large batches, automating what was once hours of painstaking manual labor. For the booming game development industry, this means potentially populating vast virtual worlds with unique assets at an unprecedented scale. In e-commerce, it allows for the rapid creation of 3D product models for virtual try-ons and augmented reality showrooms, a market projected to reach over $9 billion by 2032. The impact extends to architecture, industrial design, and film, where rapid prototyping and visualization are paramount.

Beyond Pixels: A Bet on Explicit 3D Worlds

While the immediate practical applications of Lux3D are compelling, the true significance of the launch lies in Manycore's long-term vision. The company is positioning itself as a proponent of an "explicit 3D approach" to building world models—the complex virtual environments AI uses to learn about and predict outcomes in the real world.

Many current world models learn by predicting the next frame in a 2D video, essentially becoming masters of pixel patterns. This can lead to impressive-looking videos but often reveals a shallow understanding of reality, resulting in spatial inconsistencies and illogical object behavior. Manycore is betting on a different, more foundational path. By generating entire 3D worlds, complete with object geometry, material properties, and spatial relationships, its AI aims to develop a more robust and intuitive grasp of physical reality.

"World models are beginning to follow different technical paths. Manycore has consistently pursued an explicit 3D approach—one that generates not just pixels, but three-dimensional worlds," stated Dr. Zihan Zhou, Chief Scientist at Manycore Tech. "Our starting point is to align digital and physical worlds across more dimensions, enabling digital environments to incorporate more signals from the physical world."

This strategy is about building AI that can interact with its environment, not just observe it. The roadmap includes combining Lux3D with SpatialGen, its scene-generation technology, to create interactive 3D environments. The company also plans to add dynamic properties to its generated assets, enabling them to have component relationships and motion logic. The ultimate goal, as Dr. Zhou added, is to "help AI systems trained in digital environments act more effectively in the physical world." This has profound implications for the future of embodied AI, robotics, and digital twins.

Under the Hood: Speed, Fidelity, and Unanswered Questions

The headline-grabbing speed of Lux3D is backed by a deep focus on technical quality. A key emphasis is on material fidelity—the model's ability to realistically simulate how light interacts with surfaces like metal, glass, and plastic. This is crucial for creating assets that look convincing under any lighting condition, a persistent challenge in generative 3D. By training the model on a mix of generated and reconstructed physical-world data, Manycore aims to ensure both geometric accuracy and visual consistency, mitigating common problems like non-manifold geometry that plague many AI-generated models.

However, while the company reports competitive results from internal evaluations, it has yet to release detailed public benchmarks. The generative 3D space is advancing rapidly, and independent verification against offerings from competitors like NVIDIA and others will be critical for a full assessment of Lux3D's market position. The tech community will be watching closely to see if the generated models hold up to the scrutiny of professional workflows without requiring significant post-production cleanup.

Still, the focus on solving core industry problems is clear. By addressing both the speed of creation and the quality of the final asset, Manycore is tackling the generative 3D challenge from both ends.

Fueling the Engine: Financial Muscle and Market Strategy

Manycore's ambitious technological roadmap is underwritten by formidable business performance. The company, founded in 2011 and known for its Kujiale spatial design platform, has successfully translated its expertise in 3D design into a thriving AI business. The reported 177% year-over-year revenue growth in its AI division for the first half of 2026 is a testament to the market's appetite for its spatial intelligence solutions.

This growth is fueling a virtuous cycle of investment. The 120% jump in capital expenditure, largely directed at expanding its GPU and computing power, demonstrates a firm commitment to providing the necessary horsepower for its resource-intensive AI research and development. Investors have taken note, with the company's stock seeing a nearly 7% rise following the H1 2026 update. This financial strength provides Manycore with the runway to pursue its long-term, capital-intensive vision for explicit 3D world models, differentiating it from smaller startups or research labs. By evolving from a design software provider to a foundational spatial intelligence infrastructure company, Manycore is making a calculated and well-funded play to define how AI perceives, interacts with, and ultimately builds our digital and physical worlds.