Manycore Tech Open Sources Aholo Viewer for Billion-Splat 3D Internet and Spatial Computing

Manycore Tech has announced that it is open-sourcing Aholo Viewer , a 3D Gaussian rendering tool that enables users to effortlessly navigate large-scale 3D environments through a web browser. The platform works across smartphones, desktop computers, and VR headsets, without requiring any installation. It is capable of rendering city-sized 3D scenes containing more than one billion Gaussian splats—an area comparable to the full expanse of Hangzhou’s West Lake. Advancing the Foundation for a 3D Internet “For over three decades, the internet has primarily existed in 2D forms such as text, images, and video,” said Huang Xiaohuang, co-founder and chairman of Manycore Tech. “Physical spaces are reduced to a few photos, and real-world objects become flat representations. We have continually compressed rich spatial information into lower-dimensional formats. By open-sourcing foundational 3D technologies like Aholo Viewer, we hope to speed up the transition to a true 3D internet era.” Aholo Viewer is built on a streamable Level-of-Detail (LoD) system that allows 3D data to be delivered across devices in a way similar to video streaming. This enables users to instantly load and interact with extremely dense 3D environments directly in the browser, regardless of the device being used. Industry analysts point out that a key benchmark for 3D browsers is how many Gaussian splats they can efficiently render and how quickly they perform. For instance, Fei-Fei Li’s open-source Spark 2.0 engine supports streaming of over 100 million splats on consumer hardware. Aholo Viewer reportedly pushes this limit further, handling up to 1 billion splats—about ten times more—while also improving rendering efficiency, smoothness, and memory usage. As a result, high-performance 3D browsers may soon become the main entry point for everyday users to access immersive digital environments. In the near future, exploring a 3D world could feel as simple as scrolling through short videos: one click, and users can virtually “enter” a real or imagined space. From Visualization to Interaction: Expanding Spatial Computing For Manycore Tech, simply displaying 3D content is only the starting point. The company also aims to make these environments fully usable and interactive. Alongside Aholo Viewer, it is releasing APIs from its Aholo Spatial Intelligence Platform, including tools for spatial reconstruction, cloud-based rendering, and AI-driven 3D model creation. These are designed to remove bottlenecks in producing, sharing, and deploying 3D assets. With these capabilities, 3D environments can evolve from static visualizations into dynamic, programmable assets integrated into real-world workflows. Tourism companies could recreate destinations digitally in a short time, filmmakers could build virtual production sets, and game developers could design immersive experiences based on accurate real-world spaces. By combining open-source visualization tools with spatial computing APIs, Manycore Tech is positioning itself as a key infrastructure provider for the emerging 3D internet ecosystem. Why the 3D Internet Matters for AI The impact of a 3D internet goes beyond simply upgrading digital content from flat to immersive formats. Its deeper significance lies in addressing a core weakness in artificial intelligence: limited spatial understanding of the physical world. Today’s internet is dominated by 1D and 2D data, meaning AI systems largely learn from flat representations. This contributes to a gap in capabilities—while AI can generate text and images effectively, it still struggles with spatial reasoning, physical accuracy, and real-world interaction. As technologies like Aholo Viewer make 3D content easier to create and distribute, these long-standing barriers could be reduced. If producing and consuming 3D environments becomes as simple as sharing short-form videos, vast amounts of spatial data could become available. This, in turn, would provide valuable training material for embodied AI systems such as robotics and other agents designed to operate in the physical world.