Anari AI Launched Thor X – the First “System-on-Cloud” for Processing 3D Point Clouds

By Tiera Oliver

Associate Editor

Embedded Computing Design

August 04, 2022


Anari AI Launched Thor X – the First “System-on-Cloud” for Processing 3D Point Clouds

A new Cloud-based technology “System-on-Cloud”, developed by the Anari AI team, introduces an optimized and efficient system from various different hardware and software architectures combined with machine learning models to revolutionize the way AI compute systems are utilized.

Anari AI is launching Thor X, the first  “System-on-Cloud” specialized in semantic segmentation of 3D point cloud data structures. Primarily focused on providing custom hardware acceleration to industries such as  geospatial, BIM, digital twin, and metaverse, Thor X is, according to the company, enabling 30x more efficient processing compared to the best GPUs on the market.

“3D applications are starting to gain more and more traction, from AR/VR, Retail, and autonomous vehicle use cases. I believe that quality data annotation of 3D point clouds is the key factor for numerous industries in order to provide value to the end users. AI plays a significant role in this regarding accuracy and speed, and will be crucial for semantic segmentation of the 3D point cloud in the future”, said Avi Yashar, co-founder and Chief Product Officer at Dataloop AI. 

General-purpose hardware like GPU or CPU has proven effective in processing 2D images, but it falls short when met with the complexity of 3D structures. In contrast, FPGA embedded Thor X is designed to specifically respond to the needs of 3D point cloud datasets obtained either by LiDAR scanning or photogrammetry. 

“The advantage of obtaining 3D point clouds by laser scanning is to obtain continuous, complete and metrically rigorous information of the object, producing with extreme accuracy and in three dimensions the topology of it”, said Davide Agrati, Product Manager for 3D Imaging Solutions at Trimble Italia. “This opens the world of opportunities in a wide variety of models such as BIM restitution, 3D surface and mesh models, metaverse virtual tours, and geospatial models”.

Benchmarked on one of the industry’s standard large scale 3D point cloud dataset, SemanticKITTI, Thor X’s combination of hardware, software, and machine learning model delivers up to 18x faster processing compared to CPUs. 

According to Jovan Stojanovic, Anari AI co-founder and CEO, Thor X reflects Anari’s goal to accelerate, simplify and optimize cloud computation for domain specific workloads in the cloud. 

“Every day, AI workloads are becoming more and more complex, and companies are struggling to achieve competitiveness and good ROI. Simply put, if things continue to develop further at this pace, AI won't be accessible to every company. That is why it is essential to simplify domain-specific computation, plus to make it possible for others to succeed by also designing their own domain-specific systems in a significantly easier way. Thor X is a big step forward in making a new industry category of “Systems-on-Cloud”, where 3D vision is the first step into the future”, said Stojanovic.

For more information, visit:

Tiera Oliver, Associate Editor for Embedded Computing Design, is responsible for web content edits, product news, and constructing stories. She also assists with newsletter updates as well as contributing and editing content for ECD podcasts and the ECD YouTube channel. Before working at ECD, Tiera graduated from Northern Arizona University where she received her B.S. in journalism and political science and worked as a news reporter for the university’s student led newspaper, The Lumberjack.

More from Tiera