{"id":79254,"date":"2026-10-22T19:24:45","date_gmt":"2026-10-22T23:24:45","guid":{"rendered":"https:\/\/www.globalvillagespace.com\/tech\/?p=79254"},"modified":"2026-10-22T19:24:45","modified_gmt":"2026-10-22T23:24:45","slug":"hesai-kosmo-spatial-camera-robot-training-data","status":"publish","type":"post","link":"https:\/\/www.globalvillagespace.com\/tech\/hesai-kosmo-spatial-camera-robot-training-data\/","title":{"rendered":"Hesai launches Kosmo spatial camera and cloud to turn real-world scans into robot training data"},"content":{"rendered":"<p>Hesai Group announced Kosmo on October 14, 2026: a combined hardware-and-cloud offering the company calls a 3D world data engine. Kosmo pairs a capture device (Kosmo One) that Hesai describes as an \u201cAI spatial camera\u201d with Kosmo World, a cloud platform that converts scans into high-fidelity 3D assets for robot training, world-model development and media production.<\/p>\n<h2>What Hesai says Kosmo does<\/h2>\n<p>According to Hesai\u2019s product materials and company statements reported at launch, Kosmo has three parts:<\/p>\n<ul>\n<li>Kosmo One \u2014 an imaging system Hesai bills as the world\u2019s first 8K ultra-high-definition \u201cAI spatial camera\u201d that fuses high-resolution LiDAR, a one-inch imaging sensor and a high-precision inertial measurement unit (IMU) with microsecond-level synchronization.<\/li>\n<li>Kosmo World \u2014 a cloud platform and processing pipeline that reconstructs captured data into 3D Gaussian Splatting (3DGS), mesh and depth outputs, plus semantic segmentation and generative fills for unscanned areas.<\/li>\n<li>Kosmo Spatial Core (KSC) \u2014 the backend engine Hesai says ties capture, reconstruction and export together, and supports integrations with tools such as NVIDIA Isaac Sim, Isaac Lab, Unreal Engine, Pixar RenderMan and Houdini.<\/li>\n<\/ul>\n<p>Hesai claims millimeter-level geometric accuracy from its 3DGS reconstruction, instant scanning on power-up, intelligent guidance for operators, one-click scan resumption and exportable outputs up to 6K resolution. The company says a single operator can scan a typical 100-square-meter indoor space in about 15 minutes, and that Kosmo World processes scenes \u201cup to approximately five times\u201d faster than comparable products, while producing reconstruction types (3DGS, mesh and depth) in one pass.<\/p>\n<h2>Demonstrations and early deployments reported by Hesai<\/h2>\n<p>Hesai presented several use cases at launch. Notable examples it provided include:<\/p>\n<ul>\n<li>A Dairy Queen (DQ) store deployment where Hesai says manual modeling previously took ~300 hours but Kosmo scanning and reconstruction took 15 minutes; in that example, Hesai reported the Sharpa robot\u2019s task success rate reached 99% and failure rate fell to one-tenth of the prior level.<\/li>\n<li>Film and virtual production workflows using Pixar\u2019s RenderMan; Hassai said RenderMan released a tutorial and that film companies are already using Kosmo to shorten set creation timelines and cut production costs.<\/li>\n<li>Partnerships with embodied-AI firms Galbot and Dexmal, and a mapping collaboration with Amap (reconstructions of the Gravity Art Museum and PIFO Gallery are live in Amap\u2019s Flying Street View feature, Hesai says).<\/li>\n<li>Cultural-heritage capture with TRIC for a Du Fu Thatched Cottage artwork project.<\/li>\n<\/ul>\n<h2>What is verifiable from the evidence packet<\/h2>\n<p>From the supplied company and press-report materials we can confirm the following facts as claims made by Hesai or reported about the product at launch:<\/p>\n<ul>\n<li>Hesai publicly launched Kosmo on October 14, 2026 and described it as a 3D world data engine built around Kosmo One and Kosmo World.<\/li>\n<li>Hesai describes Kosmo One as an 8K imaging system combined with LiDAR and IMU, producing synchronized geometry, color and position data, and says Kosmo World outputs 3DGS, mesh and depth in a single reconstruction.<\/li>\n<li>Hesai presented claimed performance numbers: ~15 minutes to scan 100 square meters, up to ~five-times faster processing versus comparable products, and millimeter-level geometric accuracy via 3DGS (all stated by the company and reported in the coverage).<\/li>\n<li>Hesai positioned Kosmo for embodied-AI training, world-model development, filmmaking and mapping, and announced partner names (Galbot, Dexmal, Amap, Pixar RenderMan, TRIC) at launch.<\/li>\n<li>Hesai is expanding beyond LiDAR into actuation modules and spatial-intelligence services; its financial disclosures reported growth in LiDAR shipments and early revenue from Strategic Growth Initiatives (SGI), including orders for Kosmo after prototype deliveries in July 2026 (company-cited results reported in financial coverage).<\/li>\n<\/ul>\n<h2>Where the evidence is limited or where claims need independent verification<\/h2>\n<p>The supplied materials are primarily company statements and press reports summarizing the launch. The evidence packet does not include independent third\u2011party benchmarks, peer-reviewed tests, raw sample datasets, device pricing, production volumes, or audited security assessments. Specific points that require outside verification before being treated as established facts include:<\/p>\n<ul>\n<li>Millimeter-level accuracy and the processing speed advantage versus competitors. These are claimed by Hesai but not independently validated in the provided sources.<\/li>\n<li>The 15-minute scan time for a 100-square-meter indoor space and the Sharpa robot\u2019s 99% task success figure \u2014 both are provided by Hesai as an example deployment; the underlying test protocol, repeatability, and baseline comparisons are not included in the materials supplied.<\/li>\n<li>Commercial availability, pricing, global production scale and lead times. The evidence does not specify whether Kosmo One and Kosmo World are broadly commercially available, nor the device price or service pricing tiers.<\/li>\n<li>Security model and data governance. The packet shows integrations and open-data positioning, but it contains no independent cybersecurity evaluation or explanation of how raw capture data, exports and cloud-hosted assets are protected, who holds keys, or how third-party customers retain control of derived assets.<\/li>\n<\/ul>\n<h2>Context: Hesai\u2019s business and strategic rationale<\/h2>\n<p>Hesai\u2019s move into a vertical stack for physical-AI data aligns with its broader strategy to diversify beyond LiDAR sensors into higher-value software and robotic hardware. Financial reporting cited in the packet shows the core LiDAR business remains profitable and growing \u2014 second-quarter revenue of RMB860.8 million with LiDAR shipments increasing \u2014 while Strategic Growth Initiatives (SGI) had early revenue but generated operating losses as the company invests to scale actuation and spatial services. The company stated its <a href=\"https:\/\/finance.yahoo.com\/technology\/articles\/hesai-hsai-moving-beyond-lidar-233613779.html\" rel=\"nofollow noopener noreferrer\" target=\"_blank\">Kosmo platform secured orders after prototype deliveries in July<\/a> 2026 and expects Kosmo-related revenue to begin contributing in subsequent quarters.<\/p>\n<h2>Security and policy background readers in the United States should note<\/h2>\n<p>Independent of Kosmo, US regulatory and national-security attention has focused on Hesai as a company: reporting in 2024 noted the U.S. Department of Defense designated Hesai as a Chinese military entity on a blacklist. That designation prevents Pentagon contracts but, according to reporting, does not prohibit U.S. commercial firms from using Hesai products. The available packet includes commentary raising cyber risk concerns about Chinese-made LiDAR being used in U.S. systems; those are security perspectives attributed to third parties rather than validated statements about Kosmo\u2019s specific product security.<\/p>\n<h2>Practical checklist for teams evaluating Kosmo for robot-training or mapping projects<\/h2>\n<p>Based on the product claims and typical commercial evaluation needs, here is a practical checklist procurement and engineering teams should use before adopting Kosmo or similar spatial-capture offerings:<\/p>\n<ol>\n<li>Request raw sample outputs: point clouds, 3DGS assets, mesh and depth exports from representative indoor and outdoor scenes; validate geometric accuracy against ground-truth measurements.<\/li>\n<li>Benchmark processing speed end-to-end on your target scenes and compare with your current pipeline (measure capture-to-ready-for-training time).\n<\/li>\n<li>Obtain detailed test protocols behind any vendor success claims (for example, the Dairy Queen case): number of runs, robot policy used, baseline performance and failure modes.<\/li>\n<li>Confirm export formats and integrations with your simulation and training stack (e.g., Isaac Sim, Unreal Engine), plus camera-model adaptation options for your robot sensors.\n<\/li>\n<li>Assess data governance and security: who stores raw capture data, where it\u2019s hosted, encryption-at-rest and in-transit, access control, and audit logs; request a security whitepaper or third\u2011party penetration test results.\n<\/li>\n<li>Clarify licensing, IP ownership and whether generative fills or semantic labels are retained, shared or sold by the vendor.\n<\/li>\n<li>Test labeled object accuracy for downstream perception: run your segmentation and detector models on vendor outputs and measure transfer gap versus real-world captures.\n<\/li>\n<li>Plan for operational constraints: battery and power needs for field scanning, environmental limits (bright sun, reflective surfaces), and operator training time.\n<\/ol>\n<h2>How Kosmo compares, in principle, with credible alternatives<\/h2>\n<p>Hesai positions Kosmo as an integrated capture+cloud pipeline focused on embodied-AI training and production workflows. That vertical approach is increasingly common among spatial-data companies: alternatives typically split the workflow between handheld or mobile capture hardware, third-party reconstruction software and separate cloud services. Hesai\u2019s claimed differentiators are integrated millimeter-level 3DGS accuracy, one-pass generation of multiple reconstruction outputs, and faster processing. These differentiators should be validated by direct tests against comparable capture systems and reconstruction pipelines before procurement decisions.<\/p>\n<h2>Bottom line for U.S. robotics and embodied-AI teams<\/h2>\n<p>Kosmo is a clear statement of intent from Hesai to move up the stack from sensors to spatial data services and robot actuation. The product is presented as commercial-facing and partnered with multiple firms in robotics, mapping and media. But the launch materials in the packet are company-led claims: teams considering Kosmo should request technical benchmarks, security documentation and pilot contracts that permit real-world validation at scale. Given existing policy scrutiny of Hesai at U.S. national-security levels, procurement teams should include legal and security review in their vendor-evaluation process and follow the checklist above to reduce operational and data-risk exposure.<\/p>\n<blockquote><p>&#8220;Kosmo combines Kosmo One, the world&#8217;s first AI spatial camera, with the Kosmo World cloud platform to transform physical environments into high-fidelity 3D digital spaces for robot training, world model development and immersive content creation.&#8221;<\/p><\/blockquote>\n","protected":false},"excerpt":{"rendered":"<p>Hesai unveiled Kosmo \u2014 an \u2018AI spatial camera\u2019 (Kosmo One) and Kosmo World cloud \u2014 claiming rapid, high-fidelity 3D capture intended for embodied-AI training, film and mapping. The product is commercial-facing, but independent verification of key performance and security claims remains limited.<\/p>\n","protected":false},"author":1,"featured_media":79256,"comment_status":"open","ping_status":"closed","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[6023],"tags":[6028,7425,7427,7428,6091,6446,7426],"class_list":["post-79254","post","type-post","status-publish","format-standard","has-post-thumbnail","category-latest","tag-embodied-ai","tag-hesai","tag-kosmo","tag-kosmo-one","tag-lidar","tag-robots","tag-spatial-data"],"_links":{"self":[{"href":"https:\/\/www.globalvillagespace.com\/tech\/wp-json\/wp\/v2\/posts\/79254","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/www.globalvillagespace.com\/tech\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/www.globalvillagespace.com\/tech\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/www.globalvillagespace.com\/tech\/wp-json\/wp\/v2\/users\/1"}],"replies":[{"embeddable":true,"href":"https:\/\/www.globalvillagespace.com\/tech\/wp-json\/wp\/v2\/comments?post=79254"}],"version-history":[{"count":1,"href":"https:\/\/www.globalvillagespace.com\/tech\/wp-json\/wp\/v2\/posts\/79254\/revisions"}],"predecessor-version":[{"id":79255,"href":"https:\/\/www.globalvillagespace.com\/tech\/wp-json\/wp\/v2\/posts\/79254\/revisions\/79255"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/www.globalvillagespace.com\/tech\/wp-json\/wp\/v2\/media\/79256"}],"wp:attachment":[{"href":"https:\/\/www.globalvillagespace.com\/tech\/wp-json\/wp\/v2\/media?parent=79254"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/www.globalvillagespace.com\/tech\/wp-json\/wp\/v2\/categories?post=79254"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/www.globalvillagespace.com\/tech\/wp-json\/wp\/v2\/tags?post=79254"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}