LG Innotek is collaborating with NVIDIA to build sensing solutions for physical AI and robotics. At GTC Berlin on October 20, it unveiled new sensing products, including a 3-megapixel (MP) camera module and a Multi-modal Vision System (MVS), designed to use NVIDIA's Holoscan Sensor Bridge (HSB) technology for real-time sensor connection to GPUs. HSB is an open-source, Ethernet-based sensor fabric that sends sensor data directly to an AI computer without passing through the CPU. The companies plan to expand into vision sensing and robotic hands for humanoid robots, with solutions to be unveiled at next year's CES and GTC.
Summary written by Upside AI from the original article. Read the source for full details.
Read the original on Yahoo Finance ↗
Keywords
Nvidiaartificial intelligenceroboticsSouth Korea
Open a keyword to see other news on the same topic.
Related tickers and likely impact
NVDA
• NeutralNvidia
Relevance: low
| Market | US stocks |
|---|---|
| Sector | Information Technology |
| Industry | Semiconductors |
| Market cap | $5.2T |
| Close on 10/21/2026 | 221.87 |
| Change that session | +2.49% |
Impact is scored by AI from the headline and may be wrong. It is not investment advice.
Related news
- YYForce Unveils "Workforce OS" Robotics Architecture Powered by NVIDIA Physical AI
YYForce (NASDAQ: YFOR), a provider of workforce and integrated facility management solutions, announced plans to develop Workforce OS, a modular and hardware-agnostic robotics software platform. The platform uses NVIDIA technologies including Jetson, RTX GPUs, Project GR00T, Isaac Sim, Isaac ROS and TensorRT, and will integrate Unitree humanoid robots. The phased roadmap starts with intelligent concierge services in months 0–2, moves to mobile manipulation and beverage service in months 2–4, and reaches a banquet and hospitality event MVP with multi-robot operations in months 4–6. The long-term goal is autonomous facility maintenance and industrial cleaning, and CEO Mike Fu said the company aims to own the software intelligence layer for commercial robots and has formed a dedicated engineering task force.
- What's unique about Gecko Robotics' contracts with Nvidia & the US Navy: CEO
Gecko Robotics CEO Jake Loosararian says the company's US Navy contract has grown from $71 million awarded in April to $300 million, more than quadrupling, to help the military branches stay ready. The robots gather data that feeds the company's AI systems, helping shipyards and dry docks repair and maintain ships faster and return them to service. Loosararian says the company has deployed in critical environments for 13 years and has a robust security architecture to keep sensitive data private, including with the Navy. A recently announced Nvidia partnership adds deployment of AI agents in these environments, and the Navy-specific robot Komodo was built alongside experts and designed to keep humans in control. He argues the technology gives workers insights and repair capabilities they never had before, claiming China does not have any of this.
- Bear Robotics Launches KINO-2 Through Engineering Collaboration with NVIDIA
Bear Robotics announced KINO-2, a robot-learning platform that turns task demonstrations into deployable manipulation skills on its KR1 humanoid robot, extending its robots from moving goods to handling them. Bear has deployed over 16,000 robots, mostly service robots, at over 5,000 customer locations, and is piloting manipulation tasks such as supermarket returns and warehouse crate handling. KINO-2 uses NVIDIA Isaac OS as the core of its perception runtime, and with TensorRT and quantization, Bear cut GR00T inference time from about 500 milliseconds to around 60 milliseconds, roughly an 8x speedup in internal measurements. Bear's in-house model also reached a 4.2x speedup, falling from 72.3 to 17.2 milliseconds. KINO-2 is being launched at NVIDIA GTC Berlin, where Bear is demonstrating KR1 running it.
- Sereact, NVIDIA, Nebius, CEVA Logistics and Zalando Close the Training Loop for Robotic Returns Handling
Sereact, NVIDIA, Nebius, CEVA Logistics and Zalando announced a closed training loop for Cortex, Sereact's robot brain, first deployed on Zalando returns handling at selected CEVA Logistics fulfillment centers. Episodes recorded at the returns stations are stored on Nebius, turned into thousands of simulated variants with NVIDIA Isaac Lab, and re-rendered across garment states, packaging, materials and lighting with NVIDIA Cosmos before Cortex is post-trained on Blackwell GPUs. A new policy is scored on held-out variants against a threshold set by Sereact; passing policies go to a real-world test at the stations, while failing ones trigger targeted variants for missed cases and another training run. An AI agent builds and runs the loop, previewing GPU cost and checking the gate, while Sereact engineers set goals and approve runs above a cost limit. The partners say a fix that once took weeks can now return as a tested policy update within days.
- NVIDIA Advances Isaac Platform for General-Purpose Robotics With New Reference Humanoid, Frameworks, Simulation Pipelines and Data
NVIDIA has unveiled updates to its Isaac platform for general-purpose robotics, including Isaac OS, Isaac Lab 3.0 and open source reference workflows. Agile Robots introduced Agile ONE SN, a humanoid reference platform built on Isaac GR00T and Jetson Thor, scheduled for availability in early 2027 with planned integration of the Halos safety stack. European players such as ABB Robotics, KUKA and Universal Robots are also building on NVIDIA's open platform. NVIDIA says these technologies give developers a faster path from data to real-world deployment.
- Reply Earns "NVIDIA Physical AI Specialization" for Production-Ready Robotics and Digital Twins
Reply [EXM, STAR: REY] is among the first NVIDIA partners worldwide to earn the Physical AI Specialization within the NVIDIA Partner Network, recognizing its expertise in deploying AI-driven robotics and digital twins in real operating environments. The award followed a technical assessment covering a customer implementation, solution architecture, certified OpenUSD expertise and its development roadmap. A current example is Roboverse Reply's work with Otto Group and NVIDIA on a virtual control system at the Hermes Fulfilment logistics centre in Loehne, Germany, which connects autonomous mobile robots from different manufacturers to fleet-management and warehouse systems. Simulating and validating new layouts and workflows in NVIDIA Omniverse and Isaac before deployment speeds integration, reduces the risk of testing in live operations, and provides a blueprint for other Otto Group sites.
Related markets
Latest news
- SpaceX Just Crushed Wall Street's Q3 Targets -- but a Potential Tesla Merger Is Even Bigger News
SpaceX reported third-quarter adjusted EPS of $0.45 on sales of $14.58 billion, far above analyst averages of $0.13 and $12.72 billion. Revenue rose 161% year over year and net income swung to $4.4 billion from a $1.1 billion loss a year earlier; Starlink sales grew 69% to $5.25 billion, while the superintelligence segment rose about 830% to $7.56 billion. The company guided first-quarter 2027 sales to $33–37 billion, and Musk said he expects annual recurring revenue of $500 billion–$1 trillion by the end of next year. On the call, Musk said special committees had been formed to explore a merger with Tesla, that he could not take part due to conflicts, and that merging was a good idea, though shareholder approval equal to 50% of total shares would be a significant hurdle.
- Chip stocks break a bear trap after a $2.6 trillion rout — here are the levels bulls need to hold
The PHLX Semiconductor Index (^SOX) is down about 3% after dipping to 11,000, a support level that has rescued chip stocks twice since September, so the bear-trap rebound only holds if that floor does. Since Oct. 6, all 60 stocks in Yahoo Finance's chip basket have fallen, erasing about $2.7 trillion in market value; Lam Research and Applied Materials are each down nearly 30%, AMD nearly 20%, and Nvidia 9%. Lam fell despite record revenue because it said supply limits would cap its outlook and China sales would be “meaningfully down” in December. The Nasdaq 100 is down about 6% over the same period. The article flags 10,450–10,550 as the zone to watch if 11,000 breaks, and 12,600–12,700 as the level bulls must reclaim, while semiconductor ETFs saw record flows over the past week.
- A New Mexico Judge Just Hit Meta With a $25 Billion Penalty
On October 21, 2026, a New Mexico judge ordered Meta to pay the state $25 billion over misleading Facebook users about privacy in the Cambridge Analytica case, far above the FTC's $5 billion fine in 2019. Jurors found 26 of 29 statements misleading, and counting each statement per user or resident who saw it produced more than 43 million violations, with state law allowing up to $5,000 each; all the money goes to New Mexico schools. The judge said the conduct went to the heart of Facebook's core business and found that senior leaders, including Zuckerberg and Sandberg, misrepresented its privacy practices. Meta is appealing, has paid nothing, and interest keeps accruing, while the penalty equals about 2.5 times its $9.86 billion third-quarter net income.
- Elon Musk says Tesla's Optimus robot needs to learn like a toddler
Elon Musk said on Tesla’s earnings call that Optimus humanoid robots must learn through self-play at scale, since Tesla lacks the vast car fleet that supplies training data, likening this to how a toddler learns to navigate the world by playing. Tesla cut its 2026 capex outlook to about $20 billion from $25 billion as it prioritizes usable humanoids. Optimus V3 production began at Fremont in the summer after Model S and Model X output ended there, with about 1,000 Gen 3 units already operating on the production floor. William Blair analyst Jed Dorsheimer said Musk pushed the timeline for a 1 million-unit weekly run-rate to over a year, with first applications in dangerous industrial settings by the end of 2027.
- Meta's new stock sale sends one signal: AI demand isn't slowing down
Brian Sozzi of Yahoo Finance sees Meta's new non-voting share class as a positive signal. CFO Susan Li said the class can be used for capital raising, employee pay and acquisitions while preserving the voting power of existing Class A and B shareholders. Sozzi expects Zuckerberg to remain CEO for at least 15 more years, and says raising capital through stock may be cheaper than borrowing, since debt costs could rise for borrowers such as CoreWeave. He also reads the sale as a sign that AI demand is not slowing, pointing to better ad placement and monetization that helped the quarter's results.
- Elon Musk: SpaceX Revenue Could Hit “Somewhere Between $500 Billion and a Trillion” by 2027
On the earnings call, Elon Musk said SpaceX’s annual recurring revenue could reach between $500 billion and a trillion by the end of 2027, a goal that requires building compute faster than almost any company has tried. SpaceX posted its first quarterly profit of $4.4 billion in the latest quarter, versus a $1.1 billion loss a year earlier, with revenue up more than 160% to $14.6 billion; capex of $19.6 billion could reach $40 billion next quarter. Management says new compute installations pay back in under a year if prices hold, but the target of 13 GW by end-2027, up from 1.7 GW, depends on power, cooling, and NVIDIA’s Vera Rubin systems arriving on time. Shares rose 7.79% to $159.31, and a Tesla combination would need approval from 50% of Tesla’s outstanding shares, a hard bar given many holders never vote.
Get the briefing on Telegram or Zalo
Join the Upside Telegram channel or Zalo group for the Vietnam stocks and gold briefing at 8:00 and 19:00 (Vietnam time) on trading days, plus a weekly wrap on Saturday morning.