On October 27, LG Uplus and Ecolab Korea signed an MOU in Seoul to jointly develop liquid cooling operation standards for AI data centers, build a comprehensive operation framework covering equipment installation, coolant quality management, and real-time monitoring, and conduct technical verification of the TCS loop. LG Uplus will provide the liquid cooling pilot environment at its Pyeongchon 2 Center, while Ecolab Korea will establish a coolant quality management system.
Meta announced on October 13 that it is rolling out the AI-powered "Your Algorithm" feature on Facebook, allowing users to describe preferences in natural language to influence Feed and Reels recommendations in real time. The feature was previously launched on Instagram, and Facebook's three-dot menu has also been updated to make it easier for users to manage content preferences.
LG Electronics' North American branch has reached a long-term supply agreement with HVAC services company AIR Control Concepts for advanced chillers for AI data centers totaling over 5GW, accelerating LG's push into the global AIDC cooling market.
At a Microsoft industry summit in Seoul's Gangnam District, LG Electronics showcased a smart-home AI voice assistant developed with Microsoft. It runs on the ThinQ ON smart home hub, uses speech-to-speech agent technology with Microsoft Voice Live, supports barge-in, and is planned for release in 2026, with existing users receiving the update and use cases expanding to B2B spaces like hotels, offices and retail stores.
On September 22, LG Electronics announced that its 2.5MW coolant distribution unit (CDU) received NVIDIA DSX Ready certification, granting technical access to NVIDIA's AI factory-related projects. Only three 2.5MW-class CDUs worldwide have passed the certification. LG is expanding liquid cooling products such as chillers and cold plates around the CDU, and its Korean-listed shares rose nearly 7% after the announcement.
LG and Microsoft have agreed to expand their strategic partnership to AI data centers, physical AI, and AI transformation. They will explore LG providing cooling, power, and IT infrastructure for Microsoft's data centers, while LG plans to adopt Microsoft's AI and cloud platforms including Copilot and accelerate physical AI development using Microsoft cloud infrastructure.
LG Electronics announced a new air-cooled oil-free variable-speed turbine chiller optimized for AI data centers, featuring self-developed oil-free magnetic bearings and natural refrigerant pre-cooling technology, claiming a 30% reduction in annual power consumption, along with machine learning-based anomaly detection and 180-second ultra-fast restart capability.
<body><div>OpenRouter raised $113M in Series B funding, led by Alphabet's CapitalG, with participation from NVIDIA, ServiceNow, MongoDB, Snowflake, Databricks, a16z, and M</div></body>
Apple, in collaboration with the University of Hong Kong, has introduced a new rendering framework named LGTM, which stands for "Less Gaussians, Texture More." This framework is designed to overcome the rendering bottlenecks associated with achieving 4K ultra-high-definition visual quality. The core innovation of LGTM lies in decoupling the geometric structure of 3D scenes from surface textures. This approach simplifies geometric complexity while achieving high-quality visuals through layered texture application. Traditional methods, like feed-forward 3D Gaussian splatting, can quickly convert 2D images into 3D scenes but face exponential computational demands at higher resolutions, straining device capabilities. To address this, LGTM employs a two-step strategy. First, the model learns the scene structure from low-resolution images, ensuring a perfect geometric framework for 2K or 4K rendering. Subsequently, a second network focuses on detail processing, precisely mapping high-resolution textures onto the geometry. This division of labor effectively avoids the full computational pipeline consumption typically required. For everyday users, this technology promises to significantly enhance the Apple Vision Pro headset experience. With combined displays offering approximately 23 million pixels—exceeding the pixel count of a 4K TV per eye—existing rendering techniques often create performance bottlenecks, leading to lag or blur. LGTM aims to resolve this computational challenge, delivering smoother and more immersive visuals. In practical demonstrations, 3D scenes generated with the LGTM framework show marked improvements in texture detail and text clarity, producing near-photorealistic results. This advancement indicates that future headset users will enjoy more realistic and sharper visual experiences in virtual environments.