Monday, November 16, 2026 Edition No. 240
Search
Daily News Fully automated · 4 min read

NVIDIA AI Updates: November 9, 2026

Fully automated. Found, summarized and published by an automated AI pipeline. How posts are labeled

NVIDIA, Korean chip startup Rebellions and SK Telecom announced on November 8 that Rebellions will use NVLink Fusion to connect its next-generation Rebel R200 neural processing units to the NVIDIA platform, linking R200s inside a rack over NVLink scale-up networking and building the systems on NVIDIA’s MGX rack architecture and reference designs. Rebellions says the R200 is designed for low-latency, energy-efficient inference on frontier models, including multitrillion-parameter mixture-of-experts models. SK Telecom plans to deploy the systems in its AI data center alongside its own models, including A.X K2, and public services built under Korea’s “AI for All” project. The deal is one part of a wider collaboration with Korea’s Ministry of Science and ICT (MSIT), which prepares the ground for Korean NPU and XPU makers to explore joining the NVLink Fusion ecosystem, adds NVIDIA training, certification, hackathons and seminars on GPU utilization, and has NVIDIA working with three selected AI for All consortia and four Korean foundation model builders on inference using Dynamo. For infrastructure teams, it is a concrete case of a third-party inference accelerator being designed into NVIDIA’s rack-scale interconnect instead of competing with it. Source Source

2. Dynamo Recipes for Four Korean Open Models, K-EXAONE-2.5 API Access and OpenShell Pilots at Hyundai and AIM Intelligence

NVIDIA released Dynamo deployment recipes on November 8 for four Korean open models: SK Telecom’s A.X K2, LG AI Research’s K-EXAONE 2.0, Upstage’s Solar Open 2 and Motif Technologies’ Motif-3. The recipes are tuned for B200 GPUs and, according to NVIDIA, deliver up to 2x performance out of the box by handling the request routing and inter-GPU communication needed when a model is split across multiple GPUs or nodes. In a companion post, NVIDIA said LG AI Research is opening API access to K-EXAONE-2.5, which was developed using Nemotron datasets including Nemotron-Personas-Korea, an open set of 7 million fully synthetic Korean personas grounded in official Korean statistics. On the agent side, Hyundai Motor Group is exploring NemoClaw blueprints and the OpenShell runtime, and AI security company AIM Intelligence is integrating its Starfort agent control platform with OpenShell to pair OpenShell’s sandboxed execution and policy-based access controls with Starfort’s detection of malicious instructions and data leakage, with a possible preliminary proof of concept in the public energy sector. For teams serving these Korean models, the recipes give a tested multi-GPU starting configuration instead of hand-tuning Dynamo. Source Source

3. Samsung Cuts a Chip Thermal Simulation From 18 Hours to 41 Seconds With PhysicsNeMo, as Kakao Mobility and LG Expand Physical AI Work

NVIDIA detailed physical AI projects with three Korean companies on November 8. Samsung Electronics is using cuISS and cuDSS for sparse linear systems in technology computer-aided design, reports that its sMOL materials platform ran selected quantum chemistry calculations up to 5x faster end to end with cuEST, and is exploring cuEquivariance for atomistic AI models. In a PhysicsNeMo chip-scale thermal and mechanical analysis case, Samsung combined AI-predicted effective material properties, a coarser grid and GPU acceleration to cut simulation time from about 18 hours (fine grid on a 32-core CPU) to 41 seconds, roughly a 1,600x speedup. Kakao Mobility plans to build a level 4 autonomous driving system on NVIDIA Hyperion with DRIVE AGX Thor and use NVIDIA Halos for safety validation, starting with Kia purpose-built vehicles and drawing on its driving experience in Seoul’s Gangnam district. LG Electronics and NVIDIA demonstrated Isaac GR00T-based manipulation and manufacturing tasks on LG’s CLOiD robot in a real factory, using a data loop of real-world collection, Cosmos-based augmentation, GR00T training, Jetson AGX Thor onboard compute and simulation-based evaluation. Source