The latest AI news and updates of NVIDIA — AI computing company building GPUs, inference hardware, and developer platforms for training and deployment. Covering NVIDIA's latest company news, launches, and product updates from the past 90 days.
NVIDIA5h agoNVIDIA and Meta are expanding their partnership on Meta's Muse model family to improve inference and training performance. The companies are exploring NVIDIA OpenShell for safe agent deployment and NVIDIA Sentry for behavior monitoring. Additionally, Meta plans broader deployments of NVIDIA Vera Rubin and Vera CPUs, while NVIDIA will invest $3 billion in Meta's equity offering.
NVIDIAOct 24NVIDIA expanded the Cosmos platform with new tools for building world models. Cosmos Data Factory adds five cloud-native blueprints for synthetic data generation, while the Cosmos Framework extends training and optimization skills to models like GR00T and QWEN. Additionally, new Cosmos 3 Nano variants enable multiview simulation testing, and updated NIM microservices deliver up to 2x faster inference.
NVIDIAOct 23NVIDIA released Nemotron 3.5 Bolt E4B 4GB, a small open model optimized for conversational agents and tool use. The 4GB model supports 64k context and 35 languages, enabling responsive local experiences on edge hardware like the Jetson Orin Nano. It is available now on Hugging Face with full weights, training data, and recipes for customization.
NVIDIAOct 21NVIDIA announced plans with six supply chain partners to expand global manufacturing capacity for its AI infrastructure. The agreements include a 2.5x increase in substrate capacity over three years, a 300% boost in advanced packaging by 2028, and a $1.1 billion investment in U.S. testing facilities. Optical engine capacity increases 15x by mid-2027 to support large-scale computing.
NVIDIAOct 21NVIDIA expanded its Isaac robotics platform with new tools including Isaac OS, Isaac Lab 3.0, and simulation-ready assets. Alongside this, Agile Robots introduced the Agile ONE SN, a reference humanoid powered by NVIDIA Jetson Thor and Isaac GR00T. The humanoid features 31 degrees of freedom and 2,070 FP4 teraflops of onboard compute, with availability scheduled for early 2027.
NVIDIAOct 21NVIDIA added the Agent Knowledge Foundation to its Agent Toolkit at GTC Berlin, providing agents with tools to find relevant information, make predictions, and guide decision-making. The foundation uses CUDA-X libraries and Auto-Ontology to map enterprise data into business concepts. Paired with the OpenShell secure runtime, the update enforces boundaries on agent access and actions.
NVIDIAOct 21At GTC Berlin, NVIDIA announced that European mobility leaders are showcasing autonomous use cases on its Hyperion platform. The company also introduced Halos SAFE, a framework for generating traceable AV safety evidence, and Alpamayo 2 Edge, a 2-billion-parameter reasoning model designed for in-vehicle scene understanding and motion planning at 10Hz.
NVIDIAOct 21NVIDIA and the UW Medicine Institute for Protein Design introduced RFD4-Proteina, an open model for protein design and candidate refinement. Announced at GTC Berlin, the model generates targeted protein designs for medicine and diagnostic applications. It is available through partners including Amazon Bio Discovery, Anthropic’s Claude Science, Benchling, Databricks, Google Cloud, Hugging Face, Nebius, OpenAI, and Tamarind Bio.
NVIDIAOct 21NVIDIA is turning telecom networks into distributed intelligence platforms by adding accelerated computing to data centers and cell sites. T-Mobile completed the first commercial AI-RAN call, while nine EMEA operators are building NVIDIA-powered infrastructure. NVIDIA also introduced two new Aerial RAN Computer configurations and integrated Nemotron open models into partner software to support autonomous network operations.
NVIDIAOct 21NVIDIA released a new portfolio of open models at GTC Berlin, including Nemotron, Kumo, Cosmos, Alpamayo, BioNeMo, and Ising. These models provide a foundation for European teams to customize applications with proprietary data on controlled infrastructure. The release supports specialized AI agents, robotics, protein design, and quantum computing development across the region.
NVIDIAOct 21NVIDIA introduced DSX POD, a reference design for Vera Rubin AI factories built by partners inside modular containers. By moving integration and validation to the manufacturing facility, the design reduces deployment time at a prepared site from six months to as little as 14 days. Partners will deliver these containerized AI factories worldwide starting in early 2027.
NVIDIAOct 21NVIDIA announced that European AI infrastructure leaders are adopting its DSX AI factory platform to improve efficiency. The platform uses closed-loop liquid cooling to reduce cooling water consumption by up to 100% and DSX MaxLPS software to generate the same AI output with up to 30% less power. DSX Flex also coordinates power demand to support grid reliability.
NVIDIAOct 21NVIDIA announced at GTC Berlin an expanded sovereign AI framework for European public service media, defining requirements for data governance and model security. Simultaneously, Adobe integrated NVIDIA GPU acceleration into Frame.io, delivering up to 5x faster embedding generation and 4x higher throughput for semantic video search, helping producers locate specific scenes within years of footage.
NVIDIAOct 20NVIDIA open-sourced Flora, a suite of models that converts raw video into timestamped captions, dense progress rewards, and rubric scores for robot learning. Flora labels 200,000 hours of video daily on 512 GB300 GPUs, achieving performance competitive with GPT-6 Astra at approximately 125x lower cost. The model weights are available now on Hugging Face.
NVIDIAOct 16NVIDIA open-sourced V1.0 of ASPIRE, an agentic robotics framework that autonomously discovers and refines robot control programs. The system achieved ~99% benchmark saturation after switching to the Astra backend. Additionally, Opus 5.5 slightly outperforms Astra on simulation benchmarks, while Sonnet 5.5 reaches a cost-efficiency Pareto frontier of $12 per task.
NVIDIAOct 15NVIDIA released Nemotron 3.5 Super VL, an open 120B-parameter model combining reasoning and tool use with text, image, document, and video understanding. The model activates 12B parameters per token to optimize inference for long-running agents. It is available on Hugging Face with weights, training data, and recipes, delivering 4x faster task completion on coding benchmarks.
NVIDIAOct 15NVIDIA announced planned investments in five UK AI startups: Boltz, Foresight, Hiverge, Prima Mente, and Riverlane. CEO Jensen Huang presented the move in London, framing the funding as part of an ongoing effort to support the UK’s AI ecosystem. The startups focus on biomolecular modeling, infrastructure risk prediction, algorithmic discovery, biological foundation models, and quantum error correction.
NVIDIAOct 14NVIDIA is donating $250 million in computing access over five years to support UK research on AI agent safety and security. The program provides $50 million annually through UK-based cloud partners to institutions including BOLD, SOFAIR, and Imperial College London. Researchers will use NVIDIA’s OpenShell and Sentry tools to develop safeguards for autonomous agents and secure AI architectures.
NVIDIAOct 12NVIDIA and Cisco introduced the Ultra-Dense Small Form Factor Pluggable (USFP) module, the first to support 400Gb/s per lane. This form factor delivers 4x the bandwidth density of current optics, allowing AI factories to connect more compute to existing switches without adding network tiers. The design also reduces power consumption per bit moved, with plans for an open industry specification.
NVIDIAOct 8NVIDIA researchers released PivotOPD, an on-policy distillation framework that trains multi-turn AI agents to identify and recover from early pivotal mistakes. The method uses teacher-provided gold and recovery actions to steer agents back on track. It achieved the strongest average performance across 13 baselines on ALFWorld, WebShop, and Search-based QA, while improving SWE-Bench Verified resolve rates by 3.2%.
NVIDIAOct 6NVIDIA reports that AT&T, SoftBank, and Indosat are adopting open models for autonomous network operations and local-language adaptation. NVIDIA provides the stack, including Nemotron for reasoning and OpenShell for agent control. Its 2026 survey shows 89% of telecom respondents prioritize open source models, with 66% actively using AI, up from 49% last year.
NVIDIAOct 2NVIDIA published a fine-tuning tutorial for its Nemotron 3.5 ASR model, demonstrating how to adapt it to Najdi and Hijazi Saudi Arabic dialects. The workflow reduced word error rates from 55% to 30% on target test sets. The recipe includes data curation, replay mixing, and encoder unfreezing, providing a path to adapt the model for other languages and dialects.
NVIDIASep 30NVIDIA released the Build Vision AI skill for VSS Blueprint 3.3, which builds and deploys visual AI agents from a single prompt. A bottling-line demonstration reached a live, previewable state in under 30 minutes, featuring automated alerts, video search, and incident reporting. The update also introduces Adaptive EVS, which reduces VLM processing costs by pruning redundant visual tokens.
NVIDIASep 30NVIDIA released Physis-Lang, an open self-evolving framework that adds physics reasoning to video captions. This reasoning improves world model training and generation, boosting NVIDIA Cosmos 3’s PhyGenBench score by 5.62 points without retraining. Physis-Lang with Cosmos 3 Super and Nano models currently hold the top two spots on the Physics-IQ Verified image-to-video leaderboard.
NVIDIASep 28NVIDIA introduced the Open Agent Safety Platform, an open reference design for governing AI agent behavior from testing to deployment. The platform includes OpenShell, a secure runtime for CPUs, and Sentry, an out-of-band watchdog on BlueField-4 DPUs that quarantines agents in milliseconds if they violate policies. OpenShell is now broadly available for integration across software and hardware systems.
NVIDIASep 23NVIDIA released Nemotron 3 Diarization, an open-weight 100M-parameter model that identifies speaker activity in multi-speaker conversations. The model supports up to eight speakers and handles overlapping speech. It ranked first among 12 systems in VoiceArena's initial Diarization-Bench evaluation with a 14.72% error rate, approximately 24% lower than the runner-up.
NVIDIASep 22Sluicebox deployed NVIDIA Nemotron 3 Ultra to power its AI supplier agent, Lucy, for automating product carbon footprint assessments. The model improved supplier-data extraction accuracy to 87.7% while cutting inference costs by 51–80%. These workflows enable engineers to evaluate environmental impacts during the design phase, delivering responses up to 2x faster than the company's previous production model.
NVIDIASep 19NVIDIA released AIPerf, a benchmarking tool for measuring LLM inference performance at scale. The tool uses a multiprocess architecture to prevent client-side bottlenecks while tracking metrics like TTFT, ITL, and throughput. It supports over 15 endpoint types and configurable traffic patterns, including Poisson and gamma distributions, to simulate realistic production workloads.
NVIDIASep 16NVIDIA Research introduced Axolotl3D at ECCV 2026, a multimodal and occlusion-aware 3D generation model. It conditions on images, camera data, and partial geometry to reconstruct missing object regions while preserving observed areas. The model achieves state-of-the-art performance in both single- and multi-view settings, enabling faithful 3D shape completion even under severe occlusion.
NVIDIASep 16University of Manchester researchers adapted NVIDIA Earth-2 generative AI to forecast UK air pollution at 2-3 square kilometer resolution. The model trained in two days on the Isambard-AI supercomputer and now runs inference locally on an NVIDIA DGX Spark desktop system. The team plans to open-source the workflows to enable localized pollution modeling globally.
NVIDIASep 16NVIDIA, Emerald AI, and Google launched the AI Energy Management Alliance to accelerate the interconnection of flexible, grid-enhancing data centers. The coalition establishes performance-based standards for data centers to dynamically manage electricity use by shifting workloads or reducing demand during grid stress. These measures aim to strengthen grid reliability, protect energy affordability, and unlock faster connections for AI infrastructure.
NVIDIASep 15NVIDIA expanded its AI for Media collection at IBC 2026 with new SDKs, NIM microservices, playbooks, and blueprints. These additions enable newsrooms to verify AI-generated footage, sports producers to generate smoother slow-motion replays, and broadcasters to perform lip-synced dubbing for multilingual content. The suite integrates directly into existing live production and streaming workflows.
NVIDIASep 14NVIDIA and Pinterest are expanding their collaboration to scale multimodal AI capabilities. The Pinterest Assistant now processes 25 times more visual context per request, enabling more personalized discovery. This update leverages NVIDIA’s accelerated computing platform to improve visual search and content understanding across the Pinterest ecosystem.
NVIDIASep 14NVIDIA optimized the Nemotron 3 Ultra NIM serving stack, tuning caching, memory, parallelism, and decoding. On a four-B200 GPU system, these improvements support up to 2.5 times more concurrent users while maintaining 50 tokens per second per user. The optimized configuration is available in NIM 2.0.12 for production agentic workloads.
NVIDIASep 10NVIDIA Research and USC released HorizonRelight, a method for relighting long videos consistently. By propagating target-domain context across sliding-window chunks, the system prevents visible lighting jumps and boundary artifacts common in chunked inference. The research, presented at ECCV 2026, includes a demo and paper, with code release coming soon.
NVIDIASep 10NVIDIA published guidance on using encode-prefill-decode (EPD) disaggregation in its Dynamo inference framework to separate vision encoding from LLM prefill and decode stages. This technique reduces resource contention and improves response times for image-heavy workloads, delivering up to 5x faster time to first token and 7x faster end-to-end latency in specific configurations.
NVIDIASep 10NVIDIA and Palantir are integrating NVIDIA Nemotron, cuOpt, and NeMo with Palantir Foundry and AIP to optimize supply chain operations. This AI stack, first deployed within NVIDIA’s own network, uses post-trained open models to flag emerging risks and model supply constraints. The system generates actionable recommendations and trade-off analysis while maintaining control over proprietary data.
NVIDIASep 10NVIDIA is collaborating with Australian data center providers, including Firmus, CDC, and AirTrunk, to expand land, power, and shell capacity for AI factories. Built on the NVIDIA DSX platform, these facilities provide high-performance compute for local researchers and enterprises. The partnership aims to deliver up to a 2-gigawatt buildout by 2027 to meet surging regional demand.
NVIDIASep 5NVIDIA’s fine-tuned Nemotron model scored 535.4 out of 600 on the IOI 2026 problem set, exceeding both the gold medal threshold and the top human contestant’s score. The model competed unofficially in Uzbekistan under identical constraints as human participants, including no internet access and strict time limits, marking the first AI system to outscore the highest-scoring human contestant.
NVIDIASep 4Jim Fan, co-creator of OpenAI's 2016 World of Bits project, notes that GPT-6 Astra finally achieves reliable computer use, a task that previously doomed RL-from-scratch approaches. He attributes this success to the Specialized Generalist recipe—broad pretraining followed by specialization to screen pixels—and the massive compute required to produce the emergent capability.
NVIDIASep 3NVIDIA entered a definitive agreement to acquire Hugging Face for $12.93 billion. The platform will remain an open, neutral, and independent home for the AI ecosystem, with the founders and team continuing their mission. NVIDIA plans to scale the platform’s infrastructure while maintaining support for multi-cloud, multi-accelerator development and open-weight models from all builders.
NVIDIASep 1NVIDIA and CrowdStrike evaluated an agentic cybersecurity system that connects offensive and defensive agents in a continuous loop. Using Nemotron 3 Ultra for orchestration and a fine-tuned Nemotron 3 Super for detection generation, the pipeline improved backtest detection rates to 41.9%. In live-fire testing, the system generated three detection rules that successfully caught eight previously unseen attacks.
NVIDIAAug 31NVIDIA and MediaTek are expanding their collaboration to build next-generation AI computing platforms, supported by a 3.5 billion dollar NVIDIA investment in MediaTek. MediaTek will adopt the NVIDIA NVLink Fusion platform to integrate custom XPUs into rack-scale AI factories. The companies are also continuing joint development of RTX Spark PC chips and AI-powered, software-defined vehicle platforms.
NVIDIAAug 27NVIDIA and AWS are expanding their partnership to deploy 2 million additional GPUs across AWS global infrastructure in 2027-2028. The collaboration adds NVIDIA Vera CPU-based infrastructure, extends NVLink Fusion with custom high-bandwidth memory, and builds AI factories for the U.S. government, including 100,000 GPUs for secure federal workloads. NVIDIA reports that demand is currently exceeding all previous forecasts.
NVIDIAAug 26NVIDIA introduced shadow engine recovery as a preview feature in its Dynamo inference framework. By maintaining a pre-warmed standby engine on the same GPU, the system eliminates the need for cold restarts during process failures. Tests with GLM-5.2 on B200 nodes restored serving capacity in 7.3 seconds, nearly 39 times faster than a standard cold restart.
NVIDIAAug 25NVIDIA has moved its Groq 3 LPX interactive AI inference accelerator into full production. Designed for the Vera Rubin NVL72 platform, the accelerator delivers 3,400 output tokens per second on Gemma 4 31B, providing 4x faster responsiveness for agentic workloads. Nebius is the first AI cloud to adopt the platform, with Groq also planning early adoption.
NVIDIAAug 25NVIDIA Nemotron 3.5 Lightning ranked in the top 4 open-weight models on the PinchBench leaderboard, achieving an 86.4% average success rate on OpenClaw agent tests. The 30B parameter mixture-of-experts model is optimized for high-volume execution. NVIDIA’s Nemotron 3 Ultra maintains the #1 spot on the same benchmark, highlighting the family's performance across agentic tasks.
NVIDIAAug 25NVIDIA announced that xAI is deploying Vera CPUs to accelerate its agentic AI workloads. The Vera CPU, featuring 88 Olympus cores and 1.2TB/s bandwidth, delivers up to 1.8x faster task completion than x86 alternatives. xAI is also expanding its Grok infrastructure on the Vera Rubin platform and plans to integrate optimized Vera Rubin NVL72 systems into its Starmind satellites.
NVIDIAAug 22NVIDIA and UC Berkeley open-sourced T-Rex, a tactile-reactive manipulation framework that gives robots a dedicated high-frequency control loop for touch. The release includes a 50-hour tactile-synchronized robot play dataset on HuggingFace and a three-stage training recipe. The architecture uses a Mixture-of-Transformer model to process tactile corrections at four times the frequency of visual planning.
NVIDIAAug 22NVIDIA’s AVO agent architecture achieved a 100.00 RHAE score on the ARC-AGI-3 benchmark, completing all 183 levels across 25 environments. The system, which uses persistent memory and supervision to sustain long-horizon tasks, solved the benchmark using 6,624 environment actions. This result demonstrates that system-level architecture can effectively convert model capability into sustained autonomous progress.
NVIDIAAug 22NVIDIA cuOpt is now the fastest open-source solver on Hans Mittelmann benchmarks across three optimization problem classes: linear programming, second-order cone programming, and quadratic programming. The GPU-accelerated engine provides near real-time solutions for large-scale problems, integrating into existing modeling languages for deployment across hybrid and multi-cloud environments.
NVIDIAAug 22NVIDIA benchmarked over 300 verified agent skills, finding they improved correctness by 41 points, effectiveness by 39, and efficiency by 35. The company open-sourced SkillEvaluator, a tool that measures skill impact through static checks and live task runs in isolated environments. This allows developers to test and verify skill performance before deployment.
NVIDIAAug 22NVIDIA released TensorRT Model Connect in public preview, an open-source tool that converts supported Hugging Face models into end-to-end TensorRT inference bundles in two commands. The tool eliminates intermediate ONNX export steps and provides bundles callable through native C++ APIs. NVIDIA built the project using OpenAI Codex agents, with human developers directing and reviewing the implementation.
NVIDIAAug 22NVIDIA guaranteed SB Energy’s PORTS-Pike Technology Campus in Ohio to exclusively host NVIDIA AI compute. The site will use NVIDIA’s DSX platform to provide 8 IT-GW of capacity for OpenAI. NVIDIA is investing $1.5 billion in SB Energy to support the project, which begins phased operations in 2028 and aims to create tens of thousands of local jobs.
NVIDIAAug 15NVIDIA launched NeMo Switchyard, an open-source library that dynamically routes AI agent tasks between models based on capability, cost, and latency. The system directs complex reasoning to frontier models while offloading high-volume execution to specialized models like Nemotron 3.5 Lightning. The routing layer is provider-agnostic and open-source — agent workloads stay optimized without rebuilding applications when switching or adding model providers.
NVIDIAAug 14NVIDIA and LG are expanding their collaboration to accelerate physical AI development. LG plans to unveil a bipedal humanoid robot built on NVIDIA Isaac GR00T and Jetson Thor in the first quarter of next year. The partnership also includes validating wheel-based robots in Tennessee, building AI factory reference sites with NVIDIA Vera Rubin, and developing AI-defined vehicle platforms.
NVIDIAAug 10NVIDIA signed memorandums of understanding with Apollo, BlackRock, Blackstone, Brookfield, Goldman Sachs, and KKR to establish independent compute financing platforms. These partnerships aim to mobilize over $500 billion of third-party capital to fund the global buildout of AI infrastructure. The platforms create dedicated capital pools to help customers access large-scale compute for AI factories.
NVIDIAAug 10NVIDIA announces support for Meta’s Muse Glimmer, a 30B open-weight dense model designed for long-running agentic tasks. The model runs locally on NVIDIA edge, desktop, and workstation platforms, delivering up to 20K tokens per second on a single GPU. Developers can deploy the model using NVIDIA NIM containers, SGLang, or vLLM inference recipes.
NVIDIAAug 8Firebird launched the CIS region’s largest AI factory in Armenia, integrating NVIDIA accelerated computing with Dell PowerEdge servers and Schneider Electric power infrastructure. The facility utilizes the NVIDIA DSX platform to support AI training and deployment. Firebird plans to deploy over 70,000 NVIDIA Rubin and Blackwell GPUs and 300 megawatts of capacity by the end of 2027.
NVIDIAAug 4NVIDIA is participating in the U.S. National Science Foundation’s new $100 million State and Regional AI Infrastructure Hubs program. The initiative expands access to advanced computing, data, software, and technical expertise for colleges and universities. These regional hubs aim to accelerate scientific research, strengthen AI education, and prepare students for the AI economy through shared resources and training.
Frequently asked questions