Nvidia is a US chip company that designs GPUs and AI computing platforms, and its CUDA software stack is widely used to train and run AI models. Recent coverage has centered on its stock and valuation: Apple overtook Nvidia as the world's most valuable company on Oct. 30, and Nvidia shares fell as much as 5.3% that day, with FQ2 revenue of $96 billion, up 106% year over year. Nvidia also plans a $1 billion investment in Naver for a 4.54% stake.
inDrive to test Pony.ai robotaxis in Kazakhstan with Nvidia in 2027inDrive signed an MOU with Pony.ai and Nvidia to test Pony.ai's Gen-7 robotaxis in Astana, Kazakhstan, in the first half of 2027, starting with a safety operator and aiming for fully driverless rides. Pony.ai supplies the vehicles, Nvidia provides its Hyperion platform and Halos safety system, and inDrive operates them via its app in a capital-light model.
SpaceX and Tesla Form Special Committees to Discuss MergerSpaceX and Tesla have formed special committees to discuss a merger. SpaceX's credit risk is high as it is about to borrow another $40 billion to spend on Nvidia chips. The merger's appeal depends on belief in Musk's Mars vision.
Nvidia pushes standalone CPU as AI agents boost processor demandAs AI shifts from answering questions to completing tasks, CPU demand is rising. Nvidia is rolling out its Vera Rubin platform with its first standalone CPU, claiming up to 1.5x faster performance per core than x86-based CPUs. AMD expects the server CPU market to exceed $120 billion by 2030, growing over 35% annually.
Nvidia open-sources OpenShell runtime for autonomous AI agentsNvidia published OpenShell on GitHub as a safe, private runtime for autonomous AI agents. The repository has 1,847 commits and drew discussion on Hacker News.
SpaceX works with Apollo, Nvidia on Grok partnership programSpaceX is working with Apollo Global Management, Nvidia and dozens of other firms on a partnership program to deploy and improve its Grok AI systems, aiming to boost business adoption of its AI offerings.
NVIDIA unveils Minerva for end-to-end AI factory observabilityNVIDIA launched Minerva, an AIOps platform that combines telemetry from infrastructure and AI workloads and applies purpose-built AI models to turn it into actionable insight. It builds a dynamic knowledge graph mapping GPUs, SuperNICs, switches and jobs, analyzes billions of data points per hour, predicts failures, groups event storms into a single root cause, and includes a native AI assistant.
SCMP: Huawei, DeepSeek narrow chip ecosystem gap with NvidiaThe South China Morning Post reports that Huawei and DeepSeek are narrowing the AI chip ecosystem gap with Nvidia, framing the trend as 'Jensen's nightmare.' The excerpt provides no specific performance benchmarks or market figures.
Jensen Huang urges Europe to get serious on AI raceNvidia CEO Jensen Huang scolded Europe over its approach to the AI race, urging the bloc to get serious and leverage its strengths to compete. The remarks highlight US chip leaders' attention to Europe's AI policy and industrial progress.
Google TPUs run in Meta data centers in first external salesGoogle's TPUs have been sold to an external customer for the first time and now run AI workloads in Meta's data centers. The move marks Google's custom chips entering the merchant market and could challenge Nvidia's dominance in AI accelerators.
Nvidia's Huang Calls Neoclouds a 'Good Investment'Nvidia CEO Jensen Huang said in a Bloomberg interview that neoclouds are a "good investment," without disclosing specific figures or deal details. The remark signals Nvidia's support for GPU-focused third-party compute providers.
Nvidia portfolio 80% concentrated in two AI stocksAs of Q2, 80% of Nvidia's investment portfolio was concentrated in two AI-related stocks: SpaceX at 33% and Intel at 47%. SpaceX's Q3 revenue grew 161% year-over-year to $14.6 billion, with AI business revenue up over 800%. The company has said it would rent data center capacity to Anthropic, Google, and others.
European Financial Firms Adopt Open Models for Proprietary IntelligenceAt GTC Berlin, European financial institutions including Revolut, Starling Bank, Lloyds and bunq showcased proprietary intelligence built on NVIDIA accelerated computing and open weight models. Revolut is rebuilding its stack on foundation model PRAGMA, serving an initial 80 million customers and engineered to scale 3x; Starling Bank trained an agentic model on U.K. sovereign cloud with its own transaction data.
NVIDIA Cosmos Expands Tools to Train and Deploy World ModelsNVIDIA expanded its Cosmos platform with new training tools, simulation checkpoints and Cosmos Data Factory blueprints for building physical AI world models. Ranking top five on six open model leaderboards, Cosmos now offers five cloud-native blueprints via GitHub, Azure, Nebius and CoreWeave, with companies already using them to speed physical AI development.
NVIDIA Powers European Retail Digital TransformationEuropean retailers are using open models and physical intelligence across commerce and fulfillment. Shopify post-trained NVIDIA Nemotron 3.5 Lightning, surpassing a leading frontier model on its internal benchmark while cutting serving costs 90% and delivering 7x faster responses; Zooplus rebuilt product discovery with NVIDIA technology, reaching 86% search accuracy in a prototype.
NVIDIA Earth-2 Adds Open Weather and Climate ModelsNVIDIA's Earth-2 platform added StormScope EU, a regional model predicting European cloud conditions every 10 minutes using EUMETSAT satellite data, demonstrated live at GTC Berlin. Organizations including AXA are building on Earth-2, and developers can customize models via the open source PhysicsNeMo framework for regional needs.
NVIDIA launches Kumo Relational foundation modelNVIDIA Kumo Relational is a pre-trained relational foundation model that predicts directly on enterprise databases without task-specific training or feature engineering. Across 41 tasks on 15 databases it beats supervised relational deep learning and expert-tuned gradient-boosted trees, and on SAP SALT it tops AutoGluon ensembles by 8% in mean reciprocal rank. It scales to 500B+ rows and is available via an open-source library or NVIDIA NIM.
User runs 748B DeepSeek-V4.1-Flash on a €2k PCA user ran the 748B-parameter DeepSeek-V4.1-Flash-1017 at 7.2 tok/s on a used ~€2,000 PC (RTX 3090, 128GB DDR4, i9-9900X) using a 1-bit quant and llama.cpp. A 110GB Engram table stays on NVMe and is read on demand, letting it fit. The output contained a physics error and has little practical use, the user said; Qwen models work better on the setup.
Train Microduck to walk in Isaac Lab 3.0 and deploy to real robotPollen Robotics and Hugging Face released Microduck, an affordable open-source bipedal robot, with a tutorial for training walking policies in GPU-parallel simulation using NVIDIA Isaac Lab and Newton. Actuator modeling and domain randomization prepare policies for real hardware; policies export to ONNX for deployment, can be shared on the Hugging Face Hub, and can compete in Microduck Arena.
Community debates overlooked RTX 3090 performance potentialA LocalLLaMA user says an earlier viral warning about the RTX 3090 now makes sense and asks whether others can confirm it. The poster speculates that restrictions originally aimed at mining demand may have been lifted and could benefit LLM workloads. No data is presented.
Demand for compute is extraordinary. NVIDIA and our partners are building new capacity to meet it. We’re working with Amkor, IBIDEN, KYEC, Samsung Mechanics, SPIL, and Unimicron to expand our supply around the world, from essential components to advanced pack…
Mistral and Nvidia team up on physics AI for industrial designMistral and Nvidia are combining Mistral Large 4 and Mistral's physics AI with CUDA-X, Warp and Omniverse to link design, simulation and AI training for aerospace, automotive and semiconductor industries; they also plan public benchmarks and open datasets. The roughly 1-trillion-parameter Mistral Large 4 was trained on 3,800 Grace Blackwell GPUs in Europe and ships as open weights in late October.
Nvidia DSX POD Cuts AI Data Center Setup to 14 DaysAt GTC Berlin, Nvidia unveiled DSX POD, a modular data center reference design based on Vera Rubin systems, with partners like Dell, HPE, Lenovo, and Schneider Electric pre-installing and testing units at the factory. It cuts on-site setup from six months to as little as 14 days. First units ship in early 2027, and closed-loop cooling can use up to 100% less water.
NVIDIA releases open-source Kumo-Forecast time-series modelNVIDIA released Kumo-Forecast, a multivariate time-series forecasting model built on the MOMENT-1-large backbone with about a 340M-parameter encoder. It adds a DARR mode that retrieves historical patterns and offers interpretability. Code and weights are open-sourced under Apache 2.0.
Cassava Deploys NVIDIA Nemotron on Its CAIMEx PlatformCassava Technologies is adding NVIDIA Nemotron open models to its AI Factory, offering them as locally hosted managed inference APIs via its Multi-Model Exchange (CAIMEx). The move lets enterprises build agentic applications without operating their own model-serving infrastructure.
NVIDIA showcases local AI hardware lineup at GTC BerlinAt GTC Berlin, NVIDIA showed its local AI hardware lineup: RTX Spark for Windows PCs, DGX Spark for Linux developers and DGX Station for enterprises. DGX Spark 64GB ships Oct. 23 and a Windows DGX Station arrives in December. NVIDIA Sync will support eight DGX Sparks, with day-0 support for new Nemotron, GLM, DeepSeek, Qwen and other models.
Google Cloud, NVIDIA Vera Rubin infrastructure now availableGoogle Cloud and NVIDIA announced general availability of A5X instances, opening next month with planned expansion to Europe. A5X combines Vera Rubin NVL72, ConnectX-9 SuperNICs and liquid cooling for large-scale agentic inference and reinforcement learning. Google Cloud plans up to 960,000 Rubin GPUs across multiple sites.
1 petaflop.
NVIDIA cloud partners expand sovereign AI capacity in EuropeNVIDIA cloud partners are expanding regional compute across Europe, the Nordics, Africa and Central Asia for training and inference. Nscale used NVIDIA DSX MaxLPS to fit 37% more GPUs and lift aggregate inference throughput nearly 50% within the same power budget, and plans Vera Rubin deployments across five European markets.
European manufacturers adopt NVIDIA for engineering, factoriesEuropean industrial companies are adopting NVIDIA CUDA-X, Omniverse libraries and open Cosmos and Nemotron models. The Factory Operations Blueprint (FOX) is in early access on GitHub for factory-manager agents. Siemens is integrating FOX-built agents into Intelligence Center X, estimating 20% cost savings and a 40% cut in inspection calibration time.
NVIDIA cuDSS reaches general availability for simulationsNVIDIA's cuDSS CUDA-X library is now generally available, moving large sparse solves in engineering and scientific simulations onto GPUs. Applied Materials reports up to 10x faster sparse solves, Dassault Systèmes cites 8-10x speedups in Abaqus, and ZEISS SMT measured a 4x gain in an EUV lithography simulation benchmark.
The A.I. Daily