AI news for: Hardware
Explore AI news and udpates focusing on Hardware for the last 7 days.
Alibaba to offer Nvidia’s physical AI development tools in its AI platform
Alibaba said on Wednesday that it is integrating Nvidia's AI development tools for robotics, self-driving cars and connected spaces into its Cloud Pla...
Just Released: NVIDIA HPC SDK v25.9
The new release introduces support for CUDA 13.0, and includes updated library components, bug fixes, and performance improvements....
TL;DR
NVIDIA HPC SDK 25.9 has been released with improvements and new features.
Key Takeaways:
Key Takeaways:
- CUDA version 13.0 is bundled, alongside CUDA 12.9 in some packages.
- Compatible with various Linux distributions, including RHEL/Rocky, SLES/SUSE, and Ubuntu.
- Installation instructions are provided for various Linux platforms and package managers.
Nvidia is letting anyone use its AI voice animation tech
Nvidia is open-sourcing Audio2Face, its AI-powered tool that generates realistic facial animations for 3D avatars — all based on audio input. The chan...
TL;DR
Nvidia is open-sourcing Audio2Face, its AI-powered tool that generates realistic facial animations for 3D avatars based on audio input.
Key Takeaways:
Key Takeaways:
- Developers can now use the Audio2Face tool and its underlying framework to create realistic 3D characters for their games and apps.
- The tool allows developers to create 3D characters for both pre-scripted content and livestreams.
- The training frame of Audio2Face is being made available, enabling users to tweak its models for different use cases.
Related:
NVIDIA Open Sources Audio2Face Animation Model
developer.nvidia.com
•
Sep 24, 2025
Open Secret: How NVIDIA Nemotron Models, Datasets and Techniques Fuel AI Development
Open technologies — made available to developers and businesses to adopt, modify and innovate with — have been part of every major technology shift, f...
TL;DR
NVIDIA opens its AI technology, Nemotron, to the public, enabling developers and enterprises to access and build upon its multimodal models, datasets, and techniques for AI development.
Key Takeaways:
Key Takeaways:
- Nemotron provides a transparent and adaptable foundation for building AI applications, enabling developers to understand how their models work and trust the results.
- The technology supports both generalized intelligence and specialized intelligence, making it suitable for a wide range of AI use cases across industries.
- Nemotron has been adopted by various companies, including CrowdStrike, DataRobot, and ServiceNow, for their AI development needs, and its dataset is also used to inform the design of NVIDIA's future AI systems.
Deploy High-Performance AI Models in Windows Applications on NVIDIA RTX AI PCs
Today, Microsoft is making Windows ML available to developers. Windows ML enables C#, C++ and Python developers to optimally run AI models locally acr...
TL;DR
Microsoft is making Windows ML available to developers, enabling AI model inference on PC hardware from CPU, NPU, and GPUs with optimized performance on NVIDIA RTX GPUs using TensorRT for RTX Execution Provider.
Key Takeaways:
Key Takeaways:
- Windows ML unlocks full TensorRT acceleration for GeForce RTX and RTX Pro GPUs, delivering exceptional AI performance on Windows 11.
- TensorRT for RTX Execution Provider provides 50% faster throughput compared to prior DirectML implementations on NVIDIA RTX GPUs and supports low-latency inference.
- Windows ML developers can leverage architecture advancements like FP8 and FP4 on the Tensor Cores and support a variety of model architectures, including LLMs, diffusion, CNN, and more with the new ONNX Runtime APIs.
Related:
Microsoft’s Windows AI Lab is a new way to test experimental features
www.theverge.com
•
Sep 23, 2025
Windows ML now available, empowering developers to scale local AI across Windows devices
blogs.windows.com
•
Sep 24, 2025
Microsoft opens the doors to more AI-powered Windows apps
www.theverge.com
•
Sep 25, 2025
Build a Retrieval-Augmented Generation (RAG) Agent with NVIDIA Nemotron
Unlike traditional LLM-based systems that are limited by their training data, retrieval-augmented generation (RAG) improves text generation by incorpo...
TL;DR
NVIDIA has introduced Agentic RAG, a new approach for text generation that combines LLMs with autonomous systems and external knowledge, enhancing their capabilities and versatility.
Key Takeaways:
Key Takeaways:
- Agentic RAG allows LLMs to access external knowledge, adapt to changing information, and perform complex reasoning tasks dynamically.
- The new approach can be built using NVIDIA NIM endpoints and tools like LangGraph, enabling developers to create advanced AI agents.
- Agentic RAG has applications in various domains, including customer support, IT Help Desks, and more, where agents can retrieve and process relevant information to provide accurate responses.
Nvidia Invests in OpenAI With $100 Billion to Build Out More AI Data Centers - CNET
Nvidia Invests in OpenAI With $100 Billion to Build Out More AI Data Centers CNETNvidia stock jumps on $100 billion OpenAI investment as Huang touts '...
TL;DR
Nvidia and OpenAI have partnered to invest up to $100 billion in OpenAI and supply Nvidia systems for the company's next-generation models, with the goal of securing massive computing resources and guaranteeing pace with AI rivals.
Key Takeaways:
Key Takeaways:
- OpenAI will deploy at least 10 gigawatts of Nvidia systems for AI data centers over the coming years.
- Nvidia's investment will grow in scale as each new system is deployed, with the first phase scheduled for 2026.
- This partnership positions Nvidia at the center of the AI boom, buying into the biggest AI company, and could define how quickly AI advances in the future.
Related:
Nvidia is partnering up with OpenAI to offer compute and cash
www.theverge.com
•
Sep 22, 2025
NVIDIA, OpenAI Announce ‘the Biggest AI Infrastructure Deployment in History’
blogs.nvidia.com
•
Sep 22, 2025
Nvidia To Invest Up To $100B In OpenAI
news.crunchbase.com
•
Sep 22, 2025
Nvidia (NVDA) Seen as Key AI Beneficiary Despite Risks Tied to OpenAI Funding - Yahoo Finance
consent.yahoo.com
•
20h ago
China’s ‘full stack’ AI ambitions on show at Alibaba’s Apsara Conference - South China Morning Post
China’s ‘full stack’ AI ambitions on show at Alibaba’s Apsara Conference South China Morning PostAlibaba Stock’s AI-Powered Run Isn’t Done Barron'sAli...
Related:
Alibaba's shares jump after it lands a major customer for its AI chips - CNBC
www.cnbc.com
•
Sep 17, 2025
Once an e-commerce giant, Alibaba bets future on billion-dollar AI overhaul - CNBC
www.cnbc.com
•
Sep 20, 2025
Alibaba shares jump 9% in U.S. premarket after CEO unveils plans to boost AI spending - CNBC
www.cnbc.com
•
Sep 24, 2025
R²D²: Three Neural Breakthroughs Transforming Robot Learning from NVIDIA Research
While today's robots excel in controlled settings, they still struggle with the unpredictability, dexterity, and nuanced interactions required for rea...
TL;DR
NVIDIA Research introduces groundbreaking neural innovations that transform robot learning and adaptation, enabling more accurate dynamics prediction, dexterous manipulation, and bimanual assembly.
Key Takeaways:
Key Takeaways:
- NeRD (Neural Robot Dynamics) enables accurate dynamics prediction, with less than 0.1% error in accumulated reward for 1,000-step policy evaluation.
- Reference-Scoped Exploration (RSE) streamlines learning dexterous manipulation from human demonstrations, achieving almost 20% more success rates with the Inspire hand.
- VT-Refine combines vision and touch for robust bimanual assembly, improving real-world success rates by approximately 20% in the vision-only variant and 40% for the visuo-tactile variant.
Faster Training Throughput in FP8 Precision with NVIDIA NeMo
In previous posts on FP8 training, we explored the fundamentals of FP8 precision and took a deep dive into the various scaling recipes for practical l...
TL;DR
FP8 training with NVIDIA H100 and DGX B200 GPUs delivers significant acceleration and convergence benefits over BF16, particularly for large-scale language models.
Key Takeaways:
Key Takeaways:
- FP8 training yields a speedup of up to 1.53x for large language models (Llama 3.1 405B) compared to BF16 on NVIDIA H100 GPUs.
- MXFP8 recipe on DGX B200 GPUs achieves a consistent speedup of 1.28x to 1.37x across different model sizes, with a peak speedup of 1.37x.
- Larger models experience more pronounced efficiency gains with FP8, as the reduced memory footprint and higher throughput benefit more significantly from reduced-precision arithmetic.
How to Accelerate Community Detection in Python Using GPU-Powered Leiden
Community detection algorithms play an important role in understanding data by identifying hidden groups of related entities in networks. Social netwo...
TL;DR
Leiden from cuGraph delivers results up to 47x faster than comparable CPU alternatives for community detection algorithms in Python.
Key Takeaways:
Key Takeaways:
- cuGraph's GPU-accelerated Leiden implementation outperforms comparable CPU implementations by up to 47x.
- Leiden from cuGraph is easily accessible in Python workflows through the cuGraph library or NetworkX library through the nx-cugraph backend.
- GPU-powered Leiden workflows can scale large data and solve bigger problems in far less time, making it ideal for genomics and other applications relying on community detection.
Build a Real-Time Visual Inspection Pipeline with NVIDIA TAO 6 and NVIDIA DeepStream 8
Building a robust visual inspection pipeline for defect detection and quality control is not easy. Manufacturers and developers often face challenges ...
TL;DR
NVIDIA TAO and DeepStream 8 empower developers to build real-time visual inspection pipelines using knowledge distillation, domain adaptation, and self-supervised fine-tuning for improved accuracy and performance.
Key Takeaways:
Key Takeaways:
- NVIDIA TAO enables the customization of vision foundation models using domain adaptation, self-supervised fine-tuning, and knowledge distillation.
- DeepStream 8 provides a low-code tool for deploying model ideas into standalone applications or microservices with improved inference accuracy and throughput.
- Together, TAO and DeepStream 8 accelerate model development, improve accuracy, and enable real-time visual inspection pipelines for various industries such as automotive, logistics, and manufacturing.
Nvidia and Abu Dhabi institute launch joint AI and robotics lab in the UAE - Reuters
Nvidia and Abu Dhabi institute launch joint AI and robotics lab in the UAE ReutersNvidia, Abu Dhabi Institute Open AI and Robotics Lab in UAE Yahoo Fi...
TL;DR
Nvidia and Abu Dhabi's TII launch joint AI and robotics lab in the UAE to explore AI applications and robotic innovations.
Key Takeaways:
Key Takeaways:
- The joint lab will explore AI applications in areas such as climate, energy, and genomics.
- The TII will use Nvidia's Thor chip to advance robotics research and development.
- Specific use cases are planned for robotic innovations, including transportation and logistics.
How to GPU-Accelerate Model Training with CUDA-X Data Science
In previous posts on AI in manufacturing and operations, we covered the unique data challenges in the supply chain and how smart feature engineering c...
TL;DR
GPU-accelerated tree-based models offer a balance of accuracy, speed, and interpretability for manufacturing data science, leveraging libraries like NVIDIA cuML to supercharge experimentation and deployment.
Key Takeaways:
Key Takeaways:
- Tree-based models are ideal for structured manufacturing data, outperforming neural networks in this domain.
- Popular library XGBoost can be 150x faster in inference using the Forest Inference Library from NVIDIA cuML, making it suitable for production use.
- Feature importance analysis helps engineers understand variable relationships and identify noise features, offering a robust way to filter out uninformative data.
Canada Goes All In on AI: NVIDIA Joins Nations’ Technology Leaders in Montreal to Shape Sovereign AI Strategy
Canada’s role as a leader in artificial intelligence was on full display at this week’s All In Canada AI Ecosystem event. NVIDIA Vice President of Gen...
TL;DR
Canada is accelerating its AI momentum with the launch of its first fully sovereign AI factory, powered by NVIDIA's accelerated computing, and the adoption of a national AI strategy aiming to build digital sovereignty.
Key Takeaways:
Key Takeaways:
- Canada's AI factory, built by TELUS and powered by NVIDIA, offers end-to-end AI capabilities and ensures full data residency and control within Canadian borders.
- The country's AI momentum is accelerated through collaborations between public and private sector leaders, shaping national infrastructure, innovation, and policy for AI development.
- RBC Capital Markets is using NVIDIA software to build enterprise-grade AI agents for capital markets research, enabling Canadian institutions to deploy intelligent systems tailored to local needs.
At Climate Week NYC, NVIDIA Details AI’s Key Role in the Sustainable Energy Transition
Energy efficiency in large language model inference has improved 100,000x in the past 10 years — demonstrating that accelerated computing is sustainab...
TL;DR
NVIDIA highlights AI's key role in the sustainable energy transition, with accelerated computing improving energy efficiency by 100,000x in the past 10 years.
Key Takeaways:
Key Takeaways:
- Forecasted AI-induced energy savings across energy-intensive sectors by 2035: 4.5% of projected energy demand will be saved.
- Emerald AI and NVIDIA collaborate on a reference design for grid-friendly, energy-efficient AI infrastructure, unlocking 100 gigawatts of untapped power grid capacity.
- NVIDIA's product carbon footprint summary comparison reveals a 24% reduction in embodied carbon emissions intensity between NVIDIA HGX H100 and HGX B200 baseboards.
Video Updates
AI News: ChatGPT Pulse, Gemini Robotics, Qwen3-Max, Stargate, OpenAI and Nvidia, and more!
Matthew Berman
•
Sep 26, 2025
28
Sep
27
Sep
26
Sep
25
Sep
24
Sep
23
Sep
22
Sep