Search
Articles · 5
Hesai expands U.S. lidar sales despite Pentagon blacklist
Hesai Technology is expanding U.S. lidar sales and partnering with Nvidia despite a 2024 Pentagon blacklist labeling it a Chinese military entity. Security experts warn the sensors could introduce cyber risks and data-access backdoors into autonomous systems and critical infrastructure.
Anthropic Claude Models Now Run on NVIDIA GB300 in Azure
Anthropic’s Claude models in Microsoft Foundry are now generally available on Microsoft Azure running on NVIDIA GB300 Blackwell Ultra GPUs. This gives Azure-native enterprises a new platform for building autonomous and domain-specific AI agents with enhanced performance and governed security.
Agility Robotics to go public via SPAC at $2.5B valuation
Agility Robotics is set to list publicly through a SPAC merger at a $2.5 billion valuation, according to executives who spoke to The Wall Street Journal. The move comes as major tech players like Nvidia intensify efforts in humanoid robotics for industrial use.
SpaceX inks major compute pact with open-source AI lab Reflection
SpaceX has signed a computing-power contract with Reflection AI granting the open-source startup access to Nvidia GB300 chips inside Colossus 2. The pact, valued at roughly $6.3 billion through 2029, underscores SpaceX’s commercialization of its supercluster while spotlighting rising demand for open models amid concerns over closed systems.
NVIDIA Debuts Accelerated Computing Libraries for Research at ISC
At ISC in Hamburg, NVIDIA unveiled DAQIRI, ALCHEMI NIM microservices and the upcoming cuPhoton reference code to accelerate scientific AI pipelines ranging from chemistry to dark matter searches. The libraries deliver a 14,900x speedup on LSST astronomical data loading and enable real-time analysis of more than 99 percent of CERN collision events that would otherwise be rejected.
From the feed · 48
Vera Rubin NVL72 production is ramping across 350-plus factory sites in 30 countries. • CoreWeave • Google Cloud • Microsoft Azure • Oracle Cloud Infrastructure
10x more tokens per second per megawatt on DeepSeek-R1. CoreWeave posted the first measured results for NVIDIA Vera Rubin NVL72, a 10x gain versus Blackwell.
Nvidia's Vera Rubin platform merges CPUs and GPUs into a single system. The move targets every layer of chips inside AI data centers.
Nvidia has released the technical details customers need to evaluate its upcoming Vera data center CPU for AI workloads. The move sets up a direct comparison with AMD and Intel offerings.
102.4 Tbps Spectrum-6 Ethernet switches are now in production at gigascale AI factories. • CoreWeave • Microsoft • Nebius • SpaceX • Tesla
Nvidia unveiled its Vera Rubin platform to optimize token emission in AI factories built around inference workloads. The design focuses on throughput for token-selling operations rather than raw training scale.
NVIDIA introduced new open models and real-time simulation capabilities at SIGGRAPH, extending its work in agentic and physical AI to graphics, media, and robotics.
NVIDIA Agent Toolkit now includes Omniverse libraries that equip AI agents with tools to add physical capabilities and prepare 3D content for simulation.
113-degree water is cooling NVIDIA data center hardware, warm enough for a hot tub yet still effective for the latest chips.
Bristol Myers Squibb is deploying its second NVIDIA DGX SuperPOD on Vera Rubin, naming the new cluster SuperDuperPOD to extend its existing life-sciences AI footprint.
HBM4 shipments tied to Nvidia's Vera Rubin are expected to scale through the second half of 2026. The contest among the three major memory makers now hinges on solving heat dissipation after hybrid bonding delays.
Bezos has backed CuspAI, which is now working with Nvidia to apply AI to the search for new chipmaking materials. The startup is one of several targeting physical-world problems with generative models.
Musk called reports of a $52 billion SpaceX order with Foxconn for Nvidia GB300 AI servers "fake news." The claim had circulated widely before his July 20 post on X.
Nvidia's quarterly revenue is approaching $100 billion, with CEO Jensen Huang stating that growth continues to accelerate as the company leans on AI infrastructure to navigate HBM constraints.
Just under $1 million was paid for a leather jacket worn by Nvidia CEO Jensen Huang at Sotheby's. The sale reflects collector interest in AI-era artifacts.
NVIDIA NeMo Automodel now integrates with Diffusers to support fine-tuning of video and image models at scale.
NVIDIA Vera Rubin delivers the lowest cost per token for post-training workloads via extreme codesign, maximizing intelligence per dollar for agentic AI systems.
Nvidia CEO Jensen Huang says the AI boom is still at the beginning of the cycle.
NVIDIA Nemotron 3 Embed took the top rank overall on RTEB. The model improves retrieval accuracy for agentic workflows.
Fireworks hit a $17.5 billion valuation with Nvidia backing as more companies shift to lower-cost AI models. The startup diversified revenue after earlier heavy reliance on coding tool Cursor.
Nvidia announced its Cosmos 3 Edge AI model and an expansion of its physical AI ecosystem in Japan.
Nvidia is partnering with robotics firms in Japan to advance AI development for industrial and automation uses.
Japan's government, industrial leaders and NVIDIA launched the world’s first national AI infrastructure.
NVIDIA and its partners in Japan are showcasing the latest full-stack AI and robotics deployments this week. The push targets manufacturing, infrastructure and robotics sectors.
Nvidia H200 exports to China are reaching a wider set of licensed buyers, though actual shipments remain minimal under continued scrutiny.
Nvidia is reportedly in talks with Mitsubishi Heavy Industries to supply cooling systems and energy equipment for the AI data centers it is building with partners.
A leaked internal Nvidia tool restores access to blocked GPU hotspot sensor data on RTX 50-series cards. The sensors had been disabled in public drivers.
Lenovo listed its Legion 7a with an RTX 5070 12GB at more than $3,370, nearly three times Nvidia's target price for the GPU.
Nvidia's head of automotive, Xinzhou Wu, says he competes internally with other divisions for GPU compute time. The demand from AI workloads has created shortages even for the company's own car projects.
Jensen Huang will travel to Japan for an event marking Nvidia's three-decade partnership with Sega, revisiting the tie that helped the chipmaker through its early years as AI PC competition intensifies.
NVIDIA launched a set of GeForce trading cards that highlight major moments in PC gaming history.
NVIDIA teams are using ChatGPT Work to automate workflows and cut operating costs. GPT-5.6 trains and runs on the company's GB200 and GB300 NVL72 systems.
Chinese companies are reportedly shifting more AI accelerator spending from Nvidia to domestic suppliers.
Nvidia appears to have quietly discontinued the 2019 Shield TV, though one model remains listed.
SambaNova paired Nvidia H200 GPUs with its SN50 RDUs to reach 763 tokens per second on MiniMax M2.7 in third-party benchmarks.
China plans to let its top AI firms buy a limited number of Nvidia H200 chips, The Information reports.
NVIDIA Nemotron 3 Ultra hit the highest accuracy among open models on LangChain's Deep Agents harness, completing more tasks at higher throughput and lower cost than top closed models.
SambaNova reached an $11 billion valuation after fresh funding led by General Atlantic. CNBC reports the round as investors back AI chip startups challenging Nvidia.
NVIDIA introduced Vera, a CPU built for maximum single-threaded performance at scale. The design targets sequential steps in agentic AI systems that can otherwise stall and leave GPUs idle.
Alan Wei Zhaolun, an Aperia Group executive, pleaded not guilty in Singapore to fraud and money laundering charges over alleged smuggling of Nvidia chips to China, per DigiTimes Asia.
Meta is exploring an AI cloud business that would let enterprises tap its models and spare GPU capacity, DigiTimes Asia reports. The plan targets better hardware returns from Meta's AI spend, with Nvidia and TSMC showing no broad slowdown.
Hon Hai beat sales expectations on continued AI demand as a key Nvidia supplier, Fortune reports.
Nvidia floated a datacenter financing scheme that would let it collect payments twice on the same hardware, per The Register.
Singapore authorities filed additional fraud and money laundering charges against four individuals and four companies in a case tied to servers that may contain Nvidia AI chips subject to US export controls.
Verkada is working with NVIDIA to accelerate physical AI across its security camera platform.
Cadence connected its Reality Digital Twin Platform to NVIDIA Omniverse. The link targets faster data center design and optimization for AI workloads.
NVIDIA released its BioNeMo Agent Toolkit, adding GPU-accelerated AI tools and microservices for life sciences workflows inside Anthropic's new Claude Science workbench.