NVIDIA Blackwell stories
NVIDIA says US AI demand will add USD $485 billion to GDP in 2026 as it expands chip, systems and data centre manufacturing.
Demand for sovereign AI compute has forced the Gagarin site to expand within weeks, with capacity set to rise to 5MW by year-end.
Azure customers can now deploy Claude for governed enterprise agents as Microsoft Foundry widens access to Anthropic's models on NVIDIA GB300 hardware.
Customers should see faster AI search and training on AWS as NVIDIA makes GPU indexing the default and adds new EC2 G7 instances.
Security and governance tools are being added as enterprises push agentic AI from pilots into live production systems.
The new inference cloud is aimed at cutting latency and costs for enterprise AI, with a Los Angeles site live and Together.ai first to use it.
Local AI processing and gaming rigs took centre stage as Gigabyte unveiled new motherboards, graphics cards and laptops for Computex.
The compact desktop aims to cut cloud costs for AI developers by letting them fine-tune and run large models locally on Windows.
Windows PCs with up to 128GB of unified memory could let developers and creators run larger AI models locally, Microsoft said.
Creative professionals could run large AI and rendering workloads locally as ASUS adds laptops and a mini PC built on NVIDIA RTX Spark.
AI will only set firms apart if they can harness trusted proprietary data, Dell said as it unveiled new tools and partnerships.
Enterprises can now run more AI projects on their own infrastructure as Dell adds data tools, racks and partner software to its NVIDIA tie-up.
Local deployment could cut agentic AI costs by up to 87% over two years, Dell says, while keeping enterprise data on premises.
Governance and safety controls are now central as businesses push autonomous AI from pilots into production across hybrid cloud systems.
New controls aim to let enterprises run autonomous AI agents more securely across hybrid cloud systems, with tighter governance and audit trails.
The tie-up gives organisations real-time controls against prompt injection and data leakage as enterprise AI moves into live deployment.
Cloud operators can now sell AI infrastructure with validated software controls, as Rafay joins an early NVIDIA-approved group for production deployments.
The integration aims to curb prompt injection and data leaks as enterprises push AI agents into production across cloud and on-premises systems.
Enterprises gain governed access to Nemotron 3 inference and training environments as Torque extends control across cloud, on-premises and deskside GPUs.
The update targets firms weighing private cloud for production AI, with Broadcom citing cost, security and governance pressures in its research.
.webp)
