Anthropic Ships Sonnet 5 With Agentic Safety and Pricing Details
1 July 20265 min read
Technology
Nvidia's Blackwell Ultra Chips Begin Reaching Enterprise Customers
Nvidia has started shipping its Blackwell Ultra GPU architecture to cloud providers and enterprise customers, with the hardware designed to support larger AI models and more demanding inference workloads than its predecessor.
Illustrative image. Cedar S. Insights uses editorial stock photography; images do not depict specific events described in articles.
Nvidia has begun shipping its Blackwell Ultra GPU architecture to cloud providers and enterprise customers. The hardware succeeds the Blackwell generation and is designed to handle larger AI models and higher-throughput inference workloads.
The Blackwell Ultra chips offer increased memory bandwidth and capacity compared with the previous generation, which matters for running large language models and multimodal systems that require holding substantial model weights in GPU memory during inference.
Cloud providers including AWS, Google Cloud, Microsoft Azure and Oracle Cloud Infrastructure have announced plans to offer Blackwell Ultra-based instances. Enterprise customers purchasing directly through Nvidia's partners can also access the hardware for on-premises AI infrastructure.
Nvidia continues to face questions about supply constraints. Demand for AI compute has consistently outpaced manufacturing capacity, and the company has said it expects Blackwell Ultra availability to improve through the second half of 2026.
Performance figures cited in Nvidia announcements are typically measured under controlled benchmark conditions. Real-world throughput depends on model architecture, batch size, memory utilisation and software stack. Enterprise buyers should evaluate performance against their specific workloads.
Why It Matters
AI infrastructure investment is running at extraordinary levels, and the availability of each new GPU generation shapes what models can be trained and deployed at what cost. Blackwell Ultra's increased memory capacity is particularly relevant for inference: as models grow larger and multimodal capabilities expand, the ability to hold more of a model in GPU memory reduces latency and increases throughput. The pace at which this hardware reaches customers will influence the competitive dynamics of AI services through 2026 and into 2027.
Primary Sources
Our sourcing: Cedar S. Insights provides source-led editorial analysis. Reported company, institutional and regulatory claims are attributed to their original sources unless stated otherwise.
Corrections: If a material factual error is identified, Cedar S. Insights will update the relevant article and preserve the distinction between the corrected statement and supporting evidence.
Topics
NvidiaBlackwellGPUAI InfrastructureData Centre