Nvidia (NASDAQ: NVDA) was one of the most actively traded stocks in the U.S. market on Tuesday, following its presentation on the Blackwell platform at the Hot Chips 2024 conference at Stanford University on Monday.

Wells Fargo highlighted that while there were no new details regarding Blackwell’s availability or rollout timeline, the presentation offered several important insights.

Led by analyst Aaron Rakers, Wells Fargo noted that Nvidia’s presentation, as anticipated, underscored the company’s strategy of innovating across the entire technology stack—from hardware components like the Blackwell GPU, Grace CPU, NVSwitch Chip, BlueField-3, ConnectX-7/8, Spectrum-4, and Quantum-3, to systems software and application libraries.

Nvidia emphasized its domain-specific approach, boasting over 400 CUDA X libraries optimized for highly specialized application environments.

The analysts also pointed out that the presentation reinforced their belief in the significance of Nvidia’s fifth-generation NVLink interconnect, which is supported by the company’s next-generation PHY/SerDes innovations. The NVLink 5.0 switch chip is built on Taiwan Semiconductor Manufacturing’s (TSMC) 4NP process.

Additionally, the presentation confirmed that Blackwell is based on TSMC’s 4NP process and utilizes the latest generation of packaging technologies.

Nvidia also reiterated its commitment to an annual product cycle across its portfolio, including the introduction of Blackwell Ultra GPUs (288GB HBM3E) and Spectrum X Ultra X800, scaling to 512-radix in 2025. The Rubin GPU (8-stacks HBM4), Vera CPU, NVLink 6 (3600GB/s), ConnectX-9 (1.6T), and Spectrum-X1600 + Quantum-X160 are slated for 2026, with Rubin Ultra GPUs (12-stacks HBM4) following in 2027.

Separately, on Tuesday, Nvidia introduced NIM Agent Blueprints—a collection of pretrained, customizable AI workflows designed to help enterprise developers create and deploy generative AI applications.

These newly available services include a digital human workflow for customer service, a generative virtual screening workflow for computer-aided drug discovery, and a multimodal PDF data extraction workflow for enterprise retrieval-augmented generation, leveraging vast amounts of business data for more accurate responses, according to Nvidia.

Nvidia also noted that Accenture (ACN), Deloitte, SoftServe, and World Wide Technology are facilitating the global deployment of NIM Agent Blueprints for enterprises.

In addition, Cisco (CSCO), Dell Technologies (DELL), Hewlett Packard Enterprise (HPE), and Lenovo (OTCPK: LNVGY) (OTCPK: LNVGF) are offering full-stack Nvidia-accelerated infrastructure and solutions to accelerate the deployment of NIM Agent Blueprints.

“With the NVIDIA AI Enterprise toolkit—including NeMo, NIM microservices, and the latest NIM Agent Blueprints—our expansive partner ecosystem is well-positioned to help enterprises customize open-source models, build bespoke AI applications, and deploy them seamlessly across any cloud, on-premises, or at the edge,” said Nvidia’s Founder and CEO Jensen Huang.

Article Credit: https://seekingalpha.com/news/4144140-nvidia-unveils-nim-agent-ai-workflows-for-enterprises-highlights-blackwell-at-hot-chips-conference