Here are the key takeaways from Nvidia’s AI summit By Investing.com

Nvidia (NASDAQ:)’s October 2024 AI Summit provided valuable insights into the company’s advancements in AI, with key updates on its Blackwell system, AI model performance improvements, and real-world applications.

Here are the main highlights from Evercore ISI’s analysis of the event:

Blackwell Update: The firm said in a note Thursday that Nvidia shared that eight partners are currently working on Blackwell systems, with volume production expected to ramp up in Q4 2024.

Evercore added that details about the GB200 NVL72 system components were also discussed, showing progress in AI hardware.

Networking for Inference: As AI transitions from training to inference, networking plays a more critical role due to latency requirements. Nvidia emphasized how network infrastructure is essential to delivering faster and more efficient AI outcomes.

CUDA-X Libraries: The company’s CUDA-X libraries continue to enhance system performance, according to Evercore, offering up to a 150x acceleration in RAG workflow processing, underscoring the increasing efficiency of AI-driven processes.

NIMs for AI Models: “NIMs (NVIDIA Inference Microservices) improve AI model performance 2-5x. Customers can use NeMo to improve and fine tune NIMs while benefiting from NVDA’s work to optimize NIMs for variety of hardware configurations,” wrote Evercore.

They added that Nvidia’s confidential computing stands out by offering stronger cybersecurity features for large language models (LLMs) compared to open-source alternatives.

Agentic AI: Evercore stated that Nvidia showcased “Agentic AI,” already being used in digital avatars and offering the potential for future proactive, unprompted analysis.

Real-World AI Applications: Nvidia reportedly highlighted significant real-world applications, including Lockheed Martin (NYSE:)’s use of AI to process radar data during a military attack, reducing false positives in hours instead of months.

Other examples are said to have included AI-powered autonomous drone inspections and Siemens using AI digital twins for complex product simulations.

Energy Demand: With AI contributing to a 15% increase in power demand, Nvidia noted that U.S. data centers are driving a quarter of this growth, prompting investments in clean energy, said Evercore.

“Notably, gigawatt data centers would require building their own power plants,” stated Evercore.