Nvidia has unveiled its latest innovation, Dynamo, an open-source software designed to enhance the efficiency and scalability of AI reasoning models. This new platform aims to optimise AI inference requests across extensive GPU networks, enabling service providers to grow and increase revenue.
Key Takeaways
Dynamo's Purpose: Enhances AI inference capabilities across GPU networks.
Collaboration: Developed in partnership with Perplexity.
Integration: Available via Nvidia's NIM microservices and supported by major cloud service providers.
Event: Announced at the GTC 2025 conference in San Jose, California.
What Is Nvidia Dynamo?
Nvidia Dynamo is a cutting-edge software platform that boosts GPU efficiency by managing inference across thousands of GPUs. It optimises various stages of large language model processing separately, allowing for more sophisticated AI models that can learn and adapt over time.
Jensen Huang, Nvidia's founder and CEO, stated, "Industries around the world are training AI models to think and learn in different ways, making them more sophisticated over time. To enable a future of custom reasoning AI, Nvidia Dynamo helps serve these models at scale, driving cost savings and efficiencies across AI factories."
Collaboration With Perplexity
Nvidia has partnered with Perplexity to enhance the capabilities of Dynamo. Aravind Srinivas, CEO of Perplexity, expressed enthusiasm about the collaboration, stating that their AI software is set to achieve significant advancements alongside Nvidia. Huang praised Perplexity and Srinivas, referring to them as his "favourite partners."
Availability and Integration
Dynamo will be accessible through Nvidia's NIM microservices and will be supported in future releases of the Nvidia AI Enterprise software platform. The platform will also be integrated with major cloud service providers, including:
Oracle
AWS
Microsoft
IBM
Google Cloud
This integration will allow enterprises to serve AI models across disaggregated inference scenarios, enhancing flexibility and scalability.
Supported Technologies
Dynamo supports various programming frameworks, including:
PyTorch
SGLang
vLLM
This compatibility ensures that developers can leverage Dynamo within their existing workflows, making it easier to implement advanced AI solutions.
Highlights from GTC 2025
The announcement of Dynamo was made during Nvidia's GTC 2025 conference, which took place from March 17 to 21 in San Jose, California. The event attracted approximately 25,000 attendees, with Huang delivering a keynote address at the SAP Center to accommodate the large audience. Other notable highlights from the conference included:
Introduction of Blackwell Ultra GPUs.
Nvidia's commitment to invest billions in U.S. chip manufacturing over the next four years.
As AI technology continues to evolve, Nvidia's Dynamo is poised to play a crucial role in enhancing AI inference capabilities, driving innovation and efficiency across various industries.