What Really Happened Inside NVIDIA’s GTC DC 2025

NVIDIA’s GTC DC 2025 showcased the company’s expanding influence across every layer of the AI ecosystem, from chips and data centres to networking and quantum computing. 

The event in Washington, D.C. highlighted how NVIDIA is positioning itself at the centre of the world’s AI infrastructure buildout, forging partnerships that span government, industry, and scientific research.

“We couldn’t do what we do without NVIDIA’s ecosystem of partners,” CEO Jensen Huang said, nodding to GTC’s reputation as “the Super Bowl of AI.”

That momentum is also reflected in the company’s market performance. NVIDIA’s market capitalisation is approaching $5 trillion, with the company’s stock reaching an all-time high during October 2025. Specifically, as of the most recent trading data, NVIDIA’s market cap was approximately $4.89 trillion, close to crossing the $5 trillion threshold.

In his keynote, Huang said that the world is moving from using general-purpose CPUs to faster GPU-based computing. He explained that computing is now being built around AI factories, which are large data centres that train, run, and manage intelligent systems. Huang said AI is no longer limited to research or model training but is now being used in industries and government projects.

Huang said the company expects to surpass $500 billion in Blackwell GPU sales by the end of 2026, following revenue of over $100 billion recorded in the first half of 2025.

At the heart of NVIDIA’s technological engine lies the CUDA-X stack, which powers applications across AI, data science, simulation, and quantum computing. “This really is the treasure of our company,” Huang said, as he highlighted how CUDA-X libraries like TensorRT-LLM, cuOpt, and CUDA-Q enable breakthroughs from deep learning to quantum simulation.

Telecom Solutions

NVIDIA will invest $1 billion in Nokia as part of a new collaboration to develop AI-powered radio access networks (AI-RAN) and data centre networking solutions. The partnership also includes building an AI-native 6G radio access network (RAN) platform that integrates GPU acceleration and real-time learning into next-generation communication systems.

NVIDIA also announced America’s first AI-native wireless stack for 6G. Developed with Booz Allen Hamilton, Cisco, MITRE, ODC, and T-Mobile, the project accelerates the development of smart, software-driven networks powered by NVIDIA’s accelerated computing platform.

Another major reveal was the company’s work with the US Department of Energy and federal research laboratories to develop the nation’s largest AI supercomputers with Oracle. One such project, at Argonne National Laboratory, will feature 100,000 NVIDIA Blackwell GPUs and serve as a public research platform for agentic AI and scientific simulations.

Autonomous Mobility

The company is also turning its focus to autonomous transportation. NVIDIA announced a partnership with Uber to build and scale what it calls the world’s largest Level 4-ready autonomous mobility network. 

The collaboration will use NVIDIA’s next-generation robotaxi and autonomous delivery fleets, powered by the new NVIDIA DRIVE AGX Hyperion 10 platform and DRIVE AV software designed for full Level 4 autonomy.

Under the partnership, Uber plans to expand its global autonomous vehicle fleet to 100,000 vehicles starting in 2027. 

Quantum Leap and AI Factories 

Moreover, NVIDIA introduced NVQLink to connect quantum computers with AI supercomputers. The technology enables real-time communication between quantum and classical computing systems, paving the way for hybrid architectures capable of simulating nature itself.

NVIDIA also unveiled the new BlueField-4 data processing unit (DPU), a key component built to power AI factory operating systems. This DPU combines a 64-core NVIDIA Grace CPU and ConnectX-9 networking capable of 1.6 terabits per second. The BlueField-4 supports AI infrastructure management and accelerates data movement within AI systems by handling tasks that break the traditional data centre bottlenecks.

Omniverse DSX is a comprehensive platform blueprint for designing and operating AI factories at scale—facilities delivering power from 100 megawatts up to multi-gigawatt installations. 

NVIDIA also detailed the expansion of its Blackwell platform for inference, optimised for large-scale deployment in enterprises and data centres. 

Open Models, Open Science

NVIDIA’s embrace of openness also took centre stage. Huang announced the release of hundreds of open models and datasets, organised into four key model families — Nemotron for agentic reasoning, Cosmos for synthetic and physical data, Isaac GR00T for robotics, and Clara for biomedical AI. 

The keynote also spotlighted new partnerships across industries. CrowdStrike will integrate NVIDIA’s Nemotron-based models for real-time cybersecurity, while Palantir is combining NVIDIA’s accelerated computing stack with its Ontology platform to process data at unprecedented scale. 

Meanwhile, companies such as Foxconn, Caterpillar, Johnson & Johnson, and Disney are using NVIDIA’s Omniverse to build digital twins of factories and robotic systems, accelerating America’s push toward reindustrialisation.

As the keynote came to a close, Huang summed up NVIDIA’s mission with characteristic confidence: “The Age of AI has begun. Blackwell is its engine. Made in America, made for the world.”

The post What Really Happened Inside NVIDIA’s GTC DC 2025 appeared first on Analytics India Magazine.