NVIDIA's 'Super Bowl of AI' - Huang's GTC keynote

Investing.com -- NVIDIA Corporation (NASDAQ:NVDA) CEO Jensen Huang described last year's GTC conference as the "Woodstock of AI." This year, during GTC 2025, Huang upped the comparison to the Super Bowl.

To start his keynote speech, Huang provided his vision for AI, based upon the current timeline of AI development. He described four waves of AI:

• Perception AI: Initiated about 10 years ago, focusing on speech recognition and other simple tasks.

• Generative AI: The focus of the past 5 years, involving text and image creation through predictive patterns.

• Agentic AI: The current phase in which AI interacts digitally and performs tasks autonomously, characterized by reasoning models.

• Physical AI: The future of AI, powering humanoid robots and real-world applications.

Huang pointed to an "enormous challenge" of computation in the AI industry, stating that at the current point of Agentic AI, computation takes 100 times more tokens and resources than was originally expected. He explained that this is because reasoning models require tokens for the numerous steps in their reasoning process.

However, Huang reassured that the industry is responding well, citing NVIDIA's GPU shipments for the top 4 Cloud Service Providers: Amazon (NASDAQ:AMZN), Microsoft (NASDAQ:MSFT), Meta (NASDAQ:META), and Google (NASDAQ:GOOGL). In NVIDIA GPU Hopper's peak year, the company shipped around 1.3 million units, while for NVIDIA's Blackwell in 2025, 3.6 million units have been ordered.

Huang maintained that the demand for more computation is being met, emphasizing that in just one year, the AI infrastructure segment of the market has shown incredible growth. Notably, the CEO stated that now is "AI's inflection point" as it is becoming increasingly more useful and popular.

A major highlight of his speech was the projected expansion of data center infrastructure, in which Huang expects capital expenditures of over $1 trillion by the end of 2028, driven by AI and accelerated compute demands.

Huang also projected that in the future, every company's fabs will require two separate facilities: one for manufacturing the product, and one driven by AI to manufacture the information for the product. In the chip-making industry, he referred to this idea as having one factory for wafers, and another focused on information for the wafers.

After discussing the future of AI, Huang and NVIDIA launched several announcements:

1. NVIDIA and GM Collaboration: NVIDIA and General Motors (NYSE:GM) announced a partnership to develop AI for next-generation vehicles, factories, and robots. GM will adopt NVIDIA DRIVE AGX for in-vehicle hardware and use Omniverse for digital twins of assembly lines.

2. AI-Native 6G Networks: NVIDIA partnered with T-Mobile, MITRE, Cisco (NASDAQ:CSCO), ODC, and Booz Allen (NYSE:BAH) Hamilton to develop AI-integrated wireless networks for 6G. The collaboration aims to enhance connectivity for billions of devices with improved spectral efficiency and new revenue streams for telecom companies.

3. NVIDIA Dynamo Library: NVIDIA released the open-source Dynamo inference software that doubles performance for Llama models and boosts token generation by over 30x per GPU. Adopters include major cloud providers and AI companies like AWS, Google Cloud, Meta, and Microsoft Azure.

4. Blackwell Ultra AI Factory Platform: NVIDIA introduced the GB300 NVL72 with 72 Blackwell Ultra GPUs, offering 1.5x more AI performance than previous generations. The platform supports agentic and physical AI with availability expected in the second half of 2025.

5. Spectrum-X Networking Switches: NVIDIA unveiled silicon photonics networking switches offering 3.5x energy savings and 10x resilience for AI data centers. Quantum-X InfiniBand switches will be available later this year, with Spectrum-X Photonics Ethernet coming in 2026.

6. Llama Nemotron AI Models: NVIDIA launched a family of open reasoning AI models with enhanced math, coding, and decision-making capabilities, improving accuracy by up to 20%. Models are available in Nano, Super, and Ultra sizes, optimized for different hardware configurations.

7. Isaac GR00T N1 Robot Model: NVIDIA released the first open, customizable foundation model for humanoid robots with simulation frameworks that improved performance by 40%. Early adopters include Agility Robotics, Boston Dynamics, and NEURA Robotics.

8. Cosmos World Foundation Models: NVIDIA announced new physical AI models including Cosmos Transfer, Predict, and Reason, along with blueprints for generating synthetic data. Early adopters include 1X, Agility Robotics, Figure AI, and Uber (NYSE:UBER).

In conclusion, Jensen Huang's GTC 2024 keynote speech has delivered a vision of AI's transformative potential. Through his insights into the future of AI and his company's technological announcements, NVIDIA seems poised to lead companies into the next generation of tech.

Despite Huang’s positive sentiments and ideas, NVIDIA stock fell 3.3% on the day, following suit with other tech stocks as tariff worries and market oversaturation concerns persist. These anxieties have caused investors to rotate out of the “Magnificent 7” tech stocks in favor of more defensive stocks lately.

This article first appeared in Investing.com