Nvidia’s annual GTC conference, often referred to as the “Super Bowl of AI,” has once again put CEO Jensen Huang in the spotlight.
Speaking to a packed audience in San Jose, California, Huang delivered a keynote filled with bold predictions about the future of AI and Nvidia’s role in it. With the competition heating up, Nvidia is taking an aggressive approach—betting on more computing power, faster performance, and continued innovation to stay ahead.
One of the key takeaways from Huang’s speech was that AI models will require significantly more computing power than before. He dismissed concerns that recent breakthroughs, such as DeepSeek’s R1 model, would make AI more efficient and reduce the need for Nvidia’s chips. Instead, Huang argued that advanced AI models—especially those focused on reasoning—demand more processing power due to the increasing complexity of their responses.
This need for greater efficiency and speed is at the heart of Nvidia’s latest product strategy. The company has already deployed 3.6 million Blackwell GPUs, and an upgraded version, Blackwell Ultra, promises three times the performance. Additionally, Huang revealed the Vera Rubin chip, set for release in late 2026, and hinted at the Feynman chip, expected in 2028.
Nvidia is positioning its hardware and software as the key to making AI systems faster and smarter. According to Huang, the two biggest challenges for AI systems today are:
- Handling a high volume of users simultaneously
- Delivering responses with near-instant speed
“If you take too long to answer a question, the customer is not going to come back,” Huang explained.
Nvidia’s latest chips, he argued, are the only ones capable of excelling in both areas.
In addition to hardware advancements, Nvidia is pushing software innovations. Huang announced Dynamo, a free software tool designed to accelerate AI reasoning processes. He also highlighted Nvidia’s growing influence in the automotive sector, revealing that General Motors has selected Nvidia to power its self-driving vehicle fleet.
Despite Nvidia’s confidence, challenges remain. The company’s AI chips come at a steep price, ranging between $30,000 and $40,000 each. With the return on investment for AI still uncertain, enterprise customers may hesitate before committing to large-scale purchases.
Additionally, Nvidia has faced manufacturing delays, particularly with its Blackwell chip. Supply chain issues and design flaws have slowed production, raising concerns about whether Nvidia can maintain its dominance in an increasingly competitive market.
Still, Huang remains optimistic, emphasizing that AI’s growth trajectory continues to accelerate.
“The amount of computation we need as a result of agentic AI, as a result of reasoning, is easily 100 times more than we thought we needed this time last year,” he said.
