Nvidia GTC 2025: Jensen Huang Set to Unveil Next-Gen AI Chips and Strategy

Jensen Huang is preparing to take center stage at Nvidia's annual GTC developer conference, where the charismatic CEO is expected to unveil a sweeping roadmap designed to keep Nvidia ahead of a fast-moving pack of AI chip competitors. The event, held in the heart of Silicon Valley across four days, has grown into one of the most anticipated gatherings in the technology world.

Industry analysts anticipate Huang will deliver updates spanning Nvidia's next-generation chip architectures — from the current Rubin lineup to the upcoming Feynman series — while placing heavy emphasis on inference computing, agentic AI, and data center infrastructure. With governments and corporations worldwide pouring hundreds of billions of dollars into AI-driven data centers, the stakes for Nvidia have never been higher.

Despite commanding over 90% of both the AI training and inference chip markets today, Nvidia faces mounting pressure. Rivals are closing the gap, and major customers like OpenAI and Meta are developing proprietary chips in-house. Analysts widely expect Nvidia's market share to face meaningful erosion starting around 2027 as custom application-specific integrated circuits gain traction, particularly in inference workloads.

To reinforce its position, Nvidia made a bold $17 billion acquisition of Groq in December, a startup specializing in fast, cost-efficient inference technology. Huang is expected to demonstrate at GTC how Groq's capabilities integrate seamlessly into Nvidia's established CUDA software platform. New server products combining Groq's chips with Nvidia's networking technologies are also anticipated.

The rise of agentic AI — where fleets of AI assistants autonomously handle tasks and require intelligent orchestration layers — is shifting demand toward CPUs, an area Nvidia is actively expanding into. Additionally, the company's recent $2 billion investments in laser-optics firms Lumentum and Coherent signal a longer-term push to accelerate chip-to-chip communication inside massive AI data centers.