Jensen Huang Unpacks Nvidia's Future: GTC 2026 Insights
Jensen Huang, CEO of Nvidia, offered profound insights into the company's trajectory, the future of AI, and global industry dynamics during an interview held in San Jose, immediately following his GTC 2026 keynote.
Keynote Highlights and Nvidia's Core Strategy
Huang's keynote revisited Nvidia's foundational journey, celebrating key milestones such as the programmable shader and the launch of CUDA two decades prior. Looking ahead, he outlined a clear strategic path for the company.
Nvidia's strategy centers on applying the CUDA approach to new industries, accelerating a vast array of software tools for use by AI agents, spanning applications from Excel and Photoshop to SQL databases. This vision extends beyond mere chip production.
Nvidia aims to build comprehensive AI factories and infrastructure globally, extending beyond just chip manufacturing to full-stack systems, networking, and storage.
This ambitious goal underlines Nvidia's commitment to providing complete, end-to-end solutions for the burgeoning AI ecosystem.
Advancements in AI Models and Capabilities
Huang emphasized that the scope of AI reaches far beyond conventional language models. AI now encompasses critical areas like protein AI, chemical AI, physical simulation, robotics, and sophisticated autonomous systems.
He acknowledged the limitations of current transformer architectures, particularly in managing long memory (KV cache) and processing continuous information like motion, as opposed to discrete tokens. To address these challenges, Nvidia has developed new model architectures, including a hybrid transformer-SSM architecture for Nemotron 3 and cuEquivariance for geometry-aware models.
AI has demonstrably crossed a threshold where models are useful at scale, boasting improved reasoning, reduced hallucinations, and enhanced grounding through methods like reflection and retrieval.
A significant practical application highlighted by Huang is the rise of coding agents. These agents demonstrate AI's ability to generate real economic value, empowering engineers to dedicate their focus to architecture and specification rather than laborious code generation.
The Evolving Role of CPUs and Groq Acquisition
Huang asserted a pivotal shift, declaring that Moore's Law is over and stressing the indispensable role of accelerated computing. Nvidia views CPUs as complementary components, carefully selecting high-performance CPUs, such as Grace, to ensure they do not bottleneck the powerful GPUs.
Nvidia's agent-focused CPUs, exemplified by Vera, are specifically designed for high single-threaded performance and robust I/O/memory bandwidth, crucial for supporting GPU-intensive tasks within data centers. While innovating internally, Nvidia also maintains a strategic Intel partnership to cater to x86 computing needs, particularly within enterprise environments.
The recent acquisition of Groq's technology and team marks a significant move. This acquisition aims to enhance low-latency, high-token-rate inference, particularly for demanding applications like coding agents. This involves a strategic approach to disaggregate inference, optimizing it for diverse workload types, from high throughput to sophisticated intelligence.
Industry Constraints and Geopolitics
Huang shed light on pressing industry challenges, noting that critical supply chain elements—including power, fab capacity, and land/infrastructure readiness—are all operating near their limits, making significant expansion profoundly challenging.
Addressing geopolitical dynamics, he expressed concern about the necessity of an American tech stack in China, while concurrently acknowledging China's substantial contributions to open-source AI and the remarkable inventiveness of its researchers, citing examples like DeepSeek, Kimi, and Qwen.
Huang criticized