Jensen Huang GTC Taipei 2026 Keynote: AI Enters Agentic Era, Vera Rubin Platform and Two New Chips Unveiled

TradingKey - NVIDIA ( NVDA) CEO Jensen Huang delivered a keynote speech at GTC Taipei 2026. He stated that AI is transitioning from generative AI into a new phase of agentic AI, while concurrently unveiling the Vera Rubin AI computing platform, the first standalone data center CPU "Vera," and the RTX Spark chip for the AI PC market.

The Era of Agentic AI Has Officially Arrived, Proposing “Tokenomics”

Jensen Huang announced that the era of agentic AI has officially arrived. He explained how agentic AI works, noting that it has a two-layer structure. The first layer is the large language model, which functions like a brain responsible for reasoning and planning. The second layer is an external orchestration engine that acts like an operating system, connecting the AI model to various tools.

In the past, we had to open software and click buttons. In the future, we will only need to describe what we want to the AI, and it will complete the work automatically. The primary agent of operation is shifting from humans to AI.

Some worry that agentic AI will lead to the collapse of software companies. Huang countered, "On the contrary. When AI is no longer limited by the human population, a massive number of AI agents will use far more software tools than humans do. This is the best era to be a software company."

Meanwhile, Huang introduced the concept of "Tokenomics," arguing that tokens themselves are profit-generating assets and that AI companies will build more "token factories."

He used a set of data to illustrate the trend of growing demand. There are 30 million to 40 million professional developers worldwide. The usage of AI-assisted programming grew from 300 million instances in 2023 to 400 million in 2024 and 500 million in 2025, reaching nearly 1.4 billion in just the first few months of 2026.

Based on this data, Huang refuted the claim that "AI will lead to unemployment." He stated, "The talk of AI reducing jobs is absolute nonsense; in fact, more software engineers are being hired."

Vera Rubin: NVIDIA's Most Ambitious Product in History

Jensen Huang confirmed that the next-generation AI computing platform, Vera Rubin, has fully entered the mass production phase. He called it "the most ambitious project in Nvidia's history," with 40,000 engineers across the company involved in its development.

Vera Rubin is not a single chip, but a complete end-to-end designed system. The entire platform is manufactured using TSMC's 3nm process. Microsoft ( MSFT ), Dell ( DELL ), and CoreWeave ( CRWV) have already led the way in deploying Vera Rubin engineering samples. Mass shipments are scheduled to begin in the fall of 2026.

Vera CPU secures its first major customers, including Anthropic, OpenAI, and SpaceX.

NVIDIA has unveiled the Vera central processing unit, specifically designed for agentic AI. This marks NVIDIA's first standalone data center microprocessor, which will compete directly with products from Intel and AMD.

Jensen Huang said, "AI agents will become the largest users of computing. Vera is the first CPU designed for this future."

In terms of performance, Vera completes tasks 1.8 times faster than x86 CPUs. Initial customers include Anthropic, OpenAI, and SpaceX.

Nvidia officially enters the AI PC market

Nvidia has unveiled the RTX Spark PC chip, co-developed with MediaTek, marking its first entry into the AI PC market with a self-developed chip.

The RTX Spark chip utilizes TSMC's 3nm process. The high-end version is equipped with a 20-core CPU and 6,144 CUDA cores, delivering approximately 200 TOPS of on-device AI performance and supporting 128GB of unified memory—a configuration designed to compete with Apple's M5 Pro and M5 Max.

Manufacturers including Dell, ASUS, and Lenovo will launch Windows PCs powered by the RTX Spark chip, primarily targeting power users and the enterprise market requiring on-device AI computing power.

NVIDIA and TSMC Bring AI to Wafer Fabs

NVIDIA announced at the conference that TSMC ( TSM) is applying NVIDIA's accelerated computing and AI technologies to its own wafer fabs. This is being done to address the challenges encountered in the design and manufacturing of chips using more advanced process nodes.

TSMC is using NVIDIA's CUDA-X software libraries and AI models to accelerate computational lithography, transistor simulation, process control, and factory operations optimization. Additionally, TSMC is leveraging NVIDIA's Metropolis platform to advance vision AI-based defect detection. This technology can enhance the detection precision of nanoscale defects while reducing the workload for repetitive labeling and retraining.

NVIDIA CEO Jensen Huang stated that TSMC is bringing NVIDIA's AI into wafer fabs, utilizing simulation, optimization, and AI technology to tackle the world's most complex design and manufacturing challenges.

TSMC Chairman C.C. Wei noted that by applying NVIDIA's technology, TSMC is strengthening its technological leadership and manufacturing advantages. This collaboration signifies that NVIDIA's AI technology is extending from data centers to the furthest upstream stages of semiconductor manufacturing.

During this keynote, NVIDIA introduced a series of new AI models. These include Nemotron 3 Ultra, designed to support long-running agents, and Alpamayo 2, specifically developed for Level 4 robotaxis. The company also launched Isaac GR00T, a humanoid robot reference platform for academic research, and Cosmos 3, an open-world model driving the development of physical AI.

Taiwan is a critical starting point for Nvidia’s ecosystem.

Jensen Huang opened his speech by saying, "It is great to be back home." He emphasized that Taiwan possesses "the world's best supply chain ecosystem." The ecosystem built by NVIDIA extends all the way to the absolute upstream of the entire supply chain, and that upstream is located in Taiwan.

The Vera Rubin platform already has 150 Taiwanese supply chain partners participating in mass production. Partners announced at the event included nearly 20 Taiwanese enterprises such as TSMC, Foxconn, Quanta, Wistron, ASUS, and Gigabyte.

The speech was held at the Taipei Music Center. Long queues formed outside the venue that morning. NVIDIA distributed chibi-style Jensen Huang figurines to the media.

Boosted by Jensen Huang's speech, the Taiwan stock market jumped more than 1,000 points in early trading, with the Weighted Index breaking 45,600 to hit a fresh record high. Heavyweight stocks including TSMC, Foxconn, MediaTek, and Quanta all reached all-time highs. The COMPUTEX 2026 exhibition will open on June 2 at the Nangang Exhibition Center.