NVIDIA unveils three new "trump cards"

NVIDIA unveiled three key updates at today’s GTC conference, dispelling market concerns about its product cadence and demonstrating how its business boundaries are extending from GPUs to CPUs, and even to consumer-level AI PCs.

CEO Jensen Huang announced that the flagship AI platform Vera Rubin has officially entered mass production, alleviating prior investor worries about potential delays. Meanwhile, NVIDIA launched the Vera CPU designed specifically for agentic AI, delivering significant performance advantages over x86 competitors on major workloads. The company also partnered with Microsoft to release a new generation of ARM-based Windows PCs—RTX Spark.

According to Chase Trading Desk, Citi Research analyst Atif Malik characterized these developments as “positive” signals. Particularly noteworthy is that the launch of the Vera CPU signifies the value ratio of CPU to GPU sales is evolving from the historical 1:2 towards nearly 1:1, opening new revenue growth opportunities for NVIDIA.

Citi maintains its buy rating for NVIDIA, with a target price of $300, implying about 42% upside from the current stock price of $211.14.

Vera Rubin Mass Production, Delay Rumors End

The news of Vera Rubin platform entering mass production is the most eye-catching highlight at this Computex. According to Citi Research, the market was previously concerned about whether the platform could be delivered on schedule, and Jensen Huang’s public confirmation effectively resolved this uncertainty.

The Vera Rubin platform consists of five dedicated racks, operating as an integrated AI supercomputer. It integrates the Vera Rubin NVL72 GPU system, Vera CPU, BlueField-4 STX storage, and Spectrum-6 networking. Compared to the previous generation Blackwell, this platform delivers 10 times agent throughput under large-scale deployments. Major server manufacturers and supply chain partners have started mass-producing Vera Rubin-based systems for AI labs, cloud service providers, and hyperscale data center operators.

Also entering mass production is NVIDIA’s Spectrum-X ethernet photonic interconnect technology. By combining co-packaged optics with Spectrum-X switches, it enables AI factory networking at millions of GPU scale. Compared with networks using traditional transceivers, this solution offers five times energy efficiency, five times longer AI run times, and 1.3 times faster deployment speed. CoreWeave, Microsoft, and OCI are among the first ecosystem partners.

CPU Strategy Leap, Direct Challenge to AMD and Intel

The most surprising development for analysts at this Computex is the strategic elevation of NVIDIA’s CPU business.

In his presentation, Jensen Huang described Vera CPU as the core bottleneck of the agentic AI era: “In the age of agentic AI, the CPU has become a bottleneck holding back the GPU.” Vera CPU is built specifically for agentic AI workloads, with SQL computation performance three times higher than x86 counterparts (AMD/Intel), six times higher for data processing, and about 1.8 times higher for Python, code analysis and compilation compared to x86 CPUs on mainstream agentic tools.

Additionally, the CPU/GPU ratio is shifting. Citi Research pointed out that historically, NVIDIA bundled one CPU with every two GPUs sold, but with the surge in agentic AI demand, the ratio of CPU to GPU content is now approaching 1:1. NVIDIA has now launched a standalone CPU rack product with 256 CPUs, which also support memory expansion, further enhancing commercial value. Vera CPU has already been adopted by Anthropic and OpenAI as early customers.

Citi Research believes this will substantially increase the overall value of NVIDIA’s system offerings, with CPU business contribution expected to exceed prior market forecasts.

New Era for AI PCs, RTX Spark Targets Consumer Market

In the personal computing segment, NVIDIA and Microsoft released the RTX Spark super chip, bringing agentic AI capabilities to consumer-level Windows devices.

RTX Spark uses the ARM architecture, connecting the Blackwell RTX GPU directly with a 20-core Grace CPU (co-designed with MediaTek) via NVIDIA’s NVLink C2C. It offers 1 petaflop of AI compute and up to 128GB of unified memory, targeting all-day lightweight laptops and ultra-efficient compact desktops.

In terms of capabilities, RTX Spark can render 3D scenes over 90GB, edit 12K 4:2:2 format video, locally run large language models with 120 billion parameters and support a context of millions of tokens, and run AAA games at 1440p resolution with over 100 frames per second. NVIDIA states that the product leverages thirty years of core technology, including CUDA, RTX, DLSS, FP4, TensorRT, OptiX, and more.

Citi Research describes the launch as marking a “new era” for PCs, believing RTX Spark will integrate AI creation, gaming, and personal agent capability into a single consumer device.

~~~~~~~~~~~~~~~~~~~~~~~~

The above highlights are from Chase Trading Desk.

For more detailed analysis, including real-time interpretation and frontline research, please join [Chase Trading Desk Annual Membership]

Risk Warning and DisclaimerThe market has risk, invest cautiously. This article does not constitute personal investment advice, nor does it take into account individual users’ special investment objectives, financial situation, or needs. Users should consider whether any opinions, viewpoints, or conclusions herein are suitable for their specific circumstances. Investment is at your own risk.