Nvidia deepens AI push as stock lags tech rally

Nvidia (NVDA) has introduced the Vera Rubin AI infrastructure platform, described as a new-generation system delivering performance and cost-per-compute improvements compared with its Blackwell architecture. The platform is aimed at advanced AI data center workloads for both training and inference.

The company also extended its collaboration with Micron (MU) around next-generation HBM4 memory. This partnership is designed to support future AI product plans and tighten Nvidia’s links to critical components in high-end AI infrastructure.

Nvidia is backing firms such as SiFive and Firmus Technologies involved in RISC-V CPUs and AI data centers, and is participating in Project Glasswing, a cybersecurity initiative targeting data center-scale AI systems. These moves highlight activity across CPU architectures, physical data center build-out, and security within AI stacks.

As a result, Nvidia is positioned not only as a GPU provider but as a broader AI infrastructure supplier of GPUs, systems, and software to data centers worldwide.

For its first quarter of fiscal 2027, Nvidia expects revenue of $78 billion at the midpoint of guidance, which would represent about 77% year-over-year growth. Investor attention is turning to the company’s guidance for the second quarter and the rest of fiscal 2027.

Nvidia is set to resume selling its H200 chips in China after regulatory pressure from Chinese and U.S. authorities disrupted its business there last year. According to some estimates cited in the coverage, Nvidia still maintains a 55% share of the Chinese market despite these restrictions.

Improving conditions in China could support meaningful revenue contributions, with any impact likely to begin appearing in Nvidia’s Q2 fiscal 2027 outlook. Markets are expected to scrutinize that guidance for signs of renewed momentum in the region.

Nvidia plans to release its next-generation AI chip, Vera Rubin, in the second half of 2026. The company has projected $1 trillion in combined sales from Vera Rubin and Blackwell through 2027, underscoring expectations for sustained demand for advanced AI hardware.

Recent commentary also links Nvidia’s growth prospects to rising adoption of agentic AI and physical AI. Chief Financial Officer Colette Kress said on the February earnings call that these applications, built on increasingly smarter and multimodal models, are beginning to drive the company’s financial performance.

Nvidia notes that customers such as Meta Platforms (META) and Anthropic are investing heavily in compute capacity for agentic AI systems. Analyst forecasts cited in the articles point to broad enterprise adoption of such technologies, which Nvidia expects will require extensive computing infrastructure.

Nvidia’s GPUs command an approximate 90% share of AI infrastructure chips, supported by its CUDA software ecosystem and NVLink interconnect, which help bind GPUs into large, unified systems. Acquisitions such as Mellanox have expanded its networking capabilities, while work on CPUs and language processing units has turned it into a full-stack AI infrastructure provider.

At the same time, competition is increasing. Custom AI ASICs, Alphabet’s (GOOGL) TPUs, Amazon’s (AMZN) Trainium, and GPUs from Advanced Micro Devices (AMD) are cited as emerging alternatives, especially in inference workloads where power efficiency is critical. Large cloud providers are also pursuing custom chip designs.

Nvidia’s shares have declined nearly 5% this year and recently traded around $188.63 to $188.67, with a market capitalization of about $4.6 trillion. The stock is described as roughly 30% below a consensus analyst target of $268.22, and another source cites a 12-month median target of $265.

Valuation assessments in the articles characterize Nvidia as trading close to estimated fair value on one model and at about 21 to 22 times forward earnings on others, even as its gross margin is reported at 71.07% and the dividend yield at 0.02%.