Not Just GPU! Nvidia (NVDA.US) GTC 2026 Launches LPU, CPU New Products, Comprehensively Covers Every Link of AI Data Centers

TechNews Finance APP has learned that on Monday, Eastern Time, NVIDIA (NVDA.US) officially kicked off the GTC conference in San Jose, California, unveiling several new chips and platforms all at once, including the next-generation Nvidia Groq 3 Language Processing Unit (LPU) and the Vera central processing unit (CPU) chassis designed to compete with Intel (INTC.US) and AMD (AMD.US).

It is reported that NVIDIA launched a total of five large server chassis, each targeting different scenarios within AI data centers.

The most significant announcement is the Nvidia Groq 3 chip. In December last year, NVIDIA acquired Groq-related technology licensing through a $20 billion deal and brought its founders Jonathan Ross, President Sunny Madra, and core team into the fold.

Groq processors focus on AI inference—the core process of running AI models. When users input commands into ChatGPT, Claude, or Gemini and receive responses, inference technology is at work behind the scenes.

Unlike NVIDIA’s general-purpose GPUs that can both train and run models, the launch of Groq 3 marks the company’s official possession of dedicated inference chips, addressing the urgent need for the AI market to shift from model training to model deployment.

Ian Buck, Vice President of Large-Scale and High-Performance Computing at NVIDIA, explained that although GPUs support larger memory capacities, Groq 3’s LPU memory offers faster access speeds. By combining the performance advantages of both, the new Groq 3 LPX platform was created—this server chassis integrates 128 independent Groq 3 LPUs. When working with the Vera Rubin NVL72 rack, it can increase throughput per megawatt by 35 times, creating tenfold revenue potential.

“The LPX architecture, optimized for trillion-parameter models and multi-million token contexts, complements Vera Rubin perfectly, maximizing efficiency in power consumption, memory, and computing power. This breakthrough in throughput per watt and token performance will drive ultra-high-end trillion-parameter inference services, opening new growth opportunities for all AI service providers,” NVIDIA emphasized in an official statement.

The launch of the LPX chassis responds strongly to market concerns that NVIDIA might lose its advantage amid emerging inference chip startups. Meanwhile, the independently deployed Vera CPU rack system is also noteworthy—this cluster system, built with 256 liquid-cooled Vera chips, marks NVIDIA’s first deconstruction of the Vera CPU from the “Vera Rubin superchip” (which includes one Vera CPU and two Rubin GPUs).

As intelligent AI agents rise, the strategic value of CPUs is increasingly evident. When AI agents need to perform tasks like browsing the web or extracting table information, CPU performance directly impacts efficiency. In scenarios such as data mining and personalized recommendations, where context analysis for GPUs is required, CPUs also play an irreplaceable role.

“Vera is the ultimate CPU tailored for intelligent agent AI workloads,” Buck said. “We have redefined CPU architecture—powered by NVIDIA’s Olympus cores, specially designed for AI execution, capable of faster response times under extreme conditions, perfectly suited for all reinforcement learning scenarios.”

This is not NVIDIA’s first foray into the CPU field. Last month, a deal was reached with Meta (META.US) to deploy the largest-ever cluster of the previous-generation Grace CPUs. The release of Vera as an independent product marks NVIDIA’s formal establishment of a “GPU + CPU” dual-drive strategy, targeting the data center market dominated by Intel and AMD.

In addition to the above products, NVIDIA also showcased the Bluefield-4 STX storage chassis system (which achieves performance leaps over traditional solutions) and the Spectrum-6 SPX networking chassis.

As demand for AI platforms continues to grow, NVIDIA’s new product lines are expected to further boost data center revenue. In fiscal year 2026, its data center revenue reached $19.35 billion, a significant increase from $11.62 billion in 2025. Among the $650 billion AI capital expenditures planned this year by giants like Amazon (AMZN.US), Google (GOOGL.US), Meta, and Microsoft (MSFT.US), NVIDIA is undoubtedly set to capture a substantial share.