Key Highlights
- At GTC 2026, Nvidia introduced the Groq 3 LPU, designed exclusively for AI inference operations
- The LPX server configuration features 128 Groq 3 chips and delivers up to 35x increased throughput per megawatt when combined with Vera Rubin NVL72
- A standalone Vera CPU rack was announced, marking Nvidia’s entry into direct rivalry with Intel and AMD’s data center business
- Vera processors target agentic AI applications, including autonomous web navigation and file data retrieval
- Fiscal 2026 data center revenues reached $193.5 billion for Nvidia, a significant jump from the previous year’s $116.2 billion
At Monday’s GTC 2026 conference in San Jose, Nvidia presented a comprehensive portfolio of processors and server architectures that signal the company’s expansion far beyond traditional GPU territory.
The star announcement centered on the Groq 3 language processing unit — abbreviated as LPU. Through a licensing agreement with Groq, Nvidia acquired technology and key personnel including founder Jonathan Ross and president Sunny Madra, completing a $20 billion transaction finalized last December.
Groq 3 targets inference workloads — the operational phase of AI following model training. Each interaction where users input prompts into chatbots and receive responses represents inference. This rapidly expanding AI market segment benefits from specialized silicon that can outperform multipurpose GPUs.
Nvidia VP of hyperscale and HPC Ian Buck explained that while Groq 3 features faster memory than the company’s GPUs, GPUs maintain larger memory capacity. The strategy involves leveraging both advantages simultaneously.
This approach materializes in the LPX server rack — a configuration containing 128 Groq 3 LPUs. When deployed alongside the Vera Rubin NVL72 rack, Nvidia claims customers achieve 35x superior throughput per megawatt and unlock 10x greater revenue potential. The system specifically targets trillion-parameter models and million-token context processing.
Vera CPU Enters Intel and AMD Territory
The second major announcement featured the Vera CPU rack. Previously, the Vera chip appeared only as a component of the Vera Rubin superchip — which integrated one Vera CPU with two Rubin GPUs. Nvidia now offers Vera as an independent processor solution.
The rack architecture integrates 256 liquid-cooled Vera processors into one unified system. Nvidia positions it as the optimal CPU for agentic AI — autonomous systems that navigate websites, extract information from documents, or execute complex multi-step workflows independently.
“We’ve designed a new kind of CPU, the Olympus core, engineered by NVIDIA for AI execution,” Buck said. Vera also plays a role in data mining, personalization, and context analysis that feeds into AI models.
This move positions Nvidia as a direct competitor to Intel and AMD in the data center CPU sector — a market these two companies have controlled for decades.
In the previous month, Nvidia secured an agreement with Meta for large-scale deployment of its Grace CPU predecessor — representing the largest deployment of its kind. The Vera announcement extends this strategic trajectory.
Additional Hardware Announcements at GTC
Nvidia’s presentation included the Bluefield-4 STX storage rack and Spectrum-6 SPX networking rack, completing a comprehensive data center hardware ecosystem.
Major hyperscalers like Amazon, Google, Meta, and Microsoft plan to invest a combined $650 billion in AI infrastructure throughout this year.
Nvidia’s data center division generated $193.5 billion in fiscal 2026, compared to $116.2 billion in the prior fiscal year.
Wall Street analysts maintain a consensus Strong Buy rating on NVDA, based on 38 Buy ratings and one Hold rating issued in the last three months, with a mean price target of $273.61.
Get 3 Free Stock Ebooks
Discover top-performing stocks in AI, Crypto, and Technology with expert analysis.
- Top 10 AI Stocks - Leading AI companies
- Top 10 Crypto Stocks - Blockchain leaders
- Top 10 Tech Stocks - Tech giants
