--- title: "Guangfa Securities: NVIDIA's new platform enhances the competitiveness of Agent applications, with AI inference driving a sustained upward storage cycle" type: "News" locale: "zh-HK" url: "https://longbridge.com/zh-HK/news/279716435.md" description: "Guangfa Securities released a research report stating that NVIDIA's Vera Rubin POD platform showcased at GTC enhances the competitiveness of Agent applications and drives the upward development of AI inference-driven storage cycles. The platform includes a new rack-level system specifically designed for Agentic AI workloads, emphasizing high throughput and low latency inference capabilities. It is recommended to focus on the core beneficiaries in the industry chain" datetime: "2026-03-19T03:54:05.000Z" locales: - [zh-CN](https://longbridge.com/zh-CN/news/279716435.md) - [en](https://longbridge.com/en/news/279716435.md) - [zh-HK](https://longbridge.com/zh-HK/news/279716435.md) --- > 支持的語言: [简体中文](https://longbridge.com/zh-CN/news/279716435.md) | [English](https://longbridge.com/en/news/279716435.md) # Guangfa Securities: NVIDIA's new platform enhances the competitiveness of Agent applications, with AI inference driving a sustained upward storage cycle According to the Zhitong Finance APP, GF Securities has released a research report stating that NVIDIA (NVDA.US) showcased the Vera Rubin POD at GTC, focusing on enhancing the competitiveness of its cluster computing and inference computing product lines for Agent applications. As AI advances, model innovation and CAPEX lay the foundation for the collaborative development of the AI industry chain; AI inference drives a continuous upward cycle in storage, with expansion and upgrades working in tandem. It is recommended to pay attention to core beneficiaries in the industry chain. ## GF Securities' main viewpoints are as follows: **NVIDIA launches the Vera Rubin POD platform** According to NVIDIA's official website, on March 16, 2026, NVIDIA showcased the Vera Rubin POD at GTC, which includes five new rack-level systems specifically designed for Agentic AI workloads. Due to the higher demands of Agentic workloads for high throughput, ultra-low latency inference, dense CPU sandboxing, and large context memory storage, NVIDIA has focused on enhancing the competitiveness of its cluster computing and inference computing product lines for Agent applications. The Vera Rubin POD is mainly divided into two types of racks: (1) MGXNVL rack: the Vera Rubin NVL72, which is interconnected internally via NVLink and undertakes core GPU computing tasks; (2) MGXETL rack: including Groq3 LPX rack, Vera CPU rack, BlueField-4 STX storage rack, and Spectrum-6 SPX network rack, all of which are interconnected directly through SpectrumX Ethernet or Groq3 LPU chips for collaboration. According to the schematic diagram on the official website, a Vera Rubin 1152 SuperPOD consists of 16 Vera Rubin NVL72 racks, 2 Vera CPU racks, 10 Groq 3 LPX racks, 2 BlueField-4 STX storage racks, and 10 Spectrum-6 SPX network racks, reflecting its heterogeneous collaborative system architecture built around Agentic AI. **Groq3 LPX rack is used to accelerate decoding.** The Groq3 LPX rack integrates 256 LPU processors, equipped with 128 GB on-chip SRAM and a bandwidth of 640TB/s. In the combined architecture of Vera Rubin NVL72 and LPX, the GPU is primarily responsible for Attention calculations during the Prefill and Decode phases, while the LPU accelerates the FFN calculations during the Decode phase, speeding up the decoding process for each output token at each layer, and collaborating with the Vera Rubin racks through customized Spectrum-X interconnects. According to disclosures from NVIDIA's official website, under the condition of 400 TPS per user, the combination of Vera Rubin NVL72 and LPX achieves up to 35 times the TPS improvement per megawatt compared to NVIDIA GB200 NVL72, enhancing the overall output of the system while being more suitable for low-latency, highly interactive Agent application scenarios **Vera CPU Rack Supports RL/Agent Sandbox Environment** The Vera CPU rack integrates 256 Vera CPUs and adopts a high-density liquid cooling design. A single rack can support over 22,500 concurrent reinforcement learning (RL) or agent sandbox environments for testing, executing, and validating the output results of the Vera Rubin NVL72 and LPX. **Risk Warning** The development of the AI industry and demand may fall short of expectations; AI server shipments may not meet expectations, and the technological and product progress of domestic manufacturers may be below expectations ### 相關股票 - [NVIDIA (NVDA.US)](https://longbridge.com/zh-HK/quote/NVDA.US.md) - [Spdr Select Tech (XLK.US)](https://longbridge.com/zh-HK/quote/XLK.US.md) - [VanEck Semiconductor ETF (SMH.US)](https://longbridge.com/zh-HK/quote/SMH.US.md) - [Direxion Daily NVDA Bull 2X Shares (NVDU.US)](https://longbridge.com/zh-HK/quote/NVDU.US.md) - [T-Rex 2X Long NVIDIA Daily Target ETF (NVDX.US)](https://longbridge.com/zh-HK/quote/NVDX.US.md) - [Direxion Semicon Bull 3X (SOXL.US)](https://longbridge.com/zh-HK/quote/SOXL.US.md) - [iShares Semiconductor ETF (SOXX.US)](https://longbridge.com/zh-HK/quote/SOXX.US.md) - [iShares Expanded Tech Software Sector ETF (IGV.US)](https://longbridge.com/zh-HK/quote/IGV.US.md) - [GraniteShares 2x Long NVDA Daily ETF (NVDL.US)](https://longbridge.com/zh-HK/quote/NVDL.US.md) - [YieldMax NVDA Option Income Strategy ETF (NVDY.US)](https://longbridge.com/zh-HK/quote/NVDY.US.md) - [SPDR S&P Software (XSW.US)](https://longbridge.com/zh-HK/quote/XSW.US.md) - [XL2CSOPNVDA (07788.HK)](https://longbridge.com/zh-HK/quote/07788.HK.md) - [XI2CSOPNVDA (07388.HK)](https://longbridge.com/zh-HK/quote/07388.HK.md) - [Direxion Daily NVDA Bear 1X ETF (NVDD.US)](https://longbridge.com/zh-HK/quote/NVDD.US.md) - [T-Rex 2X Inverse NVIDIA Daily Target ETF (NVDQ.US)](https://longbridge.com/zh-HK/quote/NVDQ.US.md) ## 相關資訊與研究 - [Trump Gives Nvidia Investors Score Big Win](https://longbridge.com/zh-HK/news/279478186.md) - [BREAKINGVIEWS-Nvidia $1 trln goal creates more noise than signal](https://longbridge.com/zh-HK/news/279640359.md) - [Nvidia reopens China channel, but H200 flows remain policy-constrained](https://longbridge.com/zh-HK/news/279768278.md) - [NVIDIA Ignites the Next Industrial Revolution in Knowledge Work With Open Agent Development Platform | NVDA Stock News](https://longbridge.com/zh-HK/news/279318873.md) - [Nvidia Still Top AI Pick With Massive Upside Ahead: Analyst](https://longbridge.com/zh-HK/news/279764436.md)