--- title: "NVIDIA launches a new server cabinet that is compatible with competitors' AI chips" type: "News" locale: "en" url: "https://longbridge.com/en/news/280287675.md" description: "NVIDIA launched a new MGX ETL server cabinet at the GTC Developer Conference, which is compatible with competitors' AI chips and designed to accommodate various chips used by customers. This move opens up new opportunities for NVIDIA in the Chinese market, allowing enterprises to use domestic AI chips while running NVIDIA's software ecosystem. The MGX ETL cabinet is based on NVIDIA's Spectrum-X networking technology, supporting high-speed and stable chip interconnection, enhancing the compatibility and ease of use of the servers" datetime: "2026-03-24T09:16:14.000Z" locales: - [zh-CN](https://longbridge.com/zh-CN/news/280287675.md) - [en](https://longbridge.com/en/news/280287675.md) - [zh-HK](https://longbridge.com/zh-HK/news/280287675.md) --- > Supported Languages: [简体中文](https://longbridge.com/zh-CN/news/280287675.md) | [繁體中文](https://longbridge.com/zh-HK/news/280287675.md) # NVIDIA launches a new server cabinet that is compatible with competitors' AI chips NVIDIA faces increasingly fierce competition in the artificial intelligence chip market. However, this chip design company has found a new path to **make money even with competitors' chips**. At the annual GTC developer conference held last week, NVIDIA launched a new server cabinet designed to support the operation of NVIDIA's own chips **while also being compatible with its competitors' chips**. According to two insiders familiar with the project, the new cabinet is equipped with NVIDIA's networking technology, enabling high-speed and stable interconnection between the chips within the cabinet. This move indicates that NVIDIA is beginning to adapt to the chips from manufacturers like Google and AMD, which are increasingly used by its core customers. At the same time, it opens up potential new opportunities for NVIDIA in the Chinese market—Chinese companies can install domestic AI chips in this cabinet and run them on NVIDIA's software ecosystem, allowing NVIDIA to generate revenue without having to deal with chip export restrictions from both the U.S. and China. The new cabinet, named **MGX ETL**, follows the MGX cabinet architecture launched by NVIDIA in 2023, which has now become the standard design for its data center products. NVIDIA's design is also used in the Groq and Vera chip cabinets, with a single cabinet capable of accommodating up to 256 chips. ETL is not NVIDIA's first attempt to make servers compatible with competing chips, but it is the **easiest to use** solution for customers. Since last year, NVIDIA has launched a standalone solution called NVLink Fusion, allowing companies to integrate their own chips with NVIDIA's proprietary high-speed interconnect technology NVLink for ultra-fast communication between chips within the server. For example, Amazon has committed that the next-generation Trainium4 AI chip will support NVLink. Not all companies are willing to adopt NVLink. Over 115 companies, including AMD, Google, Meta, and Microsoft, support UALink technology—an alternative to NVLink that allows companies to adopt it without needing to apply for authorization or approval from NVIDIA. The ETL cabinet is different. It is based on NVIDIA's Spectrum-X networking technology, which uses Ethernet as its underlying protocol, a foundational protocol supported by almost all chips. Although to fully leverage Spectrum-X's performance, it still requires NVIDIA's self-developed switching chips and network cards, the barrier to using the ETL cabinet is much lower than that of NVLink. For NVIDIA, the core significance lies here: even as customers increasingly turn to competing chips, the ETL cabinet still allows NVIDIA to firmly occupy a central position in AI infrastructure A spokesperson for NVIDIA stated that MGX is an open rack design that does not restrict system partners from using other components. The company described this as not a new initiative and defined MGX ETL as an "open rack reference architecture," which is a modular system design that supports various computing and networking combinations. According to two informed sources involved in the discussions, some NVIDIA employees have introduced this new type of rack to customers. For example, NVIDIA has showcased the rack to some Chinese companies, which can accommodate a combination of domestic AI chips and those from manufacturers like AMD, while still operating based on NVIDIA's Spectrum-X network and software. This new rack also helps NVIDIA respond to related accusations — the company has been criticized for forcing customers to bundle purchases of chips and networking equipment, a practice that has caused dissatisfaction among large clients like Microsoft and has led to investigations by EU antitrust regulators. The networking equipment business has become an increasingly important component of NVIDIA's revenue. As of the quarter ending January this year, the networking business contributed over 15% of its total revenue, amounting to $11 billion, with a year-on-year growth rate of 268%, far exceeding the growth rates of any other publicly disclosed business segments of NVIDIA. American companies such as Google, Amazon, and metaverse platforms, as well as Chinese companies like Huawei, are increasingly pushing customers to procure or lease alternative chips, especially for high-speed AI products and agents. NVIDIA's move is a response to this trend. At last week's GTC conference, NVIDIA CEO Jensen Huang stated that the AI industry has shifted from the model training phase to the large-scale deployment phase of AI agents, which requires a different type of processing capability — faster and more cost-effective chips, along with large-capacity storage and memory hardware. He noted that over the past two years, "the computational demand for AI inference" (a type of computation for executing model logic to solve problems) "has grown about ten thousand times, with usage increasing about one hundred times," while overall computational demand has increased by about "one million times." ### Related Stocks - [NVIDIA Corporation (NVDA.US)](https://longbridge.com/en/quote/NVDA.US.md) - [Direxion Daily NVDA Bear 1X ETF (NVDD.US)](https://longbridge.com/en/quote/NVDD.US.md) - [T-REX 2X Inverse NVIDIA Daily Target ETF (NVDQ.US)](https://longbridge.com/en/quote/NVDQ.US.md) - [Direxion Daily NVDA Bull 2X Shares (NVDU.US)](https://longbridge.com/en/quote/NVDU.US.md) - [Direxion Daily Semicondct Bull 3X ETF (SOXL.US)](https://longbridge.com/en/quote/SOXL.US.md) - [State StreetSPDRS&PSftwr&SvcsETF (XSW.US)](https://longbridge.com/en/quote/XSW.US.md) - [VanEck Semiconductor ETF (SMH.US)](https://longbridge.com/en/quote/SMH.US.md) - [YieldMax NVDA Option Income Strategy ETF (NVDY.US)](https://longbridge.com/en/quote/NVDY.US.md) - [iShares Semiconductor ETF (SOXX.US)](https://longbridge.com/en/quote/SOXX.US.md) - [GraniteShares 2x Long NVDA Daily ETF (NVDL.US)](https://longbridge.com/en/quote/NVDL.US.md) - [iShares Expanded Tech-Software Sect ETF (IGV.US)](https://longbridge.com/en/quote/IGV.US.md) - [T-REX 2X Long NVIDIA Daily Target ETF (NVDX.US)](https://longbridge.com/en/quote/NVDX.US.md) - [GraniteShares 2x Short NVDA Daily ETF (NVD.US)](https://longbridge.com/en/quote/NVD.US.md) - [Tradr 1.5X Short NVDA Daily ETF (NVDS.US)](https://longbridge.com/en/quote/NVDS.US.md) - [The Technology Select Sector SPDR® ETF (XLK.US)](https://longbridge.com/en/quote/XLK.US.md) - [XL2CSOPNVDA (07788.HK)](https://longbridge.com/en/quote/07788.HK.md) - [XI2CSOPNVDA (07388.HK)](https://longbridge.com/en/quote/07388.HK.md) ## Related News & Research - [LPU: Nvidia popularizes a new AI acronym](https://longbridge.com/en/news/279662346.md) - [NVIDIA GTC Keynote: Huang Unveils “AI Factories,” Blackwell-Rubin Roadmap, and $1T Demand View](https://longbridge.com/en/news/280016667.md) - [Nvidia Asked by US Senators to Provide Details on Groq Deal](https://longbridge.com/en/news/279931631.md) - [Nvidia In Crosshairs As Senators Question China Export Controls](https://longbridge.com/en/news/280338466.md) - [Nvidia: Expanding AI Data Center Opportunity and Deepening Moat Support Long-Term Buy Thesis](https://longbridge.com/en/news/279596280.md)