NVIDIA Corporation (NVDA.US) tops the data center Ethernet ranking, revealing investment insights: the frenzy of computing power is far from over, and GPUs no longer dominate the AI bull market.

The latest research report released by the globally renowned research institution IDC shows that the world's largest market cap company - the AI chip superpower NVIDIA Corporation (NVDA.US) has become the largest supplier of global data center Ethernet switches for the first time, measured by revenue. According to IDC's Quarterly Ethernet Switch Tracker report, in the first quarter of 2026, the global Ethernet switch market size increased by 39.8% year-on-year, reaching $15.4 billion. IDC's latest report is consistent with the views of Wall Street giants such as Morgan Stanley, Goldman Sachs Group, Inc., and Bank of America Corp, that leaders in the AI computing industry like NVIDIA Corporation are expanding their computing power chain coverage from "GPU/AI chip single point control" to "GPU cluster + network + DPU + optical interconnect system + software ecosystem" and other aspects of the AI factory system-level closed loop. The research institution pointed out in the report that as the construction of global AI computing infrastructure progresses rapidly, there is a collective surge in the complete AI computing demand associated with AI data center deployments, especially in the sub-markets covering high-speed switch infrastructure systems within ultra-large scale cloud computing and enterprise data centers. Driven by investments in AI computing infrastructure for large-scale AI inference and training workloads, it soared by 61% year-on-year, reaching $10 billion, occupying the vast majority of the $15.4 billion market size. Morgan Stanley stated that the AI computing arms race is entering a phase of system-level expansion. The institution's capital expenditure expectations for major U.S. technology giants in 2026 have been significantly raised from $433 billion a year ago to $805 billion, with the possibility of reaching $1.1 trillion in capital expenditure in 2027, compared to the previous estimate of $950 billion. Morgan Stanley's latest expectations highlight the bottlenecks in the AI computing infrastructure supply chain shifting from "large-scale purchases of GPUs/ASICs" to "simultaneously addressing data center power equipment, liquid cooling, data center CPUs, DRAM/NAND/HBM, data center optical communication/optical interconnect, high-performance Ethernet network infrastructure, transformers, gas turbines, and the entire set of AI data center deployment processes". According to IDC's detailed calculations, in the first quarter of 2026, the total revenue of global enterprise-level campus and branch office Ethernet switches increased by 12.3% year-on-year, reaching $5.4 billion. In a milestone shift, NVIDIA Corporation has become the largest supplier in the data center Ethernet switch market, measured by total revenue for the first time. IDC added that with hardware refresh cycles and support from rising component prices, the exchange business in campus and branch offices grew by 12.3% year-on-year. IDC data shows that with a 192.7% year-on-year revenue increase to $21 billion in the latest quarter reported by CKH HOLDINGS, NVIDIA Corporation's Spectrum-X platform, through its integrated AI GPU cluster and network infrastructure, has captured significant demand for AI factory-type network infrastructure from ultra-large scale cloud providers and large enterprises. The research institution noted that this structural shift is reshaping the supplier rankings in the data center networking industry. According to IDC, NVIDIA Corporation became the largest supplier in the data center Ethernet switch market in the first quarter of 2026, signaling a transformation in data center networking and reflecting the increasing influence of the latest AI computing infrastructure clusters on procurement decisions. Buyers focusing on AI factory-type campuses are beginning to factor in the impact of sustained average selling price (ASP) increases when planning budget updates. "NVIDIA Corporation's rise to number one in the data center Ethernet switch market in just one year is one of the most significant supplier landscape changes IDC has tracked in the enterprise networking space. The integrated design of Spectrum-X leading AI GPU with network infrastructure is winning many AI factory deals, a feat that current network suppliers cannot match with independent hardware alone. On the campus side, there is another, equally important story to tell: the Ethernet network infrastructure upgrade wave is real, but IT teams should plan for ASP normalization once memory supply constraints ease. Budget for this transition now, rather than waiting for price changes," said Paul Nicholson, IDC's vice president of cloud and data center networking research. While the core function of a switch relies on a "switch chip" for rapid packet forwarding, without storage chips providing software guidance, configuration storage, address addressing, and data buffering workloads in the background, the switch will not function at all. Buffer memory is typically the most critical storage determining switch performance, often being extremely high-speed SRAM or DRAM. IDC pointed out that large-scale AI inference computing deployments are accelerating in ultra-large cloud providers and large organizations, mainly used to enhance customer experiences, reduce operational risks, and empower critical business areas including IT infrastructure and operations, software development, and sales. According to IDC's report, AI workloads ranging from large-scale training clusters to enterprise edge inference are widely adopted, driving the continuous demand for high-speed, low-latency data center network infrastructure. Spectrum-X is completely reshaping the data center network infrastructure landscape, extending the narrative of the AI computing arms race from GPUs to switches in the AI power domain IDC projects that under the continued push of massive AI computing infrastructure investments by ultra-large cloud providers and enterprises, the Ethernet switch market will maintain strong growth momentum in 2026. The research institution states that with the expansion of inference deployments and training workloads, the strong demand for high-speed data center switches, especially those with 800G and higher rates, is expected to continue. IDC added that NVIDIA Corporation's position will face increasingly strong competitive responses from long-time Ethernet ecosystem providers such as Cisco Systems, Inc., Arista, and Broadcom Inc., making the data center sub-market one of the most active segments in the networking field. In the campus and branch office sector, IDC expects update cycles to continue to play out strongly, but if memory/storage chip supply constraints ease significantly and weaken ASP pricing tailwinds, overall revenue growth may slow. IDC also pointed out that major macroeconomic uncertainties, including tariff risks and regional economic fluctuations, remain factors that need close monitoring and could inhibit investment decisions in certain regions. Below is the latest compilation of the global data center Ethernet switch market competition landscape by IDC: NVIDIA Corporation (NVDA.US) IDC states that NVIDIA Corporation's Ethernet switch revenue all comes from the data center sub-market, with crazy year-on-year growth of 192.7% in the first quarter of 2026, reaching $2.1 billion, giving it a solid 21.5% share in the data center sub-market. The company's exclusive Spectrum-X high-performance Ethernet platform is an end-to-end AI data center high-performance network solution that combines Spectrum Ethernet switches with BlueField DPU and NVIDIA LinkX cable systems, designed specifically for NVIDIA's large-scale AI GPU computing clusters. It has become the preferred network interconnection solution for large-scale AI training and inference clusters, gaining significant configuration appeal among ultra-large cloud computing providers and AI-native cloud service providers leading the construction of AI factories. Arista Networks (ANET.US) Arista's Ethernet switch revenue data shows that 92% comes from the data center sub-market. Its high-performance Ethernet-related business revenue grew by 37.3% year-on-year in the first quarter of 2026, reaching $2.2 billion. Compared to NVIDIA Corporation's single Ethernet switch revenue, Arista's total Ethernet revenue is slightly higher at about $2.2 billion, but 92% of it comes from the data center, which is around $2.02 billion in data center Ethernet-related revenue. According to IDC, Arista has a 14.6% share of the overall Ethernet switch market and a 20.7% share in the data center base segment, maintaining strong growth positioning in 400G and 800G deployments for ultra-large cloud customers. Cisco Systems, Inc. (CSCO.US) IDC states that Cisco Systems, Inc. saw a 24% year-on-year growth in Ethernet switch-related total revenue in the first quarter of 2026, gaining a 29.3% market share. Non-data center revenues accounted for 60.5% of Cisco Systems, Inc.'s total revenue and grew by 14.1% year-on-year, reflecting the rebounding network infrastructure renewal in data center campuses; while data center-related revenue saw a substantial 43% year-on-year growth under the strong demand for AI computing infrastructure. According to IDC, Cisco Systems, Inc.'s router-related total revenue grew by 24.4% in the first quarter of 2026, giving the company a market share of up to 35.1%. Hewlett Packard Enterprise Co. (HPE.US) IDC points out that Hewlett Packard Enterprise Co.'s total Ethernet switch revenue system is 70.5% from the non-data center sub-market. The company's total Ethernet switch-related revenue grew by 15.4% year-on-year in the first quarter of 2026, with a market share of 6.4%. Following the completion of all acquisitions in July 2025, Hewlett Packard Enterprise Co.'s total revenue data now includes Juniper Networks, Inc. According to IDC's statistics, the company's deepened campus and branch office product portfolio system is transforming the current wave of upgrades into strong revenue growth. Huawei According to IDC's latest statistical data, Chinese tech giant Huawei saw a significant 27.2% year-on-year growth in total revenue related to Ethernet switches in the first quarter of 2026, reaching $895 million, giving the company a 5.8% market share. IDC points out that Huawei's router-related revenue saw a modest growth of 0.8% in the first quarter of 2026, with a market share of approximately 25.4%, highlighting its sustained strength in the service provider network sub-market, especially in China and some emerging markets. From GPUs to Ethernet switches, the process of constructing AI factories progresses rapidly: under Goldman Sachs Group, Inc.'s heavy asset logic for AI infrastructure construction, the bull market in the computing power chain is far from over As shown in the IDC research report, leaders in the AI computing power industry like NVIDIA Corporation are expanding the coverage of computing power chains from "GPU/AI chip single point control" to "GPU + network + DPU + optical interconnect system + software" and other aspects of the AI factory system-level closed loop. The bottleneck of AI training and inference clusters is no longer just about "buying more GPUs", but about whether GPUs can synchronize, transfer parameters, schedule inferences, and isolate tenants between GPUs with low latency, high bandwidth, low packet loss, and high utilization network to complete. In other words, NVIDIA Corporation's rise to the top of the data center Ethernet switch market demonstrates that the "crazy surge in computing power chain demand" has expanded from AI GPU and AI ASIC/TPU computing power clusters to the entire AI power chain system, including HBM, advanced packaging systems, high-performance Ethernet network infrastructure, optical interconnect systems, DPUs, data center cables, ABF/glass substrates, switch chips, and even gas turbines, power management systems, and other components of the entire AI power chain. Goldman Sachs Group, Inc.'s benchmark framework estimates that the investment around AI CapEx (namely capital expenditures related to AI) is no longer focused on large purchases of NVIDIA Corporation's Blackwell/Rubin AI GPU computing clusters, but covers the entire system of AI infrastructure assets including data center power equipment, liquid cooling, data center CPUs, DRAM/NAND/HBM storage, optical communication/optical interconnect in data centers, high-performance Ethernet network infrastructure/data center DCI high-speed interconnection, transformers, gas turbines, and other components, representing a systematic AI capital expenditure. NVIDIA Corporation CEO Jensen Huang mentioned on Wednesday that AI could usher in a new era of growth for American manufacturing and industry. According to Goldman Sachs Group, Inc., the ultra-large cloud providers' investment in 2026 is expected to reach $770 billion, close to their entire operating cash flow. The forecast for cumulative capital expenditure on AI infrastructure from 2026 to 2031 is about $76 trillion, with annual AI CapEx of at least $765 billion in 2026, rising to about $1.6 trillion by 2031. In the view of the Goldman Sachs Group, Inc. analyst team, the capital market pricing logic of the AI wave is shifting from "who can write the most powerful AI models/AI applications software" to "who can quickly build AI computing clusters, achieve large-scale power supply, large-scale cooling, accelerate the advancement of data center internal optical interconnect and data center-to-DCI between centers and continuously iterate the next generation of AI factories". In the eyes of Goldman Sachs Group, Inc. on Wall Street, the global bull market around the AI computing power chain is far from over, and the market theme has shifted from the long-term "programming/code-driven software light asset software valuation expansion" since 2008 to "re-pricing the AI computing infrastructure around a series of physical assets". The latest judgment of Goldman Sachs Group, Inc. implies that the next round of excess alpha returns will no longer just belong to the AI GPU/AI ASIC domain leaders but will systematically spread to high-performance CPUs, DRAM/NAND/HBM storage, AI printed circuit boards, liquid cooling systems, data center optical interconnect systems, ABF/Glass substrates, MLCCs, electronic cloth, and wide semiconductor foundries, all components of the full-stack AI computing power infrastructure layer. Among the Wall Street analyst community, MarketBeat shows that 54 senior Wall Street analysts have an average target price of $305.67 for NVIDIA Corporation over 12 months, with a highest target price of $500; TipRanks shows that 37 senior analysts have an average target price of $311.41, with a highest of $500; Investing.com displays that 59 analysts have an average target price of $298.93, with a highest of $500. The average target price of $305.67 corresponds to a market value of approximately $7.43 trillion; while estimated at $500, it would correspond to a market value of around $12.15 trillion, compared to NVIDIA Corporation trading near $207 in pre-market trading, with a market value hovering around $4.95 trillion.