From the Token frenzy to a $1.3 trillion AI server blue ocean: AI inference computing power is exploding across the board, and leaders like Dell are entering a "volume + price + share" golden era.

Wall Street financial giant Goldman Sachs Group, Inc.'s latest research report, "2Q26 Server Market Share and Growth Outlook," shows that Dell Technologies, Inc. Class C and a host of other global leaders in AI server cluster manufacturing are simultaneously sharing in two growth dividends: "accelerating expansion of the AI server market size" and "a substantial increase in order share from core AI cloud computing customers such as large enterprises and neoclouds," rather than merely passively benefiting from hardware price increases. Under the 650 Group basis cited in the Goldman Sachs Group, Inc. research report, the statistics show that in the second quarter of 2026, Dell Technologies, Inc. Class C AI server revenue grew 146% year over year, with shipments up 80% and average selling price (ASP) up 37%, with revenue growth significantly above the industry's 94%; AI server revenue share rose from 13% to 17%, while traditional server share rose from 15% to 29%. Accordingly, Goldman Sachs Group, Inc. maintained its "Buy" ratingits most bullish ratingon Dell Technologies, Inc. Class C (DELL.US), with a 12-month target price of as high as $570, using an 18x forward earnings per share valuation; it also maintained "Buy" on Hewlett Packard Enterprise Co. (HPE.US) with a target price of $75, using a 14x forward earnings per share valuation; and maintained "Sell" on Super Micro Computer, Inc. (SMCI.US) with a target price of $34, using a 7.5x forward earnings per share valuation. None of the three target prices were adjusted in this report. Goldman Sachs Group, Inc. emphasized in the research report that Dell Technologies, Inc. Class C is one of the most prominent branded vendors in this round of server share expansion, Hewlett Packard Enterprise Co. retains allocation value as an enterprise infrastructure platform, and Super Micro Computer, Inc. needs to confront the competitive problem of relatively lagging AI business growth. The latest forecast data provided by Goldman Sachs Group, Inc. show that the overall server market size is expected to be about $1.5 trillion in 2030, of which the AI server market size will be about $1.3 trillion and traditional servers about $192 billion. By comparison, Goldman Sachs Group, Inc.'s 2026 market size expectations are about $370 billion for AI servers and about $137 billion for traditional servers. Goldman Sachs Group, Inc.'s AI server investment coordinates: rising volume and price are unstoppable, and share is increasingly concentrating among leaders Goldman Sachs Group, Inc.'s actual research and latest model estimates and expectations show that the core change at the global server industry level is that AI servers continue to grow in both volume and price, while traditional servers have also entered a stage of significant revenue expansion, with the 2030 market opportunity further revised upward accordingly. Under the 650 Group expectation basis cited in the Goldman Sachs Group, Inc. research report, second-quarter AI server revenue grew 94% year over year, driven jointly by a 20% increase in shipments and a 62% increase in ASP; IDC's accelerated server basis shows revenue growth of 43%, shipment growth of 10%, and ASP growth of 30%. On the traditional server side, 650 Group forecast basis data show revenue growth of 91%, significantly faster than 24% in the first quarter; although shipments fell 9%, ASP still rose 111%; IDC's non-accelerated server data show revenue growth of 81%, shipment growth of 17%, and ASP growth of 56%. Both statistical bases support strong revenue expansion, but they differ in their judgment on traditional server shipments. The ASP benchmark is an average selling price indicator and is not equivalent to a simple price increase for NVIDIA Corporation Blackwell or Rubin servers with the same configuration. More importantly, 650 Group raised its annual revenue forecasts for overall servers, AI servers, and traditional servers for 20262030 by an average of about 7%, 4%, and 17%, respectively, and expects the overall market size in 2030 to be about $1.5 trillion, of which AI servers will be about $1.3 trillion and traditional servers about $192 billion; revenue compound growth rates for 20252030 are 39%, 46%, and 17%, respectively, with the 46% for AI servers supported jointly by about 29% compound shipment growth and 13% compound ASP growth. In the view of the Goldman Sachs Group, Inc. analyst team, the expansion potential of the AI server market in which Dell Technologies, Inc. Class C operates is also very considerable: for secondary cloud and computing power leasing service providers, including a host of neocloud forces such as CoreWeave, the AI server market is expected to expand from $45.209 billion in 2025 to $560.536 billion in 2030, a compound growth rate of 65%; the market size for AI server cluster supply for enterprises is expected to expand from $14.430 billion to $91.472 billion, a compound growth rate of 45%; for hyperscale cloud computing vendors, the corresponding AI server market will reach about $646.737 billion, a compound growth rate of 37%. These latest statistics and model estimates all mean that AI server growth is spreading from a small number of hyperscale cloud service vendors to faster-growing neocloud platforms and a broad range of enterprise customers, and Dell Technologies, Inc. Class C has precisely established a strong share among these two types of customers. The key to the competitive landscape is not that all server vendors benefit simultaneously, but who can obtain a higher share among the customer groups with the largest incremental growth. Dell Technologies, Inc. Class C's revenue share in the neocloud AI server market rose from 47% to 51%, and in the enterprise AI server market from 30% to 46%; its traditional server share in the neocloud and enterprise markets also rose from 15% and 21% to 34% and 36%, respectively, with corresponding customer revenue growing 502% and 219%, respectively, reflecting cross-customer and cross-product expansion capability. 650 Group estimates and forecast data show that its calendar second-quarter traditional server revenue was $11.3 billion, up 278%; Dell Technologies, Inc. Class C's disclosed traditional server and networking revenue for the fiscal quarter ended July 31 was $10.5 billion, up 122%. Note that the differences between the two in fiscal-year statistical periods and business classification should be retained. On Hewlett Packard Enterprise Co., 650 Group forecasts show its traditional server revenue grew 75%, with share changing from 13% to 12%, but AI server revenue fell 10%, with shipments down 42% and ASP up 54%, and AI share falling from 3% to 1%; therefore, Goldman Sachs Group, Inc.'s "Buy" cannot be interpreted as an increase in AI server share this quarter. If supplemented with the company's own disclosed operating data, it is easier to understand the bullish support from Goldman Sachs Group, Inc. for AI server leader Hewlett Packard Enterprise Co.: its fiscal 2026 third-quarter networking revenue grew 74.9%, of which data center networking revenue grew 112.2%; the company also raised its fiscal 2027 revenue growth expectation to 13%17% and adjusted earnings per share growth expectation to 16%20%, indicating that networking and enterprise infrastructure businesses beyond servers are also contributing to growth. Super Micro Computer, Inc. shows clear divergence: traditional server revenue grew 356%, with share rising from 4% to 10%; although AI server revenue still grew 66% and shipments grew 59%, this was below the industry revenue growth rate estimated by 650 Group, with AI share falling from 9% to 8% and neocloud AI share falling from 29% to 25%. This provides a competitive basis for Goldman Sachs Group, Inc.'s cautious stance. From the Token explosion to complete-system delivery: the golden growth period for servers in the inference era The ability of large AI model applications to move toward the AGI era, the expansion of financing scale by Anthropic and OpenAI, the two leading AI application players, and computing power procurement together with increasingly strong AI chip revenue are forming a mutually reinforcing expansion chain, providing external demand corroboration for Goldman Sachs Group, Inc.'s server growth model. Astra, which sparked AGI discussion, strengthened programming, browsing, computer operation, and complex task execution capabilities in its official disclosure, further expanding the scope of work in which AI can participate; OpenAI also confirmed that starting September 10 it will suspend new registrations and upgrades for the $200-per-month Pro 20X plan, while existing subscriptions will not be affected. At the capital market level, OpenAI is considering a new funding round at a valuation of more than $1.2 trillion; Anthropic is reportedly seeking to advance a potentially record-breaking IPO at a valuation of about $2 trillion, raising up to $100 billion. Both should currently be understood as financing intentions or preparatory plans. More directly corresponding to demand in the AI server-related AI hardware chain than valuation figures is Anthropic's already announced capacity arrangements: an agreement with Amazon.com, Inc. for up to 5 gigawatts, an agreement with Alphabet Inc. Class C and Broadcom Inc. for 5 gigawatts starting online in 2027, another $30 billion in Azure computing power, a $50 billion U.S. AI infrastructure investment related to Fluidstack, and more than 300 megawatts of additional capacity for SpaceX Colossus 1. Operating growth data are equally strong at the level of AI chips, the most core upstream segment of the AI industry chain, and memory chip demand driven by the global AI infrastructure boom: NVIDIA Corporation's fiscal 2027 second-quarter data center revenue reached $89 billion, up 117% year over year, with overall revenue guidance for the next quarter of $108 billion; South Korea's August exports rose 68.7% to $98.26 billion, of which semiconductor exports reached $46.65 billion, more than triple the same period a year earlier, and exports from September 110 rose another 82.6% to about $34.97 billion. The bullish judgment supported jointly by these different links is that AI expansion is simultaneously reflected in improved application capabilities, long-term capacity procurement, and hardware revenue realization. What AI server vendors face is no longer just training cluster construction, but continuously expanding massive-scale AI inference service demand across industries. The simultaneous explosive expansion of Token demand from B-end/C-end users and enterprise-level memory chip capacity has, so to speak, extended this round of server growth from one-time equipment procurement into multi-year infrastructure upgrades. A recent research report released by another Wall Street financial giant, Citi, shows that the monthly compound growth rate of its Token usage tracking data reached 31%, with year-over-year growth of 2,434% in August 2026; Citi further forecasts that in 2027 HBM bit demand will grow 62%, server DRAM demand about 51%, and enterprise SSD demand 52.9%, and believes that continuous learning, personal AI, and physical AI may extend the memory supply-demand imbalance until 2031. Dell Technologies, Inc. Class C management also gave an outlook at its earnings call that inference Token demand will expand about 87 times by 2030, reaching 3.610 Tokens, and expects enterprise agents to become the largest single workload by 2028; industry models point to market expansion, customer orders improve revenue visibility, and realized AI-related revenue generation and profit prove that delivery is accelerating. Together, these three constitute the substantive basis for Dell Technologies, Inc. Class C entering a multi-year high-growth phase. From the underlying engineering logic, the benefit of the inference era for complete system suppliers such as Dell Technologies, Inc. Class C lies in the fact that every AI application moving into practical work requires coordinated expansion of compute, memory, storage, networking, and operations management. When an agent executes a task, it may repeatedly perform planning, retrieval, tool invocation, code execution, and result validation, with total demand driven simultaneously by active users, task frequency, number of model invocation rounds, and expanding context length. When a model processes input, the prefill stage has strong parallel computing demand; the decode stage that generates output is often more constrained by memory bandwidth, and the capacity of the key-value cache (KV Cache) also grows with concurrent requests and expanding context length. Therefore, higher inference throughput requires both accelerator computing capability and sufficient high-bandwidth memory (HBM), reasonable cache management, and efficient interconnection. Extending further from the complete system, CPUs handle agent orchestration, tool execution, and data processing; server DRAM and enterprise SSDs support knowledge bases, data access, and tiered caching; networking is responsible for connecting compute and storage resources; enterprises also need to deploy these components into a production environment capable of continuously providing services. The differentiation specifically emphasized by Dell Technologies, Inc. Class C management is precisely engineering design, global deployment, continuous support, and financing capability from order to operation, helping customers begin generating usable Tokens faster. This explains why AI inference expansion can simultaneously drive AI servers and traditional CPU servers: the former handles model computation, while the latter undertakes the applications and data work surrounding model operation. The dual-line share gains recorded by Goldman Sachs Group, Inc. for Dell Technologies, Inc. Class C in enterprise and neocloud precisely correspond to the commercial landing point of this kind of complete delivery capabilitythe growth opportunity for server leaders is upgrading from "selling more accelerator chassis" to "delivering more large-scale AI computing infrastructure clusters capable of continuously producing Tokens," driven jointly by market expansion, higher value per system, share growth, and accompanying services.