From the Token Frenzy to the $1.3 Trillion AI Server Blue Ocean: AI Inference Computing Power Fully Explodes, Industry Leaders like Dell Usher in the Golden Era of "Volume, Price, and Market Share"

According to reports from Zhicheng Finance APP, Wall Street financial giant Goldman Sachs has just released its "2Q26 Server Market Share and Growth Outlook" research report, which reveals that leading global AI server cluster manufacturers such as Dell are simultaneously benefiting from two sources of accelerated growth: the "rapid market expansion of AI servers" and the "substantial increase in order share from core AI cloud computing customers including large enterprises and new cloud providers." This growth is not merely a passive effect of hardware price increases.

Goldman's research, citing 650 Group data, shows Dell's AI server revenue is expected to grow 146% year-on-year in Q2 2026, with shipment volume up 80% and average selling price (ASP) up 37%, significantly outperforming the industry growth rate of 94%. Dell's AI server revenue share is set to rise from 13% to 17%, while traditional servers will see their share rise from 15% to 29%.

Based on this, Goldman Sachs maintains a "Buy" rating, the most bullish, for Dell (DELL.US), with a 12-month target price of $570 using an 18x forward earnings multiple. It also maintains "Buy" ratings for Hewlett Packard Enterprise (HPE.US) with a $75 target and a 14x forward earnings multiple; and for Super Micro Computer (SMCI.US), it holds a "Sell" rating with a $34 target, applying a 7.5x forward earning multiple. None of the price targets were adjusted in this report.

Goldman's research emphasizes that Dell is one of the most prominent brands in expanding server market share in this cycle. HPE retains value through its enterprise infrastructure platform, while Super Micro Computer faces competition due to relatively slower growth in AI business. Goldman's latest forecasts show that the overall server market will reach approximately $1.5 trillion by 2030, with the AI server segment at about $1.3 trillion and traditional servers around $192 billion. For comparison, Goldman expects the market size in 2026 to be...

Goldman's AI Server Investment Coordinates: Unstoppable Momentum in Price & Volume, and Market Share Further Concentrated Among Leaders

Goldman's actual research and latest modeling expect that the core change in the global server industry is AI servers sustaining growth in both price and volume, with traditional servers also entering a stage of significant revenue expansion; thus, the 2030 market space has been revised upward accordingly.

Citing 650 Group's projections, Goldman's research indicates that AI server revenue in Q2 grew 94% year-on-year, driven by 20% shipment growth and a 62% rise in ASP; IDC's "accelerated server" category shows revenue up 43%, shipments up 10%, and ASP up 30%. For traditional servers, 650 Group estimates a 91% revenue rise (up notably from 24% in Q1), despite a 9% drop in shipments, with ASP soaring 111%; IDC's non-accelerated server data points to 81% revenue growth, 17% shipment growth, and a 56% ASP rise. Both methodologies support robust revenue expansion, though they differ on traditional server shipment outlooks. The ASP is a benchmark for average selling price and does not represent a simple price increase for identical Nvidia Blackwell or Rubin servers.

More importantly, 650 Group's predictions have raised the 2026–2030 annual revenue forecasts for total servers, AI servers, and traditional servers by around 7%, 4%, and 17% on average. By 2030, the total market is expected at $1.5 trillion, with AI servers at $1.3 trillion and traditional servers at $192 billion. From 2025–2030, compound annual revenue growth rates are expected to be 39%, 46%, and 17%, with AI server growth underpinned by a 29% CAGR in shipments and 13% in ASP.

From Goldman analysts' perspective, the expansion potential of Dell's AI server market is very considerable: Secondary cloud and compute leasing providers—including the likes of CoreWeave and other NeoClouds—are expected to see their AI server market grow from $45.209 billion in 2025 to $560.536 billion in 2030, a CAGR of 65%. The enterprise-facing AI server cluster supply market is forecast to expand from $14.430 billion to $91.472 billion (CAGR 45%), while hyperscale cloud computing vendors are projected to reach approximately $646.737 billion at a 37% CAGR. These latest statistics and model estimates indicate that the growth in AI servers is diffusing from a handful of hyperscale cloud providers to faster-growing new cloud platforms and a wide swath of enterprise customers, in both of which Dell has established considerable market share.

The key competitive dynamic is not that all server manufacturers benefit equally, but rather who can capture a larger share among the fastest-growing client groups. Dell's revenue share in the new cloud AI server market has increased from 47% to 51%, and in the enterprise AI server market from 30% to 46%. Market share for traditional servers in the new cloud and enterprise markets also rose from 15% and 21% to 34% and 36%, respectively, with corresponding client revenues up 502% and 219%, demonstrating cross-client and cross-product expansion capability. According to 650 Group's calculations, Dell's traditional server revenue in calendar Q2 rose to $11.3 billion (+278%); for the fiscal quarter ending July 31, Dell reported traditional server and networking revenue of $10.5 billion (+122%). Note the difference in reporting periods and business classifications between the two sets of data.

For HPE, 650 Group forecasts traditional server revenue up 75%, with share declining from 13% to 12%, but AI server revenue dropping 10%, shipments down 42%, and ASP up 54%; AI server market share decreases from 3% to 1%. Therefore, Goldman's "Buy" rating should not be interpreted as an immediate improvement in AI server market share this quarter. Supplementing with HPE's own disclosures, it's clearer why Goldman remains bullish: For the third quarter of fiscal 2026, networking revenue rose 74.9%, including a 112.2% increase in data center networking; the company also raised its fiscal 2027 revenue growth forecast to 13–17%, and adjusted EPS growth to 16–20%. This implies that business growth is being contributed not only by servers but also by networking and enterprise infrastructure.

Super Micro Computer exhibits stark divergence: traditional server revenue is up 356%, with share growing from 4% to 10%; AI server revenue is up 66% and shipments 59%, but this lags industry AI revenue growth as estimated by 650 Group. Its AI share drops from 9% to 8%, and new cloud AI share from 29% to 25%, providing a competitive rationale for Goldman's conservative stance.

From Token Boom to Full System Delivery: The Golden Growth Era for Servers in the Inference Age

As AI large models move toward AGI, with leading AI application firms like Anthropic and OpenAI expanding their fundraising and accompanied by surging compute procurement and ever-stronger AI chip revenues, a mutually reinforcing expansion chain is forming, providing external demand validation for Goldman's server growth model.

Astra, which has triggered AGI discussions, has officially enhanced programming, browsing, computer operation, and execution of complex tasks, further broadening the practical applications of AI; OpenAI also confirmed that, starting September 10, new registrations and upgrades for its $200/month Pro 20X plan are paused, though current subscriptions remain unaffected. In capital markets, OpenAI is considering a new financing round at a valuation exceeding $1.2 trillion; Anthropic is reportedly seeking up to $100 billion at a $2 trillion valuation, possibly in preparation for a record IPO. Both should be understood as indicative or preparatory financing stages.

Even more directly relevant to demand for the AI hardware chain linked to AI servers are Anthropic's announced capacity arrangements: agreements for up to 5GW with Amazon, further 5GW deals with Google and Broadcom starting in 2027, another $30 billion in Azure compute, $50 billion in U.S. AI infrastructure investment related to Fluidstack, and SpaceX Colossus 1 bringing over 300MW of additional capacity.

Upstream, operating data on core AI chips and booming storage chip demand driven by the AI infrastructure frenzy are also compelling: Nvidia's Q2 FY2027 datacenter revenue hit $89 billion, up 117% year-on-year, with next quarter guidance set at $108 billion. South Korea's August exports climbed 68.7% to $98.26 billion, with semiconductor exports at $46.65 billion—more than triple last year's level—while September 1–10 exports rose 82.6% to roughly $34.97 billion. All links in the chain support a bullish view: AI expansion is simultaneously reflected in upgraded application capabilities, long-term capacity procurement, and the realization of hardware revenues. AI server manufacturers are now facing demand not just for training clusters, but for ever-growing, industry-wide inference service loads.

B2C and B2B users' surging demand for Tokens and enterprise-level storage chip capacity has turned this server boom from a one-off hardware refresh into a multi-year infrastructure upgrade cycle. Another Wall Street giant, Citi, recently reported that its tracked Token usage shows a 31% monthly compound growth rate, with a 2,434% year-on-year increase projected for August 2026. Citi further predicts 62% growth in HBM demand, 51% in server DRAM, and 52.9% in enterprise SSD demand by 2027, suggesting sustained imbalances in storage supply and demand may persist through 2031 due to continual learning, personal AI, and embodied AI.

Dell's management shared on its earnings call that inference Token demand is expected to multiply 87-fold to about 3.6×10¹⁸ Tokens by 2030, with enterprise agents slated to become the single largest workload by 2028. Industry models point to market expansion, more visible customer orders drive revenue, and proven AI-related monetization is accelerating—all substantiating Dell's entry into a multi-year high-growth phase.

From an engineering perspective, the rise of inference benefits integrated system providers like Dell, as every practical AI application requires coordinated expansions of compute, memory, storage, networking, and operations management. When an agent performs a task, it may repeatedly plan, retrieve, invoke tools, run code, and verify outputs. Overall need is driven by active users, task frequency, model call cycles, and expanding context length. During input processing, the Prefill stage demands extensive parallel computation, while the Decode stage is often limited by memory bandwidth, with KV Cache capacity increasing with concurrent requests and longer context windows.

Thus, higher inference throughput requires not only accelerator compute power but also ample HBM (high-bandwidth memory), efficient cache management, and robust interconnects. Looking at complete systems: CPUs handle agent orchestration, tool execution, and data processing; server DRAM and enterprise SSDs support knowledge base, data access, and tiered caching; and networking links compute and storage resources. Enterprises need to deploy these components into production environments that can provide continuous service. Dell management especially highlights differentiation in engineering design, global deployment, sustained support, and financing from order to ramp-up, all helping clients generate usable Tokens faster.

This explains why AI inference expansion is driving both AI servers and traditional CPU servers: the former provide model computation, while the latter support applications and data operations around the models. The dual-track market share gains in enterprises and new cloud reflected in Goldman's data perfectly correspond to this comprehensive delivery capability—server leaders are growing not just by "selling more accelerator chassis," but by "delivering larger-scale AI compute infrastructure clusters capable of consistently generating Tokens," driven by market expansion, higher system value, share gains, and supporting services.