AI server shipments surge 65% year-on-year, Dell and Supermicro Computer lead global computing infrastructure

On August 1, 2026, global AI computing power market data exploded again. The latest report from market research institution Counterpoint Research shows that the global shipment volume of AI servers grew by 65% year-over-year, a 20% increase compared to the previous quarter, setting a new historical high. This figure far exceeds market expectations, once again confirming that the comprehensive deployment of large AI models, from training to inference, is driving explosive growth in computing infrastructure. ## AI Servers Become the Core Pillar of Computing Investment Unlike general-purpose servers, AI servers are equipped with accelerator chips such as GPUs and ASICs, designed for high parallel computing. Their average price is 8 to 10 times that of traditional servers. This quarter, AI servers accounted for over 30% of total global server shipments, and revenue share exceeded 70%. Among these, GPU servers accounted for over 80% of AI server shipments, with NVIDIA remaining the absolute leader, but the market share of AMD Instinct series and custom ASICs is steadily increasing. Counterpoint analysts point out that the strong growth in AI server demand stems from three levels: first, large cloud vendors are continuously increasing their capital expenditure on AI computing power; second, enterprise customers are shifting from pilot projects to full deployment; and third, policy pushes for infrastructure construction in various countries. This structural demand makes AI servers the most certain investment line in the entire computing chain. ## Dell and Supermicrocomputer Lead the Way In terms of vendor performance, Dell holds a market share of 18% in the global AI server market, closely followed by Supermicrocomputer with 15%. Recent earnings reports from both companies show that backlogged orders for AI servers have set a record, with delivery cycles extending to 2027. Dell dominates the enterprise AI market due to its global supply chain advantages and strong channel capabilities. Its AI server business revenue for the quarter first exceeded $10 billion, a year-over-year increase of over 80%. The Chief Operating Officer of Dell stated in an earnings call that AI server demand shows an "insatiable" trend, with customers expanding from leading cloud vendors to traditional manufacturing, finance, and healthcare industries. Supermicrocomputer continues to deepen its ties with NVIDIA. Its GPU systems are welcomed by large data centers due to their high-density liquid cooling solutions and rapid iteration capabilities. Supermicrocomputer shipped over 100,000 units this quarter, with HGX servers based on the NVIDIA Blackwell Ultra platform accounting for over half. Notably, Supermicrocomputer's gross margin increased by 2.3 percentage points driven by high-end products, indicating improving profitability in the AI server market. ## Cloud Vendors Unleash Capital Expenditure Power The surge in AI server shipments is inseparable from the massive investment by cloud vendors. Statistics show that in the second quarter of 2026, the combined capital expenditure of the four tech giants Microsoft, Google, Amazon, and Meta exceeded $60 billion, a year-over-year increase of about 40%, with the vast majority invested in AI servers and data center infrastructure. Microsoft's capital expenditure exceeded $18 billion this quarter, with over 70% allocated to AI-related projects. Google followed closely with $17 billion in capital expenditure, focusing on expanding its TPU cluster scale. Amazon AWS and Meta's capital expenditures were $16 billion and $9 billion respectively, both targeting large model training and inference as core investment areas. This capital expenditure intensity is expected to continue at least until 2027. Cloud vendors generally believe that AI computing power is the core engine for the next round of growth, and it is still in the early stage of "investment for growth." As AI application scenarios expand from chatbots to enterprise software, autonomous driving, and robotics, the compound annual growth rate of computing power demand is expected to remain above 60%. ##Synergy in Industry Chain Benefits, Explosion in Cooling and Interconnection Demand The soaring power consumption and performance of AI servers have also driven the comprehensive upgrade of supporting industries. This quarter, the penetration rate of liquid cooling solutions in AI servers has rapidly increased from 10% in the previous year to 25%. NVIDIA's next-generation GB300 NVL72 system form factor requires all-liquid cooling design, making liquid cooling the standard configuration for high-end AI servers. Vertiv Holdings, the leader in liquid cooling in US stocks, saw a year-on-year order increase of 110% this quarter, and its heatsinks for AI data centers developed specifically for AI have received certifications from multiple cloud vendors. Core components in the liquid cooling chain, such as quick connectors and CDUs (Coolant Distribution Units) are in short supply, and domestic suppliers like CSL and CoolIT are accelerating production expansion. Meanwhile, the demand for high-speed interconnects in AI servers is also surging. Due to the need for high-bandwidth, low-latency network interconnection among multi-GPU clusters, 800G optical modules have entered the mass deployment stage, expected to transition to 1.6T by 2027. Market research data from LightCounting shows that optical module revenue in data centers grew by 85% year-on-year in the second quarter, with Coherent and Lumentum being the biggest beneficiaries. Niche areas such as high-speed copper cables and AEC (Active Electrical Cables) are also showing explosive growth. ##Risks and Challenges: Supply Chain Bottlenecks Remain Despite the bright outlook for AI server shipments, supply chain bottlenecks cannot be ignored. HBM (High Bandwidth Memory) and advanced packaging capacity are the biggest constraints for this round of AI computing power expansion. The HBM capacity of SK Hynix, Samsung, and Micron has all been pre-ordered by NVIDIA and AMD, and the yield climb for HBM4 in the second half of 2026 remains uncertain. In terms of advanced packaging, While TSMC's CoWoS capacity is doubling, it still cannot fully meet the surging demand. This leads to delays in AI server shipments, potentially affecting the product iteration pace of downstream customers. Furthermore, there is a shortage in the supply of power management chips and passive components like MLCC. Counterpoint analyst stated: "The AI server market is in a cycle of 'demand explosion - supply tightness - capacity expansion - renewed explosion.' Although there are bottlenecks in the short term, in the long term, the investment cycle for AI computing infrastructure is far from over." He predicts that the AI server market size will exceed $150 billion in 2026 and double by 2028. ##Investment Perspective: Capturing the Certainty Opportunities in Computing Power Infrastructure For investors focusing on the US stock AI computing power sector, the sustained high growth in AI server shipments provides solid performance support for upstream chip, component, and system manufacturers. In addition to Dell and Supermicrocomputer, investors should also watch core suppliers such as NVIDIA, AMD, and TSMC. Furthermore, the elasticity in "shovel-moving" segments such as liquid cooling, optical modules, and power management should not be overlooked. Any investment decision should be based on a thorough risk assessment. Although the AI server industry is experiencing very high growth, there are risks such as changes in technological routes, fluctuations in demand rhythm, and overvaluation. Investors are advised to focus on leading companies with core technology and stable customer relationships, and to closely monitor cloud vendor capital expenditures and the progress of AI commercialization.