Industry News

The global AI server market is booming in 2026, driven by trillion-parameter models and large-scale computing demand. Flagship platforms including Huawei Atlas 950 SuperPoD, Dell PowerEdge, HPE ProLiant, and NVIDIA Vera Rubin lead the industry with revolutionary full liquid-cooling architecture, ultra-high computing density, and extreme energy efficiency. These next-generation AI servers deliver powerful performance for large model training, inference, and data center expansion, while setting new standards for scalability and stability. As core infrastructure for the digital economy, AI servers are reshaping cloud, enterprise, and industrial computing worldwide.

The global AI computing power explosion has triggered a memory shortage! The production capacity of HBM, DRAM and NAND has been locked up by giants, server storage costs have risen sharply, small and medium-sized data centers are under increasing pressure, and high-end storage has become a strategic resource for computing power.

On February 24, 2026, local time, U.S. Customs officially suspended additional tariffs imposed under IEEPA, following a final Supreme Court ruling that the tariffs were illegal. The $175 billion already collected will be fully refunded to importers. Improved customs clearance efficiency and reduced costs will significantly boost exports of IT hardware and network equipment, leading to a surge in U.S. procurement demand.

HPE has partnered with Juniper Networks to provide an end-to-end AI-native intelligent network solution for the 2026 Milano-Cortina Winter Olympics. Covering over 40 venues across 22,000 square kilometers, the solution deploys more than 1,500 enterprise-class switches and 4,900+ wireless access points, supported by high-performance network hardware. Powered by the HPE Mist AIOPs platform for automated operations and Juniper’s end-to-end security, it ensures stable, low-latency connectivity for event management, live broadcasts and more, setting a new benchmark for smart networks in large-scale international events. This collaboration also marks their first joint appearance at a top global sports event since HPE’s acquisition of Juniper, showcasing their integrated technical strength.

At the beginning of 2026, demand for AI servers surged, with Nvidia and Supermicro leading the market through their close collaboration, driving Foxconn's January revenue to a year-on-year increase of 35.5%. Due to supply-demand imbalances, Nvidia H100/H200 inference cards are experiencing price premiums exceeding 80%, and some models are out of stock. The RTX 5090Ti may be discontinued for the consumer market, only remaining available through enterprise channels, leading to significant fluctuations in the AI ​​hardware supply chain.

On February 3, 2026, Dell Technologies received positive news as Barclays upgraded its rating to "Overweight" with a target price of $148 due to strong performance in AI server orders. On the same day, Dell also upgraded its PowerStore storage system, adding AI-driven capacity forecasting, a 30%+ increase in rack density, and a native zero-trust architecture. This upgrade supports immutable data snapshots, further strengthening ransomware defense and improving enterprise-grade data infrastructure to meet the core needs of businesses in the AI ​​era.

HPE Juniper Networking has officially upgraded its MX204 router, adding 5G core network capabilities.  It supports FlexE slicing, SyncE, and IEEE 1588v2 PTP clock synchronization technologies.  With its compact 1U design and 400Gbps throughput, it is suitable for carrier edge nodes, 5G core network transport, and other scenarios. This upgrade allows for dynamic allocation of network resources, reduces deployment and maintenance costs, improves the efficiency of 5G service delivery, and is compatible with 5G-Advanced and 6G evolution requirements.  It is now available globally and has received pilot orders from international carriers.

Microsoft has launched its second-generation self-developed AI inference chip, the Maia 200.  Leveraging TSMC's 3nm process and exclusive supply of 216GB HBM3E memory from SK Hynix, it offers significantly improved computing power and energy efficiency.  The chip will primarily serve GPT-5.2, directly driving increased demand for AI server components such as HBM memory and MCIO cables.