Lenovo AI Server Backlog Hits $21 Billion as HBM Shortage Stalls China’s Compute Race - Tech Times

As China's artificial intelligence computing race accelerates, Lenovo's AI server backlog has surged to $21 billion, highlighting a global surge in demand for AI infrastructure. However, this growth i... Read original →

The HBM shortage has evolved from a supply chain bottleneck into a fulcrum of geopolitical competition in AI compute. Technically, the deep integration of 3D-stacked DRAM with EUV lithography concentrates HBM3e output among Samsung, SK Hynix, and Micron—severely limiting real-world deployment of NVIDIA’s Blackwell Ultra GPUs and forcing server vendors to redesign memory subsystems or delay DDR5 transitions. Compliance-wise, China’s Xinchuang initiative boosts domestic AI server demand, yet the National Intelligence Law raises due diligence costs for multinational buyers evaluating Lenovo infrastructure, eroding its global competitiveness. Competitively, Inspur and H3C are fast-tracking partnerships with CXMT for localized memory solutions, while Dell and HPE secure exclusive deals with Micron, deepening market fragmentation. Over the next 18 months, prolonged HBM lead times will stretch AI server delivery beyond nine months, triggering 'GPU idling'—chips ready but unshippable due to memory shortages—and ultimately accelerating China’s push for sovereign 2.5D/3D advanced packaging capabilities as an alternative technical pathway.