Dell Forecasts Tighter HBM and DRAM Supply in 2027 as AI Data-Center Demand Rises
Редакция ЦОД:Главное
Dell expects shortages of high-bandwidth memory (HBM) and conventional DRAM to worsen in 2027, driven by the memory requirements of AI data-center buildouts. The company says constrained memory availability is already limiting the pace at which it can fulfill and ship an AI-server backlog of about US$95 billion.
- Dell expects HBM and conventional DRAM shortages to intensify in 2027.
- Dell reports an AI-server backlog of approximately US$95 billion.
- Memory availability is limiting Dell’s server fulfillment and shipment pace.
- HBM and DRAM constraints could delay AI infrastructure deployment even if processors and accelerators are available.
- Dell expects some new semiconductor fabs may not reach full production until 2029–2030.
Dell Chief Executive Michael Dell has forecast a more severe shortage of HBM and conventional DRAM in 2027, attributing the pressure to heavy memory consumption from AI data-center construction.
The company reports an accumulated AI-server order backlog of approximately US$95 billion. Memory availability is constraining the rate at which those orders can be executed and shipped, indicating that supply limitations extend beyond access to compute processors and accelerators.
HBM is a critical component for AI accelerator systems, while conventional DRAM is required across the server infrastructure surrounding those accelerators. For data-center operators and AI infrastructure providers, this creates a potential deployment bottleneck: a system may be constrained by memory supply even when other major hardware components are available.
Dell also expects broader supply tightness to persist for several years. According to the company’s assessment, some new semiconductor fabs may not reach full production until 2029–2030. This is an expected timeline rather than a confirmed outcome, and no specific manufacturers or facilities were identified.
For the data-center market, sustained HBM and DRAM constraints could slow AI-server rollout schedules and increase uncertainty around infrastructure delivery timing. It could also intensify competition for server memory between AI infrastructure operators and other electronics-market segments.
Memory availability could become a material constraint on AI infrastructure deployment. AI accelerators depend on HBM, while the surrounding server platforms require DRAM; shortages can therefore delay server deliveries and data-center capacity expansion even where processors and accelerators are available.
- TechNews 科技新報 / КитайскийПервоисточник →
Материал подготовлен редакцией ЦОД: Главное на основе указанных первоисточников.