Intel Turns to Large LPDDR5X Pools for AI Inference GPU

Intel shared additional Computex details on Crescent Island, its Xe3P-based data-center GPU for AI inference. The notable memory angle is the planned use of a very large LPDDR5X pool, with configurations reaching up to 480GB. That choice matters because it shows another path around the constrained high-end AI memory market. Instead of relying only on premium HBM-style stacks, Intel is positioning LPDDR5X as a capacity-heavy option for inference workloads that need long context handling and lower power characteristics. For RamTrend, the signal is supportive for LPDDR5X demand. If more AI accelerators adopt large pools of mobile-derived DRAM, suppliers may see additional data-center pull on a technology class more commonly associated with client and mobile devices. The article does not identify memory suppliers or pricing, so the impact remains directional rather than quantified.