Intel CEO Lip-Bu Tan on the one thing troubling his customers the most: I think in terms of the AI, the biggest challenge is…
AI-summarised brief · reviewed before publication
Intel CEO Lip‑Bu Tan warned that memory scarcity is now the primary obstacle for companies building AI infrastructure, citing a five‑to‑seven‑fold rise in memory prices and limited production that forces project delays. Speaking at Splunk’s .conf26 in Denver, Tan explained that demand for high‑bandwidth memory (HBM) from Nvidia and AMD accelerators has outstripped the output of Samsung, SK hynix and Micron, prompting a shift of existing lines to HBM and reducing capacity for standard DRAM. He relayed industry forecasts that relief won’t arrive until 2028. Tan also noted CPU shortages, with Intel able to meet only half of customer orders, and highlighted the geopolitical risk of relying on Taiwanese packaging suppliers, which dominate 95% of the market. Intel’s 18A node is now in mass production, with 14A slated to start soon.
💡 Why It Matters
- · The bottleneck forces AI developers to postpone deployments, throttling the rollout of next‑generation services that rely on massive model inference and orchestration.