NVIDIA CEO Jensen Huang Warns Memory Shortage Will Last for Years
NVIDIA CEO Jensen Huang stated that the memory shortage will last for "several years," a comment made before NVIDIA announced a partnership with SK Hynix on Monday.
He pointed out that the entire supply chain, from wafers and packaging to silicon photonics, is facing shortages, primarily due to high AI demand. The collaboration will cover multiple areas including AI supercomputers, CPUs, new computers, and robotics.
In market mechanisms, the explosive demand for AI training and inference drives data center operators to continuously secure high-bandwidth memory supplies. Under this event-driven scenario, funding is shifting from general computing to the expansion of dedicated memory capacities like HBM, benefiting NVIDIA's ecosystem and memory giants like SK Hynix, while putting pressure on smaller AI companies and traditional computing demanders with lower supply priority.
Source: Public Information
ABAB AI Insight
黄仁勋过去多次在财报会议中提前预警供应链瓶颈，如Hopper到Blackwell过渡期的内存紧张，此前NVIDIA已通过深度绑定SK海力士等伙伴确保HBM优先供应，并持续推动硅光子与先进封装技术以匹配GPU性能跃升。
资本路径上，NVIDIA资源向HBM内存产能与全栈供应链集中，通过战略合作锁定关键节点，动机在于维持AI加速器领先地位并将巨额行业资本开支转化为自身生态壁垒与长期定价权。
类似早期GPU短缺期与台积电等伙伴的绑定路径，当前AI硬件处于从单纯算力竞赛向内存、封装与互连全链条优化的转型阶段。
本质属于资本集中，内存持续短缺迫使行业资源向少数高产能供应商与NVIDIA生态高度集聚，机制上通过供需失衡放大定价权并加速AI基础设施向垂直整合与高效供应链重构。
ABAB News · 认知定律
算力看似AI核心，实则内存供应链才是规模扩张的长期结构性瓶颈。
卖通用芯片烧库存，卖优先HBM收生态，顶级卖的是被合作锁定的全栈定价权。
需求不缺资金，缺的是持续产能；胜者用供应链集中重塑AI硬件的结构.