3.25 million chips per month; Samsung HBM4 capacity to rise 40% amid strong orders
Samsung is raising HBM4 semiconductor output, adding supply of advanced memory chips.
Source →The trace
How it reaches your house.
Starting from High-Bandwidth Memory (HBM3/HBM3E), we walked the supply graph hop by hop to the goods you buy and the companies you own. Each step is the engine’s reasoning, scored for confidence and cited where a source backs it.
First impact
High-Bandwidth Memory (HBM3/HBM3E) →supply upHBM is a critical input to servers and cloud hardware, and Samsung's roughly 40% HBM supply expansion should ease that bottleneck and raise downstream hardware availability.
93% confidenceDDR5 DRAM Server Memory →Shortage riskSamsung Electronics005930.KS
Samsung is redirecting manufacturing toward HBM4 and outsourcing commodity DDR5 production, so its direct DDR5 server-memory contribution tightens despite the HBM ramp.
82% confidenceHigh-Bandwidth Memory (HBM3/HBM3E) →Stands to benefitHop 2
Samsung Electronics →supply upHBM (High Bandwidth Memory) Stack
Samsung's HBM4 capacity ramp directly increases availability of HBM-family advanced memory supply to the market.
96% confidenceServers and cloud hardware →Prices riseAI Training GPU (H100/H200/B200 class)
More available HBM lets more servers and cloud hardware be built and deployed, which raises demand for the critical AI GPU input and tends to tighten its price/supply.
60% confidenceHop 3
AI Training GPU (H100/H200/B200 class) →Stands to benefitTaiwan Semiconductor Manufacturing Company (TSMC)TSM
Looser HBM constraints support higher AI GPU build volumes, which increases demand for TSMC’s sole-qualified leading-edge foundry capacity.
90% confidenceAI Training GPU (H100/H200/B200 class) →Stands to benefitMore available HBM raises the feasible supply of NVIDIA’s H100/H200/B200-class GPUs, directly supporting shipments and revenue.
86% confidenceAI Training GPU (H100/H200/B200 class) →Stands to benefitHop 4
NVIDIA Corporation →supply upAI Training GPU (H100/H200/B200 class)
Samsung's HBM4 capacity expansion should ease a tight memory bottleneck for NVIDIA's dominant AI-training GPUs, so their effective availability rises despite some generation-transition lag.
79% confidenceAdvanced Micro Devices (AMD) →supply upAI Training GPU (H100/H200/B200 class)AMD
Samsung's HBM4 capacity expansion should ease a key advanced-memory bottleneck for AMD's AI accelerator line, modestly increasing AI training GPU supply given AMD's ~10% share and long replacement lead time.
78% confidenceTaiwan Semiconductor Manufacturing Company (TSMC) →supply up
Where it lands
What you buy, and what you own.
Goods you buy
Servers and cloud hardware · Smartphones and tablets
In the MyPRIA app, this is checked against your own holdings.