Samsung has pre-allocated about 70% of its memory capacity under long-term agreements. According to Korean media reports, the main customers are Microsoft, Google and Nvidia, and they have signed supply contracts until 2031.
The reason for this move is the rapid growth of demand for HBM in the artificial intelligence industry. The move to HBM4 further exacerbates the DRAM shortage as the new modules use more memory layers, reducing available production for other products. Against this background, Samsung and SK Hynix are already considering expanding production capacity. Samsung may convert its S5 factory in Pyeongtaek to produce memory, while SK Hynix is discussing a joint venture with the Japanese company.
Long-term contracts allow manufacturers and large customers to determine supply volumes and prices in advance, reducing reliance on wild market swings. The difference from current spot prices seems particularly stark. For example, the price of a 36GB HBM3E module in the spot market is about 2.87 million won, while similar memory costs about 500-700,000 won under a long-term agreement.
For HBM4 with 16 layers of DRAM, the spot price is already around 4.8 million won. As a result, as the market faces increasing shortages, large AI companies gain a significant advantage by reserving memory in advance on more favorable terms.
Other publications on this topic
OpenAI has significantly increased its purchases of Apple computers to perform tasks related to artificial intelligence. According to The Information, the company has purchased tens of thousands of Mac minis and Mac Studio units in recent months, and Anthropic leases similar devices through Amazon AWS.
Today 19:15
0
Even before the platform’s official launch, NVIDIA RTX Spark-based laptops have been growing in popularity. ASUS and MSI report strong demand: partners have already selected the first batch of devices, and manufacturers are trying to secure additional supply from NVIDIA.
Today 17:18
7
The cost of server RAM and solid-state drives in Russia rose by 25% to 150% in the first six months of 2026, with prices of some items tripling as global production capacity was reallocated to support artificial intelligence infrastructure.
Today 16:15
2
