Nvidia's decision to reflect rising memory prices in its AI server costs has highlighted the negotiating power of Samsung Electronics and SK Hynix. The surge in demand for HBM and server DRAM, driven by increased investment in AI data centers, is affecting the prices of finished products, including smartphones and PCs.
According to industry sources, Nvidia recently informed major clients that it may raise the prices of AI server systems scheduled for shipment early next year by more than 15%. This increase will apply to next-generation AI platforms like 'Vera Rubin' and the current generation 'Grace Blackwell.' The actual price hike will vary depending on the memory configuration of each product. Bloomberg identified the rise in memory prices as a key factor behind this decision.
AI servers are equipped not only with AI accelerators but also with large-capacity HBM and server DRAM. As investments in data centers surge, memory manufacturers are prioritizing production capacity for AI-related products. The expansion of HBM production is putting pressure on the supply of general-purpose DRAM, driving up overall memory prices.
Nvidia's price increase is also impacting its clients. Major cloud service providers like Microsoft and Google are now shouldering the burden of rising Nvidia server prices. Even big tech companies building their own AI data centers are finding it difficult to avoid the impact of rising memory costs.
The increase in memory costs is also affecting consumer devices. Apple has recently raised prices on some products, including Macs and iPads. Apple CEO Tim Cook stated that memory costs are rising and are expected to continue to increase in the future. Qualcomm is also reportedly considering passing on the burden of rising memory prices to its product prices.
The potential for improved performance for Samsung Electronics and SK Hynix has increased due to these price hikes. If both HBM and server DRAM prices rise, the increase in AI data center investments will directly reflect in the revenues and profitability of memory manufacturers. As competition in AI semiconductors intensifies, the competition for securing memory is also expected to grow.
However, memory manufacturers do not have pricing power across all products. China's CXMT is expanding its production of general-purpose DRAM. While Samsung and SK Hynix focus their production capacity on high-value memory for AI, CXMT's increase in supply of general products could slow the upward trend in general DRAM prices.
Industry analysts suggest that the current strength in memory prices is linked to the investment structure in AI infrastructure, going beyond mere supply shortages. The competition for AI accelerator performance is leading to a race for HBM capacity, establishing memory as a key component that influences data center construction costs.
An industry insider noted, "In AI servers, memory has become a core component that determines both performance and cost, rather than just supporting the GPU. For the time being, Samsung Electronics and SK Hynix are likely to maintain their pricing power in the high-value memory market, where supply is tight."
* This article has been translated by AI.
Copyright ⓒ Aju Press All rights reserved.