As AI-driven memory demand explodes and the DRAM supply shortage deepens, market leader Samsung Electronics is projected to increase its DRAM production capacity by only 6% this year. The strategic choice to prioritize cutting-edge process conversion and yield stabilization over volume expansion could prove to be a liability amid the second-half supply crunch.<\/p>\n
According to data released on the 24th by market research firm Omdia, Samsung Electronics’ DRAM production capacity is expected to grow approximately 6% this year, from 7.695 million wafers annually in 2025 to 8.175 million wafers. Next year, the growth rate is estimated to fall to just 1.3%, reaching 8.28 million wafers. Notably, some production loss is expected during the conversion of legacy DRAM equipment to 10-nanometer 6th-generation (1c) DRAM, which will be used in 6th-generation HBM (High Bandwidth Memory, HBM4).<\/p>\n
A source at Samsung Electronics’ Pyeongtaek campus explained, “While steady expansion investments are being made at the Pyeongtaek campus, it appears difficult to significantly increase DRAM production capacity with currently operating lines to match market demand. The company’s basic policy emphasizes yield and productivity of cutting-edge DRAM over volume expansion, which has been Vice Chairman Jun Young-hyun’s core philosophy since last year.”<\/p>\n
Indeed, the Pyeongtaek campus is undergoing process conversion investments through retrofitting existing lines. As some lines that had remained on mature processes transition to 6th-generation technology, Pyeongtaek’s share of Samsung Electronics’ total DRAM production capacity is projected to expand from 49.5% in 2025 to 55.4% by 2027. However, since this increase comes from retrofitting existing lines rather than completing new fabs, observers note that the actual pace of supply expansion may be slower than the surface numbers suggest.<\/p>\n
In contrast, the Hwaseong 16 and 17 lines have shown stagnation or decline for three consecutive years. During its Q4 2024 earnings call, Samsung Electronics announced plans to significantly reduce its legacy DRAM and NAND flash mix, which had accounted for 30% of existing DRAM revenue, citing declining profitability of legacy products. The Hwaseong Line 12 recently completed its conversion from 2D NAND production to an end-fab for back-end-of-line (BEOL) processes for Hwaseong Line 15, and is expected to contribute to expanded DRAM shipments starting in the second half.<\/p>\n
SK Hynix is expected to maintain double-digit growth over the same period. According to Omdia projections, SK Hynix’s DRAM production capacity will grow 9.9% from 6.06 million wafers in 2025 to 6.66 million wafers in 2026, and another 9.5% to 7.29 million wafers in 2027. Consequently, the production capacity gap between the two companies is projected to narrow by approximately 39% over two years, from 1.635 million wafers in 2025 (Samsung 27.0% ahead) to 990,000 wafers in 2027 (13.6% ahead).<\/p>\n
Samsung Electronics has continued to prioritize yield stabilization over volume expansion through new capacity additions, having previously lagged in the 5th-generation (1b) DRAM yield competition. Since Vice Chairman Jun’s appointment, the company has reportedly been re-examining all product processes and designs, deliberately moderating its production capacity ramp-up to build stability starting from 5th-generation DRAM.<\/p>\n
However, some counter that wafer input-based supply capacity does not perfectly equate to actual market demand response. Newer DRAM processes offer higher bit density per wafer, meaning bit-based calculations partially offset the severe supply-demand imbalance seen in the first half. An industry source noted, “Looking only at wafer inputs, Samsung’s expansion pace appears slow, but actual supply capacity measured by bit growth is increasing. That said, with the DRAM shortage still severe in the second half, the memory market boom is likely to extend further.”<\/p>\n
Nvidia Raises Prices Amid Memory Shortage<\/p>\n
The memory supply crunch has even altered pricing strategy at Nvidia, the world’s most valuable company by market capitalization. Nvidia recently notified major customers that prices for AI chip-equipped servers will increase by more than 15%. The adjusted pricing applies to shipments starting early next year, covering systems featuring its flagship ‘Grace Blackwell’ and next-generation ‘Vera Rubin’ chips. The magnitude of price increases will vary depending on chip generation and memory configuration.<\/p>\n
With HBM and high-performance server DRAM prices skyrocketing\u2014components critical to AI accelerator performance\u2014even Nvidia can no longer absorb the cost pressure and is passing the burden to customers. No matter how powerful an accelerator’s compute capability, insufficient HBM bandwidth and server DRAM capacity inevitably create data bottlenecks that severely degrade overall system performance.<\/p>\n
Currently, the three global memory giants\u2014Samsung Electronics, SK Hynix, and U.S.-based Micron\u2014supply the vast majority of high-performance memory. But their production capacity cannot keep pace with surging AI server demand, creating a thoroughly seller-dominated market. Industry analysts suggest that HBM supply for the next three years is already sold out.<\/p>\n
Price pressure is even more pronounced in server DRAM. Market research firm TrendForce expects Q3 server DRAM contract prices to rise 13\u201318% quarter-over-quarter. For PC DRAM commodity products with publicly available average fixed transaction prices, prices surged to $24 (approximately 33,000 won) last month, the highest since tracking began in June 2016. TrendForce data also shows Q2 server DRAM contract prices jumped 53\u201358% quarter-over-quarter.<\/p>\n
Nvidia now faces a test to justify its price increases by proving chip performance to customers. This is why attention is focused on Nvidia’s Q2 earnings release scheduled for the 26th (local time). The market is watching the mass production timeline for the next-generation Vera Rubin chip, which began initial shipments to big tech customers including Microsoft this month. Another key point is whether the company will reveal a concrete roadmap for the new ‘Rubin Ultra’ product, which recently drew market attention following reports that HBM capacity would be reduced from original plans.<\/p>\n
Samsung-SK Hynix Pricing Power Strengthens… But AI Ecosystem Cost Burden Raises Boomerang Concerns<\/p>\n
With Apple, Qualcomm, and now Nvidia deciding to raise product prices, analysts suggest profitability for memory semiconductor makers like Samsung Electronics and SK Hynix could improve further. Nvidia is a highly profitable company with gross margins reaching 75%. That such a company is passing costs through to product prices signals just how much stronger memory makers’ pricing power has become.<\/p>\n
With big tech likely to accelerate AI data center construction and AI chip development in the near term, memory makers’ profitability is expected to continue its upward trajectory. Amazon, Microsoft, Google, and Meta are developing their own AI chips to reduce dependence on Nvidia. Even for these efforts, they must secure HBM and DRAM in massive quantities.<\/p>\n
However, concerns are also being raised that if cost burdens across the AI ecosystem intensify, they could boomerang back on memory semiconductor makers. The Information, a U.S. tech media outlet, reported that some server companies have been notified of AI chip price increases of around 17%, which would add at least $5 billion (approximately 6.9 trillion won) to the cost of building a 1-gigawatt (GW) class data center. In other words, while the AI investment boom drove memory prices higher, the resulting cost escalation could ultimately dampen AI investment appetite.<\/p>\n
The prevailing view is that record-breaking operating profits at Samsung Electronics and SK Hynix from the memory super-cycle will continue for the foreseeable future. The Korea Development Institute (KDI) raised its South Korean economic growth forecast for this year from 2.5% to 3.2%, reflecting the memory boom, and the Bank of Korea is also expected to present a growth rate in the 3% range.<\/p>\n","protected":false},"excerpt":{"rendered":"As AI-driven memory demand explodes and the DRAM supply shortage deepens, market leader Samsung Electronics is projected to…\n","protected":false},"author":2,"featured_media":129484,"comment_status":"","ping_status":"","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[18],"tags":[644,699,12741,1485,2295,24867,127,276,275,23385],"class_list":["post-129483","post","type-post","status-publish","format-standard","has-post-thumbnail","category-samsung-electronics","tag-dram","tag-hbm","tag-jun-young-hyun","tag-micron","tag-nvidia","tag-omdia","tag-samsung","tag-samsung-electronics","tag-sk-hynix","tag-trendforce"],"_links":{"self":[{"href":"https:\/\/www.europesays.com\/korea\/wp-json\/wp\/v2\/posts\/129483","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/www.europesays.com\/korea\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/www.europesays.com\/korea\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/www.europesays.com\/korea\/wp-json\/wp\/v2\/users\/2"}],"replies":[{"embeddable":true,"href":"https:\/\/www.europesays.com\/korea\/wp-json\/wp\/v2\/comments?post=129483"}],"version-history":[{"count":0,"href":"https:\/\/www.europesays.com\/korea\/wp-json\/wp\/v2\/posts\/129483\/revisions"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/www.europesays.com\/korea\/wp-json\/wp\/v2\/media\/129484"}],"wp:attachment":[{"href":"https:\/\/www.europesays.com\/korea\/wp-json\/wp\/v2\/media?parent=129483"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/www.europesays.com\/korea\/wp-json\/wp\/v2\/categories?post=129483"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/www.europesays.com\/korea\/wp-json\/wp\/v2\/tags?post=129483"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}