Nvidia’s stock fell more than 2% on the New York Stock Exchange, but the Philadelphia Semiconductor Index actually rose 1.3%. Expectations for AI infrastructure investment benefits spread across the broader semiconductor sector, driving simultaneous gains in AMD, Intel, and TSMC. In South Korean trading on the morning of the 9th, SK Hynix climbed more than 3% to break through ₩1.85 million (approximately $1,400), while semiconductor materials, components, and equipment names such as YEST and Nepes posted sharp gains of around 8%.

On the 8th (local time), U.S. markets fell across the board—the Dow (-1.18%), S&P 500 (-0.58%), and Nasdaq (-0.32%)—pressured by rising oil prices from Middle East conflicts and concerns about AI encroaching on software businesses. The Philadelphia Semiconductor Index, however, rose 1.30% to 11,887.87. Following OpenAI’s unveiling of GPT-6 Astra last week, concerns about software displacement have grown, but analysts note that semiconductor and data center capex beneficiaries are trading on a separate trajectory.

Among individual stocks, Intel stood out with a 9.05% surge, driven by analyst upgrades and reports of CPU price increases. Qualcomm rose 3.17% on the announcement of its AI semiconductor collaboration with Amazon. AMD gained 5.90%, while TSMC ADR climbed 2.35%, ASML advanced 2.91%, and Arm rose 3.74%. Nvidia, by contrast, fell 2.01% to $225.73, and Micron slipped 1.61% to $1,000.26, reflecting a mixed picture even within the semiconductor space.

South Korean Semiconductor Stocks Rally in Tandem

As of 10 a.m. on the 9th, Samsung Electronics was up 1.11% at ₩272,500 (approximately $200) in South Korean trading, while SK Hynix rose 3.18% to ₩1.85 million (approximately $1,400). Gains among materials, components, and equipment names were even larger. YEST jumped 8.45%, Nepes climbed 8.14%, YC rose 6.79%, HPSP advanced 6.75%, and Soulbrain gained 5.84%.

KB Securities forecast that memory prices and shipment volumes will both increase in 2027, with production capacity being allocated primarily through long-term supply agreements. The firm noted that intensifying AI model competition is expanding demand for high-capacity server memory, benefiting Samsung Electronics and SK Hynix. Limited new supply is also expected to keep memory market conditions tight.

Meritz Securities maintained a “Buy” rating on TES with a target price of ₩200,000 (approximately $150), citing SK Hynix’s M17 investment and new equipment effects. The firm assessed that the semiconductor equipment cycle is entering the early stages of an unprecedented boom. LS Securities projected that HPSP’s equipment revenue will recover on expanded memory investments by SK Hynix and China’s CXMT, setting a target price of ₩220,000 (approximately $160).

Daishin Securities raised its third-quarter operating profit forecast for AI server substrate-related names from ₩76.7 billion to ₩83.7 billion (approximately $57.3 million to $62.5 million), reflecting expanded demand for AI server substrates. That represents a 576% increase year-over-year and a 33.1% gain quarter-over-quarter. The firm maintained its “Buy” rating and ₩170,000 (approximately $130) target price, citing module and semiconductor substrate revenue expansion and a rising mix of high-value-added substrates.

Qualcomm-Amazon Collaboration Worth Up to $60 Billion

Qualcomm and Amazon have agreed to collaborate on AI data center semiconductors and optical communications worth up to $60 billion (approximately ₩80.3 trillion). The two companies will jointly develop custom AI semiconductors and connectivity technologies, with Qualcomm providing Amazon warrants valued at approximately $4 billion (approximately ₩5.4 trillion) tied to purchase performance. However, the $60 billion figure represents the maximum potential purchase amount, not a fully committed order volume.

ASML and TSMC are jointly pursuing a transition to 12-inch large-format photomasks for High-NA EUV lithography. TSMC plans to introduce High-NA EUV for advanced-node mass production by 2030, with the two companies aiming to establish a pilot production line for large-format photomasks by 2031 and complete an advanced semiconductor production system by 2033.

OpenAI reportedly plans to begin deploying its own AI chips, co-developed with Broadcom, starting later this year. The chips will be used alongside accelerators from Nvidia and AMD, diversifying the chip supply chain for AI computing infrastructure.

Memory Spot Prices Rise, Led by eTT and MLC

DRAM spot prices showed notable strength in eTT products while standard DDR4 products stalled. DDR4 16Gb (2Gx8) 3200 fell 1.61%, reversing the previous session’s gain. In contrast, DDR4 16Gb (2Gx8) eTT rose 0.94%, and DDR4 8Gb (1Gx8) eTT climbed 2.02%. DDR4 8Gb eTT in particular extended its gains after rising 1.67% in the prior session.

DDR5 16Gb (2Gx8) 4800/5600, used in high-performance AI server memory, edged up 0.06%, a smaller increase than the previous session’s 0.12%. DDR4 8Gb (1Gx8) 3200 also rose just 0.02%, a sharp slowdown from the prior session’s 0.96% gain.

NAND flash spot prices continued to climb for MLC products while SLC products declined across the board. MLC 64Gb 8GBx8 rose 2.66% to extend its strength, and MLC 32Gb 4GBx8 gained 1.09%. SLC 1Gb 128MBx8, however, fell 1.31%, widening its decline from the previous session’s -1.01%.

Meanwhile, China’s CXMT has reportedly expanded its R&D workforce by approximately 60% year-over-year, reaching roughly 78% of SK Hynix’s level. Analysts view this as a rapid acceleration of investment in DRAM competitiveness and next-generation memory development, including LPDDR6. China has also decided to impose provisional anti-dumping deposits of up to 99.2% on Japanese-origin dichlorosilane, a key material used in semiconductor thin-film deposition. The measure could affect the supply chain for Japanese materials into China.

Samsung Electronics has entered into a strategic partnership with French AI company Mistral AI to jointly develop proprietary AI models and platforms specialized for semiconductor design and manufacturing. The collaboration will combine manufacturing data from Samsung’s DS division with Mistral AI’s model technology for applications including data analytics, defect prediction, and process optimization. Samsung has also made a significant equity investment in Mistral AI to strengthen the partnership, though the investment amount was not disclosed.

Samsung Electronics is also partnering with ASML on joint development of 12-inch photomasks and plans to apply High-NA EUV to DRAM mass production starting in 2028. The goal is to improve process productivity by reducing the exposure area limitations of existing 6-inch masks. Reports also emerged that the 2-nanometer production capacity at Samsung’s Taylor, Texas fab has already been fully booked before operations begin. Tesla’s AI5 and AI6 chips, Broadcom’s next-generation communications chips, and Arm-related AI chips were cited among the expected volume, with initial capacity set at 50,000 wafers per month. Trial operations are expected as early as late September to early October, with mass production projected for 2027.