The semiconductor industry remained firmly focused on AI infrastructure this week, with major developments spanning NAND Flash, HBM, DRAM, AI servers, advanced packaging, and power management. At the same time, rising memory costs are beginning to reshape procurement strategies and investment priorities across the supply chain. Here are the key developments from August 24–30, 2026.
01. Kioxia and SanDisk Commit $31 Billion to Expand Japan Memory Production
Kioxia and SanDisk announced plans to invest more than $31 billion in Japan through 2032, targeting expanded flash-memory production capacity and supporting infrastructure. The investment will strengthen their long-running manufacturing partnership and is aimed at meeting growing demand from AI, data centers, and high-capacity storage. Kioxia also announced plans to expand production at its Kitakami facility with a new Fab3/K3 building, designed to increase capacity for advanced 3D flash memory. The companies said future investment will depend partly on market conditions and government support. The move highlights how AI-driven storage demand is encouraging memory suppliers to commit substantial capital to NAND Flash capacity and advanced 3D memory technologies.
02. SK hynix Starts Indiana HBM Facility, Targets HBM4E Production in 2029
SK hynix has begun construction of its first U.S. HBM production site in West Lafayette, Indiana, with investment exceeding $4 billion. The facility is scheduled to begin volume production of next-generation HBM4E in the second half of 2029. Under the planned operating model, advanced wafers will be produced in Korea before being shipped to Indiana for advanced packaging and testing, creating a more localized U.S. HBM supply chain. The project is supported by the U.S. CHIPS Act and is expected to strengthen SK hynix's relationships with major AI customers including NVIDIA, Microsoft, and Google. CEO Kwak Noh-jung also expects the current memory supply shortage to continue through 2030.
03. NVIDIA Reportedly Raises AI Server Prices as HBM Costs Continue to Climb
NVIDIA is reportedly preparing to increase prices for AI server systems based on Grace Blackwell and Vera Rubin by more than 15% for systems shipping from early 2027. Industry analysts believe rising HBM and DRAM costs are a major factor behind the expected adjustment. The higher system prices could give Samsung Electronics and SK hynix greater negotiating leverage in future HBM contracts, with some market forecasts pointing to substantial increases in HBM pricing during 2027. NVIDIA's high-margin AI systems provide some room to absorb higher component costs, but the reported pricing strategy suggests that part of the increase may ultimately reach hyperscalers and data-center operators.
04. Memory Costs Take a Larger Share of Cloud AI Infrastructure Spending
TrendForce expects global cloud service providers to sharply increase capital expenditure as AI infrastructure expands. Total CSP CapEx is projected to rise 98% year-on-year in 2026 and another 50% in 2027. More importantly for the semiconductor supply chain, DRAM and NAND Flash are expected to account for 47% of CSP capital spending in 2026, rising to 68% in 2027. Server DRAM and enterprise SSD pricing are already under pressure from strong procurement demand, while HBM remains particularly constrained. Although new fabs and process migrations should increase memory bit supply in 2027, TrendForce expects higher contract prices to keep memory as one of the largest cost components in AI infrastructure.
05. Qualcomm Brings Samsung and SK hynix Into Its HBC Memory Strategy
Qualcomm is reportedly working with Samsung Electronics and SK hynix to commercialize its new High Bandwidth Compute (HBC) architecture. Introduced earlier this year, HBC places compute and low-power DRAM in a tightly integrated 3D structure, aiming to improve memory bandwidth, energy efficiency, and AI inference performance. Qualcomm says its first-generation HBC technology is designed to deliver up to 133 TB/s of effective memory bandwidth per card, with commercial sampling planned for mid-2027. Samsung and SK hynix are expected to contribute their memory expertise while Qualcomm handles compute and system integration. The development signals growing competition between HBM-based architectures and alternative high-bandwidth memory solutions for next-generation AI infrastructure.
06. Infineon Acquires C2i to Strengthen AI Data Center Power Solutions
Infineon Technologies announced the acquisition of C2i Semiconductors, a Bangalore-based company specializing in software-defined multiphase controllers and smart power stages for AI data centers. The deal expands Infineon's capabilities in intelligent power management and strengthens its strategy for delivering power solutions from the grid to the processor core. C2i's technology combines digital control, power semiconductors, and software algorithms to respond more dynamically to the rapidly changing power demands of advanced AI processors. The companies also plan to advance vertical power delivery architectures, including Substrate Integrated Voltage Regulators. The transaction is expected to close in the third quarter of 2026, further expanding Infineon's AI data-center power portfolio and engineering capabilities in India.
07. Micron Taiwan Faces Possible Labor Action as Memory Profits Surge
Micron's Taiwan operations are facing potential labor disruption after unions called for changes to the company's incentive and profit-sharing system. According to industry reports, mediation is expected to continue through late August and September, with a strike vote potentially taking place in September if negotiations fail. The two major Taiwan unions represent roughly 10,000 employees, and a preliminary survey reportedly showed around 80% support for possible action. The dispute comes as strong DRAM and HBM demand improves profitability across the memory industry. Similar compensation discussions are also taking place at SK hynix and Samsung, highlighting how the current memory upcycle is affecting not only pricing and capacity allocation, but also workforce negotiations.
Outlook
This week's developments point to a semiconductor market increasingly shaped by the AI infrastructure buildout. Memory remains at the center of the trend: HBM capacity is expanding, NAND investment is accelerating, DRAM prices remain under pressure, and AI servers are becoming more expensive as component costs rise. At the same time, semiconductor manufacturers are investing closer to end markets, while new architectures such as Qualcomm HBC and intelligent power-delivery solutions are emerging to address the next generation of AI computing.
At Futuretech Components, we support global customers with electronic component sourcing, semiconductor supply chain solutions, traceable components, inventory management, and shortage sourcing. By combining global supplier networks with quality control and supply-chain expertise, Futuretech helps customers manage changing lead times, pricing pressure, and component availability with greater confidence.
