Amazon & Google: Estimated Spending on South Korea's Two Giants (Samsung/Hynix) (2026F) 1. Core Spending Data Brief (Estimated Spending) Based on the logic of doubling the storage value in the AI server cost structure, the estimated procurement scale for the two giants in 2026 is as follows: Amazon storage spending: $30-36 billion Google storage spending: $25-30 billion. Key Variable: At the beginning of 2026, due to the extreme shortage of HBM capacity, procurement heads from Amazon and Google were reported to have "long-term rented hotels in South Korea" to seek to sign 3-5 year long-term supply agreements (LTA), which may lead to actual spending exceeding the budget. 2. Business Line Deep Breakdown (Segment Breakdown) SK Hynix: The Absolute Beneficiary of AI Premium Google Logic: Google's TPU v7 architecture has extremely high bandwidth requirements for HBM. Hynix, as the number one supplier of HBM3E, accounts for over 60% of Google's AI accelerator storage demand. Amazon Logic: Amazon's Trainium 2 chip is ramping up production, and Hynix has customized dedicated HBM for it. Profit Quality: Hynix raised the price of HBM3E by 20% in January 2026, and both giants almost fully accepted it to "maintain volume." Samsung: All-Round Backup and NAND Dominance NAND/SSD Demand: A significant portion of Amazon's $200 billion spending is allocated for data center expansion, and Samsung's dominance in the enterprise SSD (eSSD) field is irreplaceable, expected to secure over 50% of flash memory orders from both companies. HBM Catch-Up: Although Samsung is slightly behind Hynix in HBM3E progress, it has become the safety supplier for both giants amid the shortage of "general-purpose DRAM (DDR5)" due to its massive capacity reserves. 3. Supply Chain Game: The "Expectation Gap" of 2026 (The Strategic Shift) From "On-Demand Purchasing" to "Long-Term Rental for Goods": In previous years, storage was considered a cyclical product, and giants tended to push prices down. However, in 2026, due to HBM occupying general DRAM capacity, traditional server memory prices skyrocketed (Q1 price list increased by 60%-70% in a single quarter). Expectation Gap: Of Amazon's $200 billion expenditure, the funds originally reserved for storage may face a shortfall of over $5 billion, and this portion of profit will flow directly from Amazon's financial statements to Hynix and Samsung. Impact of Self-Developed Chips: Both Google and Amazon are increasing their self-developed ASICs (TPU/Trainium). This means their storage requirements are no longer "standardized" but rather "customized." Hynix, through this deep binding, is transforming from a "supplier" to a "co-developer."...