SK Hynix, the dominant leader in HBM memory, is reportedly receiving unprecedented offers from global tech giants. Companies like Microsoft, Google, and Amazon are proposing massive investments in its new production lines. Their plans include directly funding the purchase of increasingly expensive chip manufacturing equipment, such as ASML lithography machines and advanced HAR etching and thin-film deposition tools, to expand capacity. This collaborative approach aims to secure as much long-term supply of HBM and DRAM/NAND memory chips as possible amidst intense competition. Such investment and procurement models are unprecedented in the global memory chip industry, highlighting the severe shortage of these components. As the AI boom fuels surging demand for computing infrastructure, memory chip makers are struggling to keep pace with the exponential growth. Memory chips are critical components for the massive new AI data centers being built globally, as well as for smartphones, PCs, wearables, and other electronics. Both Google's vast TPU AI clusters and NVIDIA's massive AI GPU clusters rely on HBM memory systems fully integrated with AI chips. Furthermore, tech giants accelerating the construction or expansion of AI data centers require large-scale purchases of server-grade DDR5 memory and enterprise-class high-performance SSDs/HDDs. Samsung Electronics, SK Hynix, and Micron Technology are positioned across these three core memory segments: HBM, server-grade high-performance DRAM (including DDR5/LPDDR5X), and high-end data center SSDs. They are direct beneficiaries of the "AI memory and storage stack," reaping the "super dividends" of the AI infrastructure wave. Regarding DRAM/NAND price trends, Goldman Sachs' latest assessment indicates that the 2026 price increase will far exceed its previous optimistic forecasts. The firm recently raised its DRAM price increase forecast from about 150% to 250%-280%, and its NAND price increase forecast from about 100% to 200%-250%. Goldman Sachs argues this is not an ordinary inventory recovery cycle but a "super supply shortage cycle" driven by unprecedented AI-driven demand surges, HBM's complex manufacturing and packaging consuming more capacity, and limited supply elasticity for general DRAM/NAND. GPUs generate intelligence, HBM/DRAM feeds data at high speed, enterprise NAND/eSSD handles hot data and caching, while HDDs store massive amounts of cold/warm data long-term. Therefore, Goldman Sachs believes the AI computing arms race led by cloud giants is transforming memory chips from cyclical commodities into scarce strategic assets. The 2026 DRAM/NAND price hikes are not the end but potentially the initial stage of a super cycle. SK Hynix as a Key Chokepoint in the AI Arms Race According to the latest media reports, six sources familiar with the matter stated that SK Hynix's largest customers have been proposing a series of unprecedented offer models to the Korean chipmaker. These include investing in dedicated memory chip production lines—meaning an entire line would belong to a specific major customer. Three of these sources indicated another proposal involves customers providing substantial funding for semiconductor manufacturing equipment purchases, such as ASML's extreme ultraviolet (EUV) lithography machines or even more expensive high-NA lithography tools. This equipment is used for the world's most advanced chip manufacturing processes, including printing circuits on silicon wafers, etching, thin-film deposition, and CMP, with values reaching billions of dollars. However, two sources noted that the Korean chipmaker, with its strong cash flow, is cautious about accepting such financial commitments from customers. Such deals could tie the company to specific buyers and potentially require supplying memory chips at lower-than-market prices in exchange for longer-term, stable revenue guarantees. One source stated, "Regardless of the type of offer, available memory capacity is essentially zero right now. There isn't even a small portion of extra capacity that can be allocated to a specific customer." Another source mentioned a proposal targeting the first phase of a large fab SK Hynix is building in the Yongin Chip Cluster in South Korea. This manufacturing facility will likely focus on DRAM chip production. The details of these offer models are being reported for the first time. SK Hynix and its main rivals, Samsung Electronics and Micron Technology, have stated they are negotiating multi-year supply contracts with large customers but have not provided specific agreement details. The sources cited by media declined to be identified due to the confidential nature of the discussions. An SK Hynix spokesperson declined to provide details on its contract terms with customers but stated, "We are comprehensively reviewing various offer methods and structural alternatives different from traditional long-term agreements." The company is Asia's third-largest by market capitalization, trailing only TSMC and Samsung. Its stock has surged 154% this year, reaching record highs, driven by increasing investor bets on the memory super cycle fueled by the AI infrastructure frenzy. As of Friday's close in South Korea, Samsung Electronics' market capitalization remained stable above $1 trillion. Samsung reached a milestone on Wednesday, with its year-to-date gain of about 80% pushing its market cap past $1 trillion, becoming the second Asian company after TSMC to achieve this level. AI training relies heavily on large-scale parallel computing, while inference—especially for long-context, multi-turn dialogues, and agentic AI workflows—requires continuously saving KV Cache, context states, and intermediate results. When memory/storage space is insufficient, models must recompute historical states, leading to lower GPU utilization and higher token generation costs. Therefore, HBM, DDR5, LPDDR, enterprise SSDs, and even HDDs/data lakes are forming an "AI memory chain" from near-GPU to far storage, determining an AI system's throughput, latency, concurrency, and per-token economics. This explains the synchronized surge in stocks of memory and data storage companies like Micron, Samsung, SK Hynix, SanDisk, and Western Digital: demand is not concentrated solely on HBM but spills over across the entire chain of DRAM, NAND, SSDs, and HDDs along the AI server architecture. In this so-called "memory chip super cycle" driven by an unprecedented AI infrastructure boom, South Korea's two memory giants—Samsung Electronics and SK Hynix—together account for nearly 50% of the weighting in the Kospi Composite Index. They are powerful engines attracting global capital and core drivers behind the South Korean stock market's record-breaking performance and significant outperformance relative to global markets. The Kospi Composite Index's year-to-date gain for 2026 has already surpassed its wild 76% surge from 2025, which led global markets. However, unlike the full year of 2025, the 2026 gain exceeding 76% has occurred in less than five months from the start of the year. It is unclear which global tech giants are making investment offers to SK Hynix, but sources indicate these large customers include, but are not limited to, U.S. cloud and tech leaders such as Alphabet's Google, Meta (Facebook's parent company), and Microsoft. These tech giants recently outlined plans to significantly increase AI computing infrastructure spending. Meta stated on its earnings call, "We are actively investing to meet our computing infrastructure needs," adding this includes "a series of deals across the supply chain to secure the necessary AI components for future capacity." Microsoft management also stated on its call that it expects overall capital expenditures to rise significantly to $190 billion this year, including an additional $25 billion due to rising costs of components like chips. Record-Breaking Boom Cycle The offers presented to SK Hynix are highly unusual for the memory chip industry, historically known for extreme boom-bust cycles. They are leading memory chip makers to believe the current upcycle—already stronger than the record-breaking 2018 super cycle—will last longer. SK Hynix and Samsung Electronics management stated last month that the current memory chip shortage will persist long-term, as chipmakers need more time to build additional supply and capacity to keep up with the "exponential growth" in AI computing demand. SK Hynix stated, "Due to current supply constraints, there are significant limitations in meeting all customer requests," adding that demand from customers for long-term supply contracts to secure volume is increasing sharply. Memory chip makers have argued in recent years that multi-year contracts would help smooth out volatile demand and reduce the massive investment risks in this cyclical industry, which often requires billions of dollars to significantly expand capacity. The unprecedented AI infrastructure wave and memory super cycle are pushing semiconductors into a new phase characterized by greater "material intensity, process control intensity, and forward-shifted packaging processes." Logic-side 3D structures and new materials, memory-side HBM stacking and interconnect upgrades, and packaging-side CoWoS/hybrid bonding that converts system performance into manufacturing difficulty—these three forces collectively increase the value density in key areas like deposition/etching/CMP/advanced packaging/core metrology. They are also shifting semiconductor equipment demand more clearly from "cyclical fluctuations" to a "structural mega-expansion cycle." The critical semiconductor equipment foundation for memory expansion includes not only ASML lithography machines but also expensive, high-end tools for high-aspect-ratio (HAR) etching/deposition, CMP (chemical mechanical polishing), metrology/inspection, and hybrid bonding required for HBM/DRAM/NAND. Applied Materials, a ubiquitous semiconductor equipment giant in fabs, noted in its latest technical commentary that HBM manufacturing adds approximately 19 extra materials engineering steps compared to traditional DRAM. It claims its most advanced semiconductor equipment covers about 75% of these steps and has also launched a hybrid bonding system for advanced packaging/memory chip stacking, positioning HBM and advanced packaging manufacturing equipment as strong mid-to-long-term growth vectors for the company. However, chip industry executives, investors, and analysts widely note that several unresolved issues remain, such as ensuring customers do not cancel deals and pricing memory chips advantageously. Samsung stated that, unlike past long-term agreements, its recent contracts with some customers are "binding" but did not provide further details. One source familiar with chipmakers' internal briefings mentioned one structure discussed as part of long-term agreements between chipmakers and their largest customers involves a price band mechanism. This would set floor and ceiling prices for annual contract pricing, effectively eliminating quarterly or seasonal price negotiations. Another structure discussed internally by chipmakers involves advance payments, requiring customers to prepay 30% to 40% in cash. However, another source noted that chip suppliers are acting very cautiously in allocating scarce capacity to avoid antitrust regulatory scrutiny or perceptions of favoring specific customers. "They don't want to 'bet on one horse' in the AI race and back the wrong one," the source said.