On July 4, 2026, a prediction from Trendforce became the starting point for this market cycle: the price of traditional DRAM is expected to increase by 13% to 18% quarter-on-quarter in the third quarter of 2026. After years of cyclical fluctuations, such a price increase expectation was interpreted by the market as a clear signal that AI computing power demand has begun to substantially encroach on storage capacity. Alongside the rising price curve is the rhythm of expansion and financing: Micron Technology's expansion project in Japan has already broken ground, with a planned investment of approximately 1.5 trillion yen, and the earliest new production line will not be able to ship until the summer of 2028. The Japanese Ministry of Economy, Trade and Industry is preparing to provide up to 500 billion yen in financial support, trying to secure a position for the domestic supply chain in the next round of the AI chip and storage game. On the other side of the capital market, SK Hynix is promoting the issuance of ADR in the United States, intending to issue no more than 2.5% of its total share capital as new shares, with a basic underwriting fee set at 0.5% and accompanied by floating rewards. This issuance is expected to rank among the largest stock offerings in history and will directly provide ammunition for its storage expansion around AI. On the same day, Serenity released data on Chinese private VC funding flows, revealing that large model-related projects attracted approximately 23.56 billion USD, while AI infrastructure and technology layers garnered about 15.74 billion USD. Funding is flowing from pure software algorithms further into physical AI and world models, indicating that the long-term demand surrounding "computing power - storage - hardware" is being pre-emptively bet on. Price expectations rising, capacity layout extending, equity financing increasing, and VC doubling down on the physical layer—these originally scattered actions are converging into a consensus: AI demand is pushing storage back to cyclical highs, dragging the entire AI industry chain into a new round of high investment, high risk but equally high return expansion track.

13%-18% Price Increase Expectation: Storage Manufacturers Welcome a New Cycle

Following the trajectory of capital flowing from software to "computing power - storage - hardware," the price side is also providing the same answer. On July 4, 2026, Trendforce put forward a significant forecast: the traditional DRAM price is expected to increase by 13% to 18% quarter-on-quarter in the third quarter of 2026 (according to a single source). Even if this is just the assessment from a single institution, it was quickly interpreted by market sentiment as a signal that "price increases are on the way"—more critically, this discussion is about "traditional DRAM" rather than the already heated high bandwidth memory (HBM), indicating that the demand overflow brought by AI is spreading from the most upstream components like training cards and HBM to the entire supply chain of server memory and general DRAM. Large model training needs to fill racks with GPUs, which requires accompanying sufficient high-capacity DRAM; the inference side must accommodate vast online requests, stacking more end-side and edge scenarios, with consumption of capacity and bandwidth far exceeding the replenishment rhythm under the previous single PC or smartphone replacement cycles, resulting in an elongated demand curve that is more continuous.

For an industry like storage, which has always been highly cyclical, a price jump of 13%-18% within a quarter means that once realized, both the speed of profit recovery and cash flow will be exponentially magnified, and the industry's "high operational leverage" attribute will begin to work in favor of shareholders. Historically, price upswings usually accompany profit recovery and increased capital expenditures among manufacturers, but this time it is compounded by the AI narrative: the capital market will write this round of prosperity into valuation models ahead of time, prepaying higher price-to-earnings and price-to-book ratios to top manufacturers, providing a story basis and pricing anchor for subsequent expansion plans, equity financing, and various long-term projects. Under this chain of "price expectation - profit recovery - valuation uplift - expansion and financing," AI-driven DRAM price increases are no longer seen as a brief replenishment but more like a new cycle starting point capable of supporting years of capital spending.

1.5 Trillion Yen Bet on Japan: Micron's Capacity Gamble

While the capital market writes the prosperity of AI storage into valuations, Micron has turned the story into concrete—launching a total investment of approximately 1.5 trillion yen, equivalent to about 9.3 billion USD, for an expansion project in Japan, with a groundbreaking ceremony held recently. This is not a simple technical reform but a gamble on future AI storage demand: the Japanese Ministry of Economy, Trade and Industry plans to provide up to 500 billion yen in financial support to include this production line in key support projects for semiconductor and AI-related industries, thereby signaling Japan's intention to secure an important position in the autonomy of AI chips and storage supply chains.

But the money invested will not immediately convert into spot supply to alleviate price hikes. According to public planning, the new production line for Micron's Japanese expansion project will not begin shipping until the summer of 2028 at the earliest; there is a several-year gap between the current storage price cycle propelled by AI demand and the actual landing of new capacity. For Micron, this is a trade-off of potential supply years later in exchange for high prosperity and policy support in the current cycle; for Japan, it means accepting short-term price fluctuations in exchange for a seat at the table in long-term AI storage discourse with a national strategy, while this mismatch of time also plants uncertain but highly tense foreshadowing for subsequent supply-demand evolution and price trends.

One of the Largest ADRs in History: Why SK Hynix is in a Hurry to Finance

Unlike Micron, which receives national subsidies and slowly builds factories, SK Hynix is directly betting its chips on the capital market. Its plan to issue ADRs in the United States appears to "only" involve new shares not exceeding 2.5% of total share capital, yet it is interpreted by multiple parties as potentially ranking among the largest stock offerings in history. In the issuance terms, the 0.5% basic underwriting fee is merely a "base salary," while the real highlight is the distribution of floating rewards based on the issuance results—underwriters earn more if they sell more. This design essentially maximizes the risk appetite of global investors: as long as the market is willing to pay for the AI storage narrative, SK Hynix is ready to open the floodgates in the primary market.

Why the urgency? Because increasing production of high-performance DRAM and related products is not a business where one can decide today and produce an extra wafer tomorrow. The demand for memory bandwidth and capacity from AI training and inference clusters is pushing storage manufacturers to the start of a new long-cycle capital expenditure, and building new production lines and transforming processes require significant investments, with the need to complete layouts before the turning point of prosperity. Choosing to issue ADRs in the United States is both a connection to the nearest incremental capital for AI computing power demand and a signal binding global investors to the long-term imagination of "AI storage" with massive financing—this money is not only for its own capacity expansion but also to leverage the entire sector, ultimately pointing to one question: can the market absorb the future demand that has been pre-paid?

Chinese VC Shifting Focus to Physical AI: Funding Shifts from Models to Hardware

While upstream manufacturers are busy leveraging in the capital market, local VCs in China are quietly rewriting their betting directions. Analysis released by Serenity on July 4, 2026, shows that in Chinese private VC funding, large models and LLM-related fields account for approximately 23.56 billion USD, while AI infrastructure and technology layers received about 15.74 billion USD. These two figures outline the basic pattern of the previous stage: "first pile up the models, then talk about landing"; the software side still occupies the top of the funding pyramid, but the base supporting its operation has secured enough weight in chips.

The real turning point lies in another judgment provided by Serenity: institutional funding is flowing massively into physical AI and world models, with investment focus shifting from pure software algorithms to embodied intelligence and hardware elements closer to the physical world. The brief did not provide a complete figure for the funding volume in physical AI but clearly stated that this new track is absorbing incremental "bullets"—from cloud computing power to robots and terminal devices in the physical space, VCs are beginning to ask, "which layer can truly turn inference capacity into sustained cash flow." As downstream betting on embodied intelligence and world models increases, it means longer cycles and higher densities of computing power and storage consumption are being written into business plans.

This forms a mutually reinforcing closed loop with the expectations of upstream production expansion and price increases: on one side is Trendforce's forecast of a 13% to 18% DRAM price increase in the third quarter of 2026, and Micron's investment of approximately 1.5 trillion yen in new production lines in Japan, along with SK Hynix seeking among the largest-scale ADR financing in history, all trying to lock in capacity and funds ahead of the demand explosion; on the other side, Chinese VCs are moving their funds from "more competitive parameters" LLM to physical AI and infrastructure that "need more chips and storage piled up." The upstream sees the long-term project pipelines of the downstream, giving more reasons to ramp up production for future demand; VCs observing upstream expansion and financing moves are also more willing to set higher valuations for hardware and embodied intelligence projects. This cycle of mutual verification of funding and capacity forces the entire AI industry chain to continue down the path towards a "heavier physical world" under higher stakes and stronger path dependence.

From Demand to Capacity: The Next Scene in the AI Storage Chain

Looking ahead along this chain, the expectation of a 13%-18% quarter-on-quarter price increase for traditional DRAM in the third quarter of 2026 has become a quantitative anchor for this round of "demand pushing prices": upstream manufacturers see that AI computing power and storage continue to squeeze, daring to assume higher future prices. Micron locks this expectation with a Japan expansion project totaling approximately 1.5 trillion yen, with new capacity expected to land no earlier than the summer of 2028. The highest subsidy of 500 billion yen from the Japanese government upgrades corporate judgment to a national industry bet; in response, SK Hynix chooses to raise funds through ADR in the United States, offering no more than 2.5% of total share capital, along with a 0.5% basic underwriting fee and floating rewards, aiming for an issuance that is expected to rank among the largest in history, filling the "bullets" for AI storage expansion and R&D in the coming years in advance. On the other end, the 23.56 billion USD in LLM investment and 15.74 billion USD in AI infrastructure funding reported by Serenity, along with the trend of funding shifting toward physical AI and world models, provide dual backing for this round of "price increase - financing - expansion" from the demand side: upstream believes the downstream will use these chips, and VCs believe upstream will build factories. The cycle of demand - price - investment - capacity is amplified under a more globalized and financialized framework. However, prepaying the demand for the next few years onto the balance sheet also means the risks are shifted backward overall: once new capacities in places like Japan are concentrated in release around 2028, or if geopolitical issues and industrial policies rewrite supply chain routes, this chain may rapidly switch from "grabbing goods" to "destocking." At that time, who will be under pressure from high leverage and high capital expenditures, and who can cross the new round of fluctuations relying on product structure and technological barriers will be reshuffled; before that, the chips around AI-related storage and physical AI are likely to remain a battleground where global funds are most reluctant to leave easily.

Join our community to discuss and become stronger together!
AiCoin exclusive Hyperliquid benefits: https://app.hyperliquid.xyz/join/AICOIN88
AiCoin exclusive Aster benefits: https://www.asterdex.com/zh-CN/referral/9C50e2
On-chain Telegram community: https://t.me/AiCoinWhaleData
On-chain community: https://www.aicoin.com/link/chat?cid=N6OVMor5g
AiCoin on-chain Twitter: https://x.com/aicoinwhaledata

免责声明：本文章仅代表作者个人观点，不代表本平台的立场和观点。本文章仅供信息分享，不构成对任何人的任何投资建议。用户与作者之间的任何争议，与本平台无关。如网页中刊载的文章或图片涉及侵权，请提供相关的权利证明和身份证明发送邮件到support@aicoin.com，本平台相关工作人员将会进行核查。