Written by: Rita

The discussion around the storage chip cycle is shifting from "has it peaked?" to "what's different this time?"

In the global technology webinar held in August, Morgan Stanley connected three topics: storage, financing of hyperscale manufacturers, and SpaceX, with the central theme being that the capital formation for AI infrastructure is reshaping the financing methods and pricing logic in the semiconductor and communication industries. In the storage sector, Morgan Stanley put forward three key judgments: how long-term agreements (LTAs) are changing the pricing mechanisms, where the current cycle stands, and the real meaning of inventory. LTA refers to long-term supply contracts in which storage manufacturers and customers lock in capacity and pricing mechanisms in advance, replacing the traditional spot pricing model.

LTAs Transition from Intent to Disclosure

Long-term agreements (LTAs) represent the most significant structural change in this storage cycle. Samsung has clarified specific numbers for the first time: it plans to include 60% to 70% of its capacity in long-term agreements with a rolling five-year term. There are five signed contracts, and five are in the final negotiation stage, with customers including AWS, Microsoft, Google, Meta, and Oracle. SK Hynix has signed about 10 agreements but has not disclosed the allocation ratio. Micron aims to include more than 50% of its revenue in LTAs, having signed 16 so far, with customers including four large and three medium-sized manufacturers, and has received about $22 billion in prepayments and commitments. SanDisk aims to lock in more than 50% of its wafer capacity for fiscal year 2027 and about three-quarters of its capacity for fiscal year 2028, having signed 8 customers with minimum revenue commitments reaching $93.9 billion and remaining performance obligations (RPO) of $59.8 billion. Kioxia aims to have 50% of its sales in LTAs by 2028, with pricing principles of "price priority and profit priority."

Morgan Stanley pointed out that the focus of industry discussions has shifted. Previously, the question was "Should we sign an LTA?" Now, the question is "How many should we sign, for whom, and under what pricing mechanism?" The pricing mechanism is shifting from purely spot pricing to a two-way protection mechanism with price floors and ceilings, and customers are now required to provide upfront deposits or equipment guarantees.

DRAM and NAND Pricing Nears Peak

The DRAM contract price for 3Q26 increased approximately 15% quarter-on-quarter, lower than the previously expected 20%; NAND increased approximately 20%, also with a narrowing rate of increase. Morgan Stanley expects the price increase in 4Q26 to further slow, with DRAM contract prices expected to fall from cycle peaks year-on-year.

However, Morgan Stanley believes this does not mean the cycle is over. The DRAM NTM price-to-earnings ratio usually leads the NTM earnings per share by about two months, while the breadth of memory earnings adjustments has receded from extreme levels since the end of June. This typically occurs in the mid-to-late stages of the earnings upgrade phase, not at the end of the cycle.

The changes in pricing mechanisms are also noteworthy. The price protection mechanisms under the LTA framework are replacing the purely spot pricing game. Even if spot prices drop, LTA customers still execute at contracted prices, with price floors providing a revenue baseline for suppliers, while price ceilings offer cost certainty for customers. This bilateral protection theoretically lowers the earnings volatility for storage manufacturers compared to any previous cycle.

Inventory Increases but Structure is Different

DRAM and NAND inventory weeks have increased in 3Q26. Supplier inventories remain low, but module manufacturers and downstream customers have built up DRAM and NAND inventories over the past few quarters.

This appears to be "inventory accumulation," but Morgan Stanley believes it's necessary to distinguish between proactive builds and passive stockpiling. Under the LTA framework, customers locking in capacity in advance is a form of proactive inventory build. The upfront deposits and irrevocable purchase commitments in the contracts mean that this inventory is supported by actual demand rather than speculative hoarding.

This is fundamentally different from the passive inventory accumulation observed in 2018 and 2022. In the previous two cycles, inventory build-up was accompanied by a price crash and suppliers being forced to cut production. Currently, LTAs lock in price floors and minimum purchase quantities, meaning that even if end demand fluctuates, suppliers' cash flow visibility is significantly higher than before.

SpaceX and Credit Market Validate Capital Intensity

SpaceX's Q2 results exceeded expectations, and Morgan Stanley maintains an overweight rating and a target price of $300. The company's goal is to reach a total ARR of at least $100 billion by the end of 2026, with a projected single-quarter ARR of about $22 billion in 4Q26 (previously forecasted at $19.1 billion). Starlink V3 satellites will begin deployment from the 14th flight of Starship, expecting to have 1,000 satellites in orbit by Q2 2027, compared to a previous prediction of 840. Capital expenditure guidance shows that the second half of the year will be flat quarter-on-quarter but approximately $18 billion higher than previous forecasts. Morgan Stanley's SOTP target price breaks down into: Space business $80+, Connectivity business $118, X and Grok business $6, Enterprise AI business $166. The pullback after the lock-up period for shareholders is viewed by Morgan Stanley as an entry opportunity.

In the credit market, the five major investment-grade TMT companies have uninitiated lease commitments totaling over $1 trillion. The credit ratings of Microsoft, Nvidia, Google, Meta, and Amazon range from A- to Aaa, with 5-year CDS spreads ranging from 57 basis points for Amazon to 171 basis points for SpaceX. Morgan Stanley's credit research team noted that contract and chip-backed debt financing is becoming a new focus. Tech giants are increasingly leveraging their balance sheets to provide financing guarantees for the AI supply chain rather than relying solely on equity financing or their cash flow. This further validates that the capital intensity of AI infrastructure investments is changing the financing structure of the entire technology industry.

The cyclical logic of storage chips is being rewritten by LTAs, and the financing methods for AI infrastructure are being reshaped by the balance sheets of tech giants. Morgan Stanley believes that the market's pricing of the storage cycle may still be trapped in the old framework of traditional commodity cycles, and this structural change implies that valuation methods may need to be adjusted.

Disclaimer

This article is a compilation and interpretation of third-party brokerage research reports (Morgan Stanley, August 11, 2026) by Tide Research, combined with publicly available market information. The ratings, target prices, profit forecasts, and related judgments quoted in the text represent the views of the analysts of that brokerage firm and do not represent the views of Tide Research nor constitute any investment advice.

The market is risky, and decisions should be made independently. This article should not be used as the basis for buying or selling any securities.