The memory industry is witnessing an unprecedented development: major suppliers and hyperscale cloud customers are beginning to sign Long-Term Agreements (LTAs) incorporating partially fixed pricing. This is more than a minor business model adjustment; it directly challenges the fundamental logic that has underpinned the valuation of memory stocks like Micron Technology for decades. With profit volatility set to be systematically reduced, the "super-cycle discount" is no longer justified. This is the core rationale behind UBS's significant price target increase for Micron. According to reports, UBS analyst Timothy Arcuri issued a research note on Tuesday, raising the 12-month price target for the memory chipmaker from $535 to $1,625, the highest on Wall Street. This target implies a potential doubling from the pre-market price around $800, corresponding to a market capitalization of approximately $1.8 trillion. It suggests Micron would solidify its recent entry into the ranks of the world's largest companies and cement its status as one of the most crucial players in the AI investment theme. In Arcuri's view, LTAs are fundamentally altering Micron's profit profile, meaning "the market will begin to assign the stock a more 'normal' valuation multiple." Specifically, this implies a benchmark closer to Nvidia's forward P/E of around 15x, rather than the approximately 5x implied by previous sum-of-the-parts (SoTP) models. Numerically, the model raises fiscal year 2027/2028/2029 EPS estimates to approximately $155, $167, and $117, respectively. Notably, this EPS trajectory remains above $100 even under a scenario assuming a roughly 50% decline in DRAM spot prices by 2029. Cumulative free cash flow over this period is projected to exceed $400 billion. Following the report's release, Micron's shares surged 8% in pre-market trading. The stock closed at $751 on Monday, marking a year-to-date increase of over 160%.

LTAs Are Not Ordinary Sales Contracts

While the memory industry has had "take-or-pay" agreements in the past, they typically locked in volume, not price. The core difference in this round of "enhanced LTAs" is the introduction of a partially fixed pricing framework, with contract durations of 3-5 years, structured as "2+3" (two years fixed + three years variable) or "3+2". Supply chain research indicates that an estimated 20%-30% of the industry's 2027 DDR shipments are expected to be covered by these enhanced LTAs. By company share: Samsung around 30%, Micron around 20%, and SK Hynix around 18%. More critically, hyperscale cloud providers have already secured approximately 60%-70% of the industry's server DDR5 capacity—meaning customers like Microsoft, Google, Amazon, and Meta are proactively ensuring memory supply for the coming years. Why are buyers willing to lock in prices? Two reasons: supply security and enhanced predictability for future data center deployment costs. They are essentially trading price for certainty. For suppliers like Micron, the benefits are equally clear: a smoother revenue curve and a material improvement in return on invested capital across cycles. Model calculations show that compared to a baseline scenario with no volume or price locks, LTAs can compress the peak-to-trough price volatility of DDR by approximately half. This is the direct mechanism allowing Micron to maintain EPS above $100 even during a mild downturn projected for 2029.

HBM Pricing: Preparing to Rebuild Premiums

Beyond LTAs, another key assumption revised upward is the average selling price (ASP) for High Bandwidth Memory (HBM). Previous models assumed a 35% year-over-year increase in Micron's HBM ASP for calendar year 2027; this has been raised to 50%. The underlying logic is that Micron, SK Hynix, and Samsung all intend to re-establish price premiums for HBM starting in 2027—a reversal from the competitive pressure suppressing HBM prices in 2025-2026. Specifically, Micron's HBM ASP is projected to rise from approximately $13.07/GB in calendar 2026 to $18.60/GB in 2027, and further to $23.26/GB in 2028. For HBM shipment volume, the forecast is about 859 million GB for 2026 and roughly 1.361 billion GB for 2027, representing year-over-year growth of about 58%. Corresponding HBM business revenue is projected to reach $25.3 billion in 2027 and approximately $44.7 billion in 2028. Notably, HBM gross margins are also on a sustained climb—the model forecasts an increase from around 66% in 2026 to about 75% by 2028, by which point they would approach or even surpass the profitability levels of standard DRAM.

Supply-Demand Tight Window Extended Beyond Prior Expectations

The duration of the DRAM supply shortage has been extended further: from a previously expected end in Q4 2027 to Q2 2028. The NAND supply shortage window has also been pushed back from Q3 2027 to Q4 2027. This shift is driven by constraints on the supply side: Samsung, SK Hynix, and Micron have all shown less willingness to add new capacity in NAND compared to DRAM—their commitment to new cleanroom capacity for NAND is notably more conservative. This alleviates some market concerns about a premature peak in NAND prices. Nevertheless, NAND price expectations remain significantly weaker than for DRAM: the model projects NAND ASP to decline approximately 56% year-over-year in calendar 2029, while non-HBM DRAM ASP is expected to fall about 45% over the same period. This explains why Micron's projected NAND revenue for 2029 (approximately $31.2 billion) shows a notable decline from 2027 (around $48.5 billion).

Valuing Like Nvidia, Not a Cyclical Stock

The jump in price target from $535 to $1,625 corresponds to a shift in valuation methodology—abandoning the SoTP model in favor of applying a forward P/E multiple of approximately 15x. This multiple is anchored to a calendar 2029 EPS of about $117, discounted back to 2028 using a cost of capital of around 12%. The logic for using 2029 earnings as the anchor is that the model already assumes a mild DRAM downturn by then, yet with LTAs providing a floor, EPS remains above $100. This year is seen as more accurately reflecting Micron's "through-the-cycle earnings power." The 15x P/E benchmark is based on the judgment that "there is no reason Micron should trade at a significant discount to Nvidia on a P/E basis." The current weighted average forward 2026 P/E for semiconductor peers (including Nvidia and Broadcom) is approximately 26x. In contrast, even at the new target price, Micron's 2027 P/E would only be about 10x. The potential for valuation expansion hinges on the market accepting this structural re-rating thesis. A downside scenario (target $250) assumes a 33% year-over-year decline in HBM shipments, a 70% crash in non-HBM DRAM prices, a 77% drop in NAND prices, and applies an 11x P/E. An upside scenario (target $1,850) corresponds to a 16x P/E, with non-HBM DRAM prices declining 42% and HBM volume and pricing outperforming the base assumptions.