Micron's Earnings Report Coming Next Week: Profits Expected to Rise Tenfold, But the Market Only Wants a Sixfold Valuation

According to the company's August 26 guidance, fourth-quarter revenue was $50 billion (± $1 billion), with a non-GAAP gross margin of about 86% and non-GAAP earnings per share of about $31 (±.1 USD). The seller's consensus expectation was even higher: revenue of $50.82 billion and adjusted earnings per share of $31.43. For comparison, the company reported revenue of $41.46 billion in the previous quarter (ending May 28), up 345% year-over-year, with a non-GAAP gross margin of 84.9%; Going further to the same period the previous year, the gross margin was about 45.7%.

Análisis de TMGM: noticias de mercados financieros, calendario económico e información del mercado

Growth is supported by two main factors. In the third quarter, DRAM revenue reached $31.3 billion, up 343% year-on-year, accounting for about 76% of total revenue, with average selling prices up over 60% quarter-on-quarter; NAND revenue was $9.9 billion, up 361% year-on-year, accounting for about 24%, with average selling prices rising over 80% quarter-on-quarter. By business, cloud storage revenue was $13.8 billion, up 78% quarter-on-quarter, and core data center business was $11.5 billion, more than doubling quarter-on-quarter.

On September 22, Micron closed up 5% at $1,096.16, with a turnover of $30.885 billion, ranking first among U.S. stocks that day, with a market value surpassing $1 trillion. Alongside this is another fact: based on expected earnings for fiscal year 2027, its forward P/E ratio is about 6.4 times, while the S&P 500 is about 21 times.

This is not a pricing mistake; memory chips are close to commodities, and DRAM from different manufacturers can be interchangeable, with prices determined by supply and demand. When prices are high, all manufacturers expand production; A few years later, capacity is ramped up simultaneously, oversupply causes prices to fall. Historically, DRAM has repeatedly slipped from peak profit margins to near zero operating profit within 24 months. Therefore, the market habitually gives memory companies the lowest multiples when profits are highest: it assumes these profits cannot be sustained

The bears' statements are even sharper. Michael Burry disclosed in early July that the stock has fallen more than 30% 34 times over the past 42 years, with a median return on invested capital of only 4% and return on equity of 7%, "destroying capital one quarter every three quarters." He also pointed out that Micron's stock price is above the 200-day moving average at more than any point since 1984.

"This Time Is Different"—What Proof Do You Need?

Micron's answer is to turn spot business into contracts. The company has signed 16 strategic customer agreements, covering about 20% of DRAM capacity and 33% of NAND capacity. These are binding underwriting contracts, typically extending from 2026 to the end of 2030, specifying both purchase volume and price floor and ceiling; Based on minimum price and minimum shipment volume, the guaranteed minimum income for performance obligations totals about $100 billion, customer deposits and financial commitments are about $22 billion, of which about $18 billion is cash deposits. Management expects that after full execution of the agreement, it will contribute more than half of the company's revenue.

Additionally, supply and demand sides are supporting this claim. Micron's HBM capacity for 2026 is fully sold out, with related underwriting contracts totaling about $100 billion; HBM uses about three times the wafer consumption per unit of HDRAM, effectively draining a considerable amount of DRAM capacity. The pricing side is even more direct: On September 20, TrendForce sharply raised its NAND price forecast, raising the sequential growth rates for Q3 and Q4 2026 and Q1 2027 to 21%, 17%, and 15% respectively, compared to previous forecasts of 18%, 2%, and 0%. The reason for the increase is that enterprise eSSD demand from U.S. cloud providers is nearly doubling from previous estimates. During the same period, most DRAM products saw Q3 contract prices rise by 20% to 30% quarter-on-quarter. In the spot market, 16Gb DDR5 averaged $57 and DDR4 3200 averaged $84

However, equating scarcity directly with perpetual demand is still premature. TrendForce predicts DRAM tightness will persist until 2027, but NAND supply is expected to ease in the second half of 2027; SK Hynix has given a longer timeframe, suggesting supply and demand will only balance by 2030. There is a three-year gap between the two judgments, which is the physical source of the divergence between bulls and bears.

More Noteworthy Points

First is the guidance for the next quarter. The market expects revenue of $56.9 billion, a 317% year-over-year increase, and earnings per share of $34.95 for the first quarter of fiscal year 2027; TD Cowen expects the company's guidance to be around $37, higher than the consensus market expectation.

Second is where gross margin stands. This quarter's guidance is about 86%, compared to 84.9% last quarter. Krish Sankar of TD Cowen believes the gross margin expansion cycle has completed about 80%, with the peak expected in the second quarter of the 2027 calendar at around 89%, after which it stabilizes above 80%. When gross margin approaches 85%, the room for prices to continue beating expectations narrows—the flip side of low valuation is low tolerance for error.

Third is the number of new contracts signed. If the 16 agreements continue to expand, the market may gradually accept "cyclical weakening" and shift the valuation method from cyclical stocks to stable growth stocks.

There are two other risks to clarify. First, most contracts have price caps, and if prices rise beyond the threshold, companies cannot fully benefit from them; At the same time, DRAM bit shipment growth is expected to slow from mid-to-high double digits in 2026 to 15% to 20% in 2027. Second, Micron factories in Taiwan face labor risks; local unions require 15% of global operating profits to be used for profit sharing, and as of September 15, preparations for a strike are still underway. Taiwan is Micron's largest production base for DRAM and HBM.

Why It Has Become A Key Validator For The AI Market

Looking at the bigger market, Micron's uniqueness lies in its rare "physical constraint" segment in AI transactions. Morgan Stanley predicts that by 2027, storage and interconnect chips will account for about 50% of the capital expenditure of the world's top 14 cloud service providers; The same agency estimates that from 2026 to 2028, the cumulative power demand for U.S. data centers will reach 97GW. After deducting projects under construction and available grid capacity, the gap is 57GW, about 50% higher than previous estimates. Storage and power are two unavoidable hurdles in computing power expansion.

This AI rally itself has already shown cracks. On September 22, the Nasdaq closed up 0.45%, hitting a new high, but the number of S&P 500 stocks hitting 52-week lows has already surpassed those hitting 52-week highs. The last time this happened was in December 1999, on the eve of the dot-com bubble peaking. During the same period, the 10-year Treasury yield hovered around 4.96%, and the market priced a December rate hike at 90%; Credit markets and stocks are also diverging. As index gains increasingly depend on a handful of large-cap tech stocks, any link in the validation chain that fails to meet expectations can be magnified.