Micron Trades at 6 Times Next Year's Earnings. The $38 Billion of Capacity That Ends This Cycle Doesn't Open Until 2028. |

Memory specialist Micron Technology (MU +0.29%) trades near $878 as of this writing, at about 20 times its trailing earnings but only about 6 times the earnings analysts expect over the coming year. Earnings are climbing so fast that next year’s expected number dwarfs the trailing one. And the market is paying about 6 times for it because it assumes the good times end soon.

Memory has always worked that way: High prices attract new supply, and new supply ends the boom.

But last Friday, the supply that is supposed to end this one got a price tag and a schedule. SK Hynix committed about $38 billion to two new fabs — and the first of them doesn’t open a clean room until December 2028.

A boom still accelerating

The earnings the market is discounting are not hypothetical. In its fiscal third quarter (ended May 28, 2026), Micron’s revenue more than quadrupled year over year to $41.5 billion, up from $23.9 billion just one quarter earlier. Gross margin ran 84.6%, compared to 74.4% in fiscal Q2 and 37.7% a year ago. Operating cash flow, meanwhile, more than quintupled year over year to $25.4 billion.

Management expects more. Guidance calls for fiscal fourth-quarter revenue of about $50 billion, gross margin of about 86%, and earnings per share of about $30.73.

Annualize that guided quarter alone and the stock trades at about 7 times earnings.

Data center demand is driving all of it. Micron’s data center revenue exceeded $25 billion in fiscal Q3 (more than $100 billion annualized).

And in prepared remarks for its June earnings call, the company said industry demand for DRAM and NAND “continues to significantly exceed industry supply.”

The supply response now has a date

For a memory stock, what matters is when supply arrives. On Friday, SK Hynix’s board approved 54 trillion won (about $38 billion) for the two new fabs. The bigger piece, 35.2 trillion won, goes to a DRAM plant in Yongin, South Korea, called Y2. The rest, 19.1 trillion won, funds a NAND plant in Cheongju called M17.

I’d argue the schedule matters more than the dollar figure. M17 breaks ground in February 2027 and opens its first clean room in December 2028. Y2 doesn’t break ground until July 2027, and its first clean room opens in June 2029.

And a first clean room typically marks the start of equipping a fab, not the start of volume output. Capacity decided on today, in other words, is 2028-and-beyond capacity.

That squares with what Micron itself has been saying. In the same June remarks, Micron said it expects tight conditions “to persist beyond calendar 2027 as a result of AI-driven demand across all segments coupled with structural supply constraints.”

Of course, some new supply lands sooner. SK Hynix’s first Yongin fab, already under construction, is slated to open its first clean room in February 2027. Micron itself is spending at record levels, too ($7.1 billion of capital expenditures in fiscal Q3 alone).

However, even with all of that in motion, Micron still expects tightness through 2027. The fabs approved last week arrive after that.

Contracts built for the turn

Micron has also spent this boom locking in what happens when it ends. The company has signed 16 strategic customer agreements — take-or-pay contracts, meaning customers commit to buying specific volumes over multiple years. The agreements typically run five years, from calendar 2026 through the end of calendar 2030.

Together, they cover roughly 20% of Micron’s DRAM volume and about a third of its NAND volume over that period. Management expects half or more of company revenue to eventually fall under these agreements.

The largest of them generally carry price ceilings set at calendar second-quarter 2026 market prices, with price floors that hold through the term. In a downturn, those floors should put a boundary under how far Micron’s contracted revenue can fall.

CEO Sanjay Mehrotra said in the June earnings release that these agreements “will significantly enhance the durability and predictability of Micron’s strong financial performance.”

Ultimately, the cycle will still turn. Memory cycles always have, and record prices are financing the capacity that could end this one. But at a valuation of about 6 times expected earnings, the stock is priced as if that ending is close.

The construction schedules the industry itself has published put the big additions in 2028 and 2029, and Micron’s contracts run through 2030. To me, the business looks likely to keep earning at something like this pace longer than the market is paying for.

The main risk isn’t the construction schedule — memory prices could fall without a single new fab opening if artificial intelligence (AI) demand cools. But based on what the industry has committed to build, the turn arguably sits further away than the price assumes.