Micron's HBM chip products have sold out with gross margins reaching 81% peak; the June 24 earnings report will further confirm supply constraints.
Micron will enter its June 24th earnings release with a memory business that Wall Street no longer knows how to value. The old commodity cycle playbook cannot explain the 81% gross margin guidance, HBM contracts signed years in advance, or the stock that has already surged significantly before the earnings announcement.
Micron's next earnings call is not a routine update. It is a test of whether investors should still view the company as a cyclical memory supplier, or should view it as one of the companies profiting from AI infrastructure buildout. If you own the stock, compete with it, or depend on chips entering NVIDIA-class servers, this distinction matters.
The disclosed numbers are substantial enough. Investor's Business Daily reported that Micron's adjusted earnings per share for the second fiscal quarter ended May was $12.20, with revenue of $23.86 billion, exceeding FactSet's expectations ($9.19 per share, revenue $19.97 billion). For the third fiscal quarter, Micron projects adjusted earnings per share of $19.15 and revenue of $33.5 billion. MarketWatch also noted that the company's adjusted gross margin target for the May quarter was 81%, up from 74.9% in the February quarter.
These are not Micron's usual numbers. For years, an honest assessment of Micron was simple: memory prices rise, memory manufacturers overproduce, prices fall, and shareholders remember why the market never gave DRAM the high valuations software companies receive. This history hasn't disappeared. But high-bandwidth memory has changed the rhythm of the debate.
High-bandwidth memory is a part of the AI server story that you cannot easily skip over. NVIDIA's current Blackwell systems rely on HBM3E, and next-generation accelerators push memory demands even higher. When Microsoft, Google, Amazon, and Meta commit tens of billions of dollars to data centers, they are not just buying GPUs. They are also pulling scarce memory capacity ahead of schedule from SK Hynix, Samsung, and Micron.
The key point is: you cannot solve HBM shortages by calling a supplier just a quarter before you need the product. You need to sign contracts in advance, commit capital, and hope your allocation stays reserved. Tom's Hardware reported in December that Micron said DRAM supply constraints would persist into 2026 and beyond, with the company establishing multi-year supply commitments. This is why this earnings report is more important than headline-level beats or misses. The real question is how much of 2027 has already been booked.
Micron remains the smaller HBM player. SK Hynix leads the market, Samsung remains a powerful force, and Micron must prove that its market share gains are not just a lucky byproduct of everyone being short simultaneously. But when customers have urgent demand and supply is locked in for years, even ranking third can still be a very profitable position.
The company's spending indicates it believes the shortage will continue. PC Gamer reported in January that Micron signed an $1.8 billion agreement to acquire Powerchip Semiconductor Manufacturing's Taiwan facility, including a 300,000-square-foot 300mm wafer fab cleanroom, although the facility is not expected to make a material contribution to DRAM output until the end of 2027. Tom's Hardware later reported that Micron is beginning construction of a $24 billion 3D NAND fab in Singapore, while also mentioning a separate HBM assembly project there. In the United States, Micron has linked its expansion plans to new and existing facilities in Idaho, New York, and Virginia.
This time lag is the underpromoted part of the bullish argument. New memory capacity is expensive, slow, and operationally unforgiving. If AI demand continues to grow before new supply arrives, Micron will gain pricing power without having to pretend it has become a different type of company. If demand falls faster than expected, investors will discover how much protection those multi-year commitments actually provided.
Barron's reported last week that Micron will release earnings after the close on June 24, with the stock having surged significantly beforehand, raising expectations. The same report noted that the stock reached all-time highs multiple times before 2026, then fell along with the broader semiconductor sector. It's an awkward situation: investors are not walking into this earnings report with skepticism. They have entered after the stock has been repriced.
So the quarter itself may not be enough. Revenue numbers in line with guidance and another solid margin number will confirm that the AI memory cycle is still running. What you should watch for is management commentary on HBM supply, contract duration, customer concentration, and 2027 capacity. If the answer remains scarcity, Micron's old valuation argument becomes weaker. If the answer becomes muddier, the market will notice quickly.
Micron doesn't need to prove that AI demand exists. NVIDIA, hyperscale data center capital spending plans, and memory shortages have already done that. It needs to prove that this cycle is not only larger in scale, but that the contract terms are more favorable than cycles that hurt investors previously.