The semiconductor industry used to treat memory chipmakers like commodity farmers. You build a massive factory, flood the market with plastic rectangles, and pray prices don't collapse before you turn a profit.
That old playbook is dead. Discover more on a related issue: this related article.
Micron Technology just dropped its fiscal third-quarter 2026 earnings, and the numbers are honestly dizzying. We aren't just talking about a standard earnings beat. Micron pulled in a record $41.46 billion in revenue—quadrupling its performance from the same period last year.
But the headline that should make every board member at Nvidia and Meta sweat is the profitability. Micron posted a staggering non-GAAP gross margin of 84.9%. More reporting by The Verge explores related views on the subject.
Let that sink in. A hardware manufacturer, heavy with physical factories and massive silicon wafers, is keeping 85 cents of every single dollar it brings in. That's not just higher than Nvidia’s hardware margins. It beats out the software-driven, ad-heavy machine of Meta. Micron is officially tech’s new margin king, and it's driven entirely by a brutal, structural AI memory crisis that shows no signs of slowing down.
The Shortage Inside the Shortage
If you want to understand why Micron is suddenly printing cash faster than a central bank, you need to look at what goes into an AI data center. Everyone knows about Nvidia’s graphics processors. They're the brains. But a fast brain is completely useless if it's starved of data.
That’s where High-Bandwidth Memory (HBM) comes in.
HBM isn't your standard desktop RAM. It’s a hyper-dense, complex stack of DRAM layers interconnected vertically. When Nvidia sells a Blackwell or a next-generation Vera Rubin architecture platform, it doesn't just need logic chips; it needs piles of HBM bolted directly onto the accelerator.
Right now, the world can't make enough of it. Micron CEO Sanjay Mehrotra bluntly stated during the earnings call that the company can only fulfill half to two-thirds of actual customer demand for HBM. The company's entire 2026 allocation is completely sold out.
To lock down what little supply exists, big tech hyperscalers are doing something unprecedented. They're paying upfront. Micron has already collected $22 billion in customer cash deposits and prepayments just to guarantee shipments. When companies are handing you billions in cash before a product is even manufactured, you don't have customers anymore. You have dependents.
Why This Supercycle Is Different
In past semiconductor cycles, high margins were a temporary glitch. A manufacturer would figure out a new process, print money for two quarters, and then competitors would ramp up production, trigger a supply glut, and crash the price.
This time around, hitting the brakes on supply isn't so simple. Manufacturing HBM requires roughly three times the wafer capacity of conventional DRAM. Every single wafer diverted to build high-margin AI memory means three wafers taken out of the supply chain for standard computers, smartphones, and servers.
The spillover effect is a total supply crunch across the board. Look at the numbers from Micron’s latest report:
- DRAM Revenue: Hit $31.3 billion, making up 76% of total revenue.
- Pricing Power: While actual bit shipments only grew by low-single digits, the average selling price surged in the low-60% range quarter-over-quarter.
- NAND Flash: Surged 99% sequentially to $9.9 billion as AI model training requires massive, ultra-fast storage arrays.
This isn't just a temporary bump. It’s a massive structural shift. Micron is already ramping its next-generation HBM4 chips, which are scaling twice as fast as the previous generation. HBM4 revenue alone has already crossed the billion-dollar mark.
The Trillion-Dollar Question for Investors
Whenever a cyclical stock goes vertical, you have to ask when the music stops. Micron crossed the one-trillion-dollar market capitalization milestone in late May, trading at prices that assume these historic margins are the new normal.
But let's be realistic. Shortage economics are driving this 85% gross margin. If big tech's capital expenditure on AI infrastructure slows down even a fraction, or if rivals like SK Hynix and Samsung manage to successfully scale their own HBM4 lines faster than expected, those margins will compress quickly. We saw a preview of this volatility just a day before earnings, when a messy rumor about South Korean production shifts sent Micron shares tumbling 13% in a single afternoon, before the blockbuster earnings report erased the losses.
For now, the bears don't have a leg to stand on. The market size for HBM is projected to expand at a 40% compound annual growth rate through 2028, ballooning from $35 billion last year to over $100 billion.
What You Should Do Next
If you're managing an enterprise IT budget, running data center procurement, or investing in the semiconductor space, the strategy has to shift immediately.
- Lock in multi-year agreements: If you rely on enterprise memory or flash storage, waiting for prices to drop is a losing game. Supply constraints are projected to last well past the end of the year. Secure your allocations now, even if it requires upfront commitments.
- Re-evaluate hardware life cycles: With standard DRAM and NAND pricing climbing significantly due to the wafer capacity squeeze, upgrading client devices (like corporate laptops) will become noticeably more expensive. Extend current hardware lifespans where possible to avoid premium component costs.
- Watch hyperscaler capex like a hawk: Micron's forward guidance of $50 billion for the next quarter proves the AI build-out is still accelerating. Keep a close eye on quarterly capital expenditure reports from Microsoft, Alphabet, and Meta. They are the ultimate engine behind Micron's newfound royalty status.
