Sponsored by edgeful

Alright, let's talk about what just happened with Micron, because this is one of those moments where the numbers say one thing and the stock does the complete opposite - and honestly, both sides have a point.

Micron just reported $23.9 billion in revenue. That's up nearly 200% year-over-year. EPS came in at $12.20 versus the Street at $9.19. Gross margins? 74.9% - in a memory company. For context, memory companies historically lived in the 30-50% gross margin world. This is software-territory profitability from a company that makes DRAM and NAND chips.

And then the guide. The guide was absurd. $33.5 billion in revenue for next quarter, EPS of $19.15 versus the Street at $12.03, and gross margins guided to 81%. That's higher than NVIDIA's gross margins right now. Let that sink in for a second.

So why is the stock down 5%?

Here's the thing about memory stocks - they have a long, painful history of peaking before the numbers do. The best quarter is rarely the best time to own the stock. Investors know this, and it's baked into the DNA of anyone who's traded this sector for more than a cycle.

The overnight debate was essentially: is 81% gross margin the ceiling? And if it is, does it matter that earnings are still going up if the rate of improvement is slowing? Memory stocks trade on the second derivative, not the first. The moment the market thinks margins have peaked - even if they stay at 80% for another year - the stock can de-rate hard.

Add to that: DDR5 spot pricing has been flattening. Contract pricing is still strong, but spot is the leading indicator, and it's sending a different signal than what the contract market is saying. That divergence matters.

This wasn't just an HBM story. That's the important part people might miss. Yes, HBM is the headline, and yes, MU started volume shipping HBM4 for NVIDIA's Rubin platform this quarter. But the real story was pricing power across the entire portfolio.

DRAM pricing was up mid-60s percent quarter-over-quarter. NAND pricing - and this is the sleeper - was up in the high-70s percent range. Bit growth was modest. This was almost entirely a pricing quarter, not a volume quarter. That distinction is critical because it tells you the supply/demand balance is genuinely tight, not just in one product but across the board.

NAND was the standout nobody expected. KV cache offload for AI inference, vector databases, Gen6 SSD share gains, and an ongoing HDD shortage are all pulling demand through data center SSDs. AI storage is no longer theoretical - it's showing up in the numbers now.

Management's message was clear: demand still far exceeds supply, and material new capacity from major cleanroom additions doesn't show up until fiscal 2028. That's two years out. They raised FY26 capex to over $25 billion from $20 billion, and signaled a "meaningful step-up" for FY27 - some estimates put that at $40-45 billion.

Now, here's where it gets interesting. Bulls hear that and say: great, supply stays tight, pricing stays elevated, earnings are durable. Bears hear that and say: this is exactly how every overbuild starts. "Visibility is great" turns into "the industry overbuilt" faster than anyone expects in memory. Every. Single. Cycle.

The wildcard is the first-ever 5-year Strategic Customer Agreement Micron just signed. That's not a typical one-year LTA - that's a customer saying memory is mission-critical infrastructure for the next half decade. If more of these get signed, it fundamentally changes how you model downside risk. But management gave almost no detail on terms - fixed vs. variable pricing, guaranteed volumes, penalty structures - so investors can't yet underwrite it as downside protection.

Let's run the math both ways.

Bulls are modeling around $120 in FY27 EPS. At 5-6x, that's $600-720. The stock closed at $462. That's 30-55% upside, and even at a "peak multiple" the risk/reward looks solid if you believe these earnings are sustainable.

Bears are at roughly $80 FY27 EPS with a material drop in FY28 as margins normalize back toward 70% or below. At 3-4x peak earnings, you get $240-320. From $462, that's meaningful downside.

The gap between those two scenarios is enormous, and it comes down to one question: is this a cycle or a structural shift? If AI demand really does keep memory tight through CY27 and beyond, and if these SCAs provide genuine downside protection, then this is a different animal than past memory cycles. If it's still ultimately cyclical - just a bigger, longer cycle - then the bears' math works eventually.

The honest answer is this is hard to have a strong view on. The numbers are undeniably spectacular. The demand backdrop from AI is real and broadening. But 81% gross margins in memory is historically unprecedented, and the spot market is already sending early signals of stabilization in DRAM.

If you're looking for the cleaner way to play the memory theme right now, there's a reasonable argument that names like Samsung or Sandisk/Western Digital offer more room for earnings revisions and margin expansion from here - they haven't had the same re-rating MU has already gotten. There's also a read-through to semicap equipment names like AMAT and LRCX from the massive capex ramp, though the shift toward EUV may create some mixed signals there.

Bottom line: Micron just proved that AI-driven memory demand is broader and more durable than almost anyone modeled. But the stock's reaction tells you the market is already asking the "what's next?" question. And in memory, that question always comes sooner than you think.