Micron's New Supercycle: Riding the High-Bandwidth Memory Wave

When I wrote the original narrative, the thesis was straightforward: Micron was riding a powerful AI driven memory supercycle, and HBM was the ticket.

That thesis hasn’t changed. What has changed is the speed and scale of what’s unfolding. The market has caught up. And it caught up fast. It has blown past even the optimistic case.

The stock is up 160% since I published it five months ago.

The Numbers Have Gone Parabolic

Micron just posted fiscal Q1 2026 revenue of $13.64 billion: up 57% YoY, and guided Q2 at a staggering $18.7 billion, which would represent 132% year over year growth. But the real headline is gross margins. When I wrote the original piece, Micron was sitting at 41% for fiscal 2025. The Q2 2026 guide? 68%. That's not a memory company margin, that's software territory. The pricing power here is extraordinary, and it tells you just how tight supply has become relative to demand.

Free cash flow hit a record $3.9 billion in the quarter, beating the company's previous all-time high by over 20%. This matters because it means Micron isn't just growing revenue, it's converting that growth into real, deployable cash at an unprecedented rate.

HBM: Sold Out and Pulling Forward

The entire HBM production line is 100% sold out through calendar year 2026. Every single wafer is spoken for. More importantly, Micron started shipping its next generation HBM4 modules a full quarter ahead of schedule, hitting speeds above 11 Gbps and bandwidths over 2.8 TB/s. That's a meaningful technical win, it reinforces that Micron isn't just riding HBM demand, it's competing credibly at the leading edge against SK Hynix.

Perhaps the most significant revision: management now expects the HBM market to hit $100 billion by 2028, pulled forward by two full years from prior projections. That single data point tells you how dramatically the AI infrastructure buildout has accelerated.

The "Crowding Out" Effect Is the Hidden Story

Here's something that doesn't get enough attention. Manufacturing HBM requires roughly three times the wafer capacity of standard DRAM. So as Micron, Samsung, and SK Hynix all sprint to build more HBM, they're starving the conventional PC and smartphone memory markets of supply. The result?

Standard DRAM contract prices are forecast to nearly double in Q1 2026. Not because PCs are suddenly flying off shelves, but because the fabs that used to make that memory are now dedicated to AI chips.

This is a structural dynamic, not a temporary blip. As long as HBM remains a huge share of the production mix, it acts as a natural cap on total industry bit supply. That's a fundamentally different setup than previous memory cycles, where oversupply was always lurking around the corner...

"This Time Is Different!" ...To A Degree

The memory industry has a brutal history of boom-and-bust as I mentioned in my original narrative. Suppliers overbuild during good times, demand cools, prices crash, margins go negative. It's happened repeatedly. So the obvious question: could it happen again?

Honestly? Yes, it absolutely could. That risk hasn't disappeared. But several things are genuinely different this cycle:

Contract structures have teeth now. Micron has negotiated multi-year agreements with non-cancellable volume commitments and stronger pricing floors. In previous cycles, customers could walk away from orders when the market turned. These new contracts provide a much thicker buffer.

HBM isn't a commodity. Unlike standard DRAM, HBM4 requires deep integration with specific processors, custom logic dies manufactured at TSMC, and advanced packaging. You can't just swap suppliers. This stickiness transforms memory from a spot-market commodity into something closer to a bespoke enterprise product.

The triopoly is behaving rationally. All three major players are prioritising margins over market share, allocating capacity to premium HBM rather than flooding the standard DRAM market. That discipline is key, and it needs to hold.

The Bear Case You Need to Watch

The biggest vulnerability is hyperscaler spending. U.S. cloud giants are guiding for roughly $700 billion in combined 2026 capex, with ~75% going to AI infrastructure. Micron's revenue is heavily concentrated in this segment. If those companies decide to pause and digest what they've already built, because AI software revenue isn't justifying the spend, the transmission to Micron would be fast and painful.

Here's the mechanism: HBM orders stall, fabs pivot back to making standard DRAM to stay utilised, that floods the consumer market, and suddenly prices collapse across the entire portfolio. A classic memory bust.

In that scenario, a 50–70% drawdown is entirely plausible.

The signposts to watch: quarterly capex guidance from Microsoft, Meta, Google, Amazon, and Oracle. If you start hearing words like "optimisation" and "digestion" replacing "acceleration" and "expansion," that's your warning...

Updated Valuation View

My original fair value estimate was ~$200 per share, based on 14% annual revenue growth and 20% net margins within five years. Given the acceleration in HBM revenues, the pull-forward of the $100B TAM target, and the margin expansion to 68%, that estimate needs revising upward. The structural pricing power and contract visibility today are materially stronger than what I modelled six months ago.

That said, I want to be honest: at these margin levels, the market is pricing in a lot of perfection. The higher the stock climbs on supercycle euphoria, the more violent the correction will be if the cycle turns. This remains a high-conviction, high-beta position, brilliant if AI infrastructure spending continues, but requiring active monitoring of the downstream signals that would flag a turn.

I wouldn't be adding to my position at these prices. In fact, I've trimmed and locked in some gains. Micron's history is littered with corrections that arrive fast and hit hard. At a minimum, the stock probably needs a period of consolidation at these levels before it can sustainably move higher. The risk of a meaningful pullback is real, but I don't see it as imminent. For now, I'm holding (a smaller position) but with one eye firmly on those capex signals.

Bottom line: The thesis is stronger than when I first wrote it. The execution has been exceptional, the market dynamics are even more favourable, and the structural shifts toward HBM customisation and contract rigidity provide genuine reasons to believe this cycle has more durability than past ones. But memory is memory, respect the cycle, watch the capex signals, and size your position accordingly.

Good luck!