For most of the summer, the story on Micron was simple.

AI needs memory. Micron makes memory. Back up the truck.

Then the stock fell 28.7% in July, and that simple story suddenly needed an asterisk.

It clawed back above $1,000 on Friday, up 6.1% in a single session, riding the same wave that sent SanDisk up almost 12% and SK Hynix up 8%.

So the easy read this week is that the supercycle is back on and July was just noise.

I went digging into what Nvidia itself disclosed about its own memory spending.

Buried in there is a specific number, tied to a specific year, that tells you almost exactly how this plays out.

Not whether. When.

Missed Friday's Playbook on Broadcom?

The tension here is obvious: Trump wants rate cuts, but Fed Chair Kevin Warsh is still sounding hawkish ahead of the September 15 to 16 decision. The latest inflation data could be decisive, with economists expecting August CPI at 3.4% year over year and core CPI at 2.4%. In plain English, politics is pushing for easier money, but the Fed may still be more worried about inflation credibility than market comfort.

Snowflake’s business momentum may be real, but investors should be careful after the recent run. The stock is up 67% in 2026 and recently surged about 17% after strong Q2 results, helped by renewed AI software excitement. The catch is valuation: when a stock runs this hard, even a good company can become a risky buy if expectations start outrunning fundamentals.

Lululemon beat EPS expectations, but the market focused on the ugly parts: weak Americas sales, softer guidance, and a turnaround that now looks harder than hoped. Reuters reported Americas revenue fell 8%, while management cut fiscal 2026 revenue guidance to a 5% to 7% decline and lowered EPS guidance to $9.48 to $9.73. In plain English, this was not a normal “beat and raise” quarter. It was a reminder that the brand still has a demand problem in its core market.

Private real estate has long been a cornerstone of institutional portfolios, but individual investors have often had limited access to the asset class - or have had to invest through pooled funds and intermediaries.

Lightstone DIRECT offers a different approach. Accredited investors can invest directly in select real estate opportunities alongside Lightstone, a vertically integrated owner/operator with more than four decades of experience and a $12 billion real estate portfolio.

We invest 20% or more of the equity in every offering, aligning our interests with yours from day one. Our in-house team oversees acquisitions, asset management, and investment execution, with property operations managed by Lightstone and, where appropriate, experienced third-party operators.

No blind funds. No intermediaries. Just direct partnership.

Accredited Investors only. $100k minimum investment.

Explore Opportunities

^{This communication is for informational purposes only and does not constitute an offer to sell or a solicitation of an offer to buy any securities. Any such offer or solicitation will be made exclusively through the definitive offering documents. All investments involve risk of loss, including the potential loss of principal. Past performance is not a guarantee of future results. Any targeted returns or projections are forward-looking statements, are based on current assumptions, and are not guarantees of future performance. Actual results may differ materially.}

Micron just posted the loudest quarter in its history.

Revenue hit $41.5 billion in the quarter that ended in May, up 346% year over year.

Guidance for the current quarter calls for $50 billion, with gross margin near 86%.

The obvious read is that memory is the best trade in tech right now, and Micron is the proof.

Here's what that read leaves out. CEO Sanjay Mehrotra told analysts that Micron's entire 2026 HBM allocation is already sold out under fixed-price contracts.

Customers are asking for roughly 50% more supply than Micron can actually deliver.

That sounds like unlimited pricing power.

Then I found the number Nvidia disclosed about its own memory buying, and it made me ask a completely different question.

Sold out until when?

Start with what already happened, because it's real.

Non-GAAP gross margin hit 84.9% last quarter, a company record.

Non-GAAP earnings per share came in at $25.11.

That is not a chipmaker squeezing out incremental improvement. That is a company operating in a market where it can basically name its price.

Wall Street's bulls agree. BMO Capital slapped an Outperform rating and a $1,300 price target on the stock on August 21, calling Micron a primary beneficiary of the memory supercycle.

New Street Research put a $1,250 target on it around the same time.

The backlog is the real anchor here. Fixed-price contracts are locked in for all of 2026, and demand is still running about 50% ahead of what Micron can ship.

That is about as close to a guaranteed order book as a cyclical semiconductor company ever gets.

Here's where the sold-out story gets more complicated.

Nvidia disclosed roughly $267 billion in memory purchase commitments running through its fiscal 2029, which ends in January 2030.

About $92 billion covers the rest of fiscal 2027.

$87 billion is earmarked for fiscal 2028.

$88 billion for fiscal 2029.

Then it falls off a cliff. Fiscal 2030 and beyond, Nvidia's committed spending drops to $6 billion, then $5 billion, then $1 billion.

Not a gradual slowdown. A wall.

That money isn't all Micron's, either. It's split across Micron, SK Hynix, and Samsung, and not every dollar Nvidia commits turns into Micron revenue.

One read on this: once Nvidia's current commitments run out, it may not need to lock in HBM at today's premium prices anymore.

Negotiating leverage flips back to the buyer. One analyst's prediction is that Micron stock could peak somewhere between December 2028 and January 2029.

There's a second complication sitting quietly behind the first one.

China's CXMT is expanding DRAM capacity fast enough that analysts are now flagging real risk of global oversupply and price declines by 2028, right as Nvidia's spending schedule starts to taper.

Because a supercycle with a published expiration date and a well-funded challenger waiting in the wings isn't really a supercycle.

It's a very good multi-year contract.

The chart tells its own version of this story, and it's more dramatic than the headlines suggest.

Micron hit an all-time high of $1,255 at the end of June and looked unstoppable.

Then it lost more than 30% of its value in about six weeks, sliding under $850 as fears about AI-hardware valuations and a soft SK Hynix earnings report spread through the whole sector.

For a while it traded like a stock that had lost its story completely, grinding lower in a descending channel with heavier volume on every leg down.

Friday changed the tone. A 6.1% jump put it back above $1,000, a level that had capped every bounce on the way down and is now trying to flip into support on the way back up.

The next real test sits close to that old $1,255 high.

Reclaim that, and the supercycle story gets its confirmation. Fail there again, and this starts looking like a stock that peaked in June and has been trading its own aftershocks ever since.

This is the next real checkpoint. Micron reports fiscal Q4 results on September 30, against guidance that already assumes $50 billion in revenue.

I covered Micron back when it first crossed $1 trillion in market cap, in Pragmatic Friday: Micron hit $1 trillion. Now the hard part starts.

The setup has changed since then. Back in May the question was whether the rally could hold. Now the question is whether Micron can beat a number that already assumes close to perfect execution.

China's CXMT is reportedly targeting a roughly $4.2 billion listing to fund DRAM expansion.

Bull case: this is years away from touching Micron's high-margin HBM business, which is where the real profit lives. Bear case: cheap Chinese DRAM has undercut pricing before, and a well-capitalized competitor moves faster than most models assume.

BMO and New Street Research sit at $1,300 and $1,250. Citigroup just cut its target from $1,400 to $1,150 while warning of DRAM and NAND price softening by 2027, and still calls it a Buy.

When bulls and skeptics agree on direction but disagree by hundreds of dollars on the number, that gap is the market arguing with itself in public.

Nvidia's memory commitments fall from $88 billion in fiscal 2029 to single digits the year after.

Wildcard: shipments typically lag commitments, so Micron could keep filling backlog well into 2029 even after new orders slow down. Bear case: markets usually start pricing in a cliff long before it actually arrives.

Here's where I've landed after sitting with all of this.

Micron is not overvalued relative to what it's earning today. A company printing 84.9% gross margins and guiding to $50 billion in quarterly revenue isn't priced like a bubble. It's priced like a business that finally has real pricing power for the first time in its history.

But "sold out" and "permanent" are not the same word, and this week's rally is treating them like synonyms.

Nvidia didn't hide the cliff. It published it, with dates attached, in its own disclosures.

Everyone chasing Friday's 6.1% pop saw the sold-out headline. Almost nobody read far enough into the filing to find the year the buying slows down.

I'd rather own Micron into the September 30 print than short it. But I'm watching $1,255 far more closely than I'm watching this week's bounce, because that old high is the only level that will tell us whether the market has actually done this math.

Get revenue agents, workflows, and automations across every stage of your motion. Access customer data in real time through Attio's web app, MCP, API, and SDK.

Then Ask Attio anything about your business and get instant answers.

It's the CRM that runs the work behind every win.

Try Attio Now

The best businesses in the world can still be bad stocks to buy at the wrong moment in the cycle.

Micron's numbers this year are not in question. The revenue is real, the margins are real, and the backlog is real.

What's less certain is whether the customer footing most of that backlog has already told everyone, in writing, when it plans to stop paying up.

A sold-out order book feels like safety. Sometimes it's just a countdown with good PR.

The stocks that hurt investors most aren't usually the bad businesses. They're the great businesses bought at the exact moment their best-known buyer starts planning its exit.

If One Big Idea sharpened your week, send it to one friend who needed it too.