To The Moon

$NVIDIA(NVDA)$  

$Micron Technology(MU)$  

$SK hynix(SKHY)$  

August 2026 was not a month. It was a stress test. In the span of four weeks, the market experienced its sharpest single-day crash of the year, a record high close, a Warsh hawkish shock, an NVDA blowout, a China memory threat, and a SanDisk Investor Day that rewrote the NAND story permanently. If your positions survived all five of those in sequence and you still have conviction, you have earned it. If you did not, August just taught you something worth learning before September.

Here is the honest review, and the September plan that follows from it.

August: What Happened and What It Proved

The month opened badly. Three triggers converged in 72 hours: a guidance miss on one forward quarter from the sector's best performer, a July payrolls print that shed 23,000 jobs against an expectation of a gain of 80,000, and a first-ever earnings report from the year's most-watched IPO that revealed quarterly capex so far above estimates it spooked the entire AI infrastructure trade simultaneously. Memory stocks fell 40 to 57% from their highs. The crash was fast, broad, and brutal.

What the crash proved: at a Shiller CAPE above 41, stocks priced for perfection have zero tolerance for any disappointment on any line in any quarter. The businesses that fell the hardest were not impaired. Their order books, their customer agreements, their margin structures were all intact. The stocks fell because the valuation left no room for a 5% guidance shortfall on a future quarter. That is the environment you are operating in and it does not change in September.

The recovery was equally fast and driven by real substance. An administration official said publicly that Apple sourcing memory from Chinese competitors was not something the administration favoured, fuelling a sector-wide rebound. Then came the Investor Day that changed the framing permanently. The NAND market was confirmed to be doubling in size by next year. Supply constraints were reaffirmed to extend into 2028. Contracted customer agreements covering tens of billions of dollars at floor pricing were disclosed for the first time, which meant the thesis shifted from analyst forecast to contractually obligated revenue. The market re-rated immediately.

Then NVDA delivered a $108 billion revenue forecast and said explicitly that memory supply is constraining its growth ambitions. When the world's most important AI chip company says memory is its bottleneck, every memory supplier's Q4 order book gets repriced upward. That is not a sentiment signal. That is a demand signal from the most important customer in the ecosystem.

Then the China threat reopened. Reports emerged that Apple may source DRAM and NAND from Chinese semiconductor companies. Memory stocks gave back recent gains in a single session. The same stocks that had just recovered 35% in a week fell 5 to 6% in a day on a sourcing report that has not been confirmed. That is how thin the sentiment floor is at current multiples.

Then Jackson Hole. A hawkish tone from the Fed chair strengthened September rate hike expectations. The S&P 500 trimmed gains and chip stocks pulled back further. Rate hike odds that had fallen to 30% mid-month crept back up into month end.

Five separate directions in four weeks. The positions that held through all five are the ones worth keeping into September.

What the Month Actually Delivered

The S&P 500 climbed more than 5% over August and sits roughly 19% above where it was a year ago. Tech led all sectors with gains above 7% month to date. Memory names were among the market's top performers for the year. The sector's best stocks delivered record quarterly revenue, record gross margins, record free cash flow, and expanded buyback authorisations in the same quarter that the market punished them for a forward guidance miss. That disconnect between business performance and stock price behaviour is the defining feature of trading at elevated valuations.

Three things changed permanently in August that carry directly into September.

First, the NAND supply story is now contractually locked rather than analyst-projected. Tens of billions in customer commitments at floor pricing were disclosed publicly for the first time. Even if spot NAND prices soften, the blended revenue base does not collapse the way it would have in prior cycles. This is a structural shift in how the memory business behaves, and it only became visible in August.

Second, the China memory risk is now a live September variable rather than a background concern. The Apple sourcing reports, while unconfirmed, demonstrated that the market will sell memory stocks immediately and aggressively on any suggestion that Chinese supply is entering the demand chain. That sensitivity travels directly into September without resolution.

Third, NVDA's explicit statement that memory is constraining its growth is the strongest possible demand-side confirmation the sector has ever received. When the world's largest AI chip company increases its purchase commitments and publicly attributes growth constraints to memory supply, the thesis stops being a macro argument and becomes a customer order.

The September Playbook

September is historically the weakest month of the year for equities, averaging a decline of 1 to 2% over the past decade. Entering it with a CAPE above 41 and a Fed chair who just delivered a hawkish message at the most watched central bank forum of the year, that historical pattern has more teeth than usual.

Three forces will define September specifically.

The first is the Fed. The September FOMC meeting is the single most important macro event of the month. A hold with neutral language is the base case that keeps the AI trade intact and allows memory stocks to consolidate their August recovery. A hold with explicitly hawkish language compresses multiples immediately regardless of earnings momentum. An actual hike puts growth stocks under sustained pressure through Q4 in a way that no earnings beat can fully offset.

The second is the China memory binary. The Apple-CXMT-YMTC sourcing story is not resolved. The administration's stated preference against Chinese sourcing is meaningful but it is a policy position, not a confirmed restriction. If Apple confirms in its September product cycle that it is sourcing memory components from Chinese suppliers, it removes a significant demand pillar from the US memory supply chain at the same time as it adds Chinese supply. That scenario has not been priced by any analyst model because it is genuinely binary and genuinely uncertain.

The third is the NVDA read-through. The $108 billion revenue forecast confirmed that memory supply is constraining the growth of the most important company in the AI ecosystem. September earnings guidance updates from memory names will be read against that backdrop. Any confirmation that Q4 order books are tracking toward the high end of guidance, in the context of NVDA explicitly needing more supply, changes the September narrative entirely.

The September Pick Levels

Three names, three specific entry frameworks.

The first name at approximately $1,484 has retraced meaningfully from its post-Investor Day peak. The 12-month analyst average target implies more than 40% upside from current levels. The contracted revenue base at floor pricing provides a floor that did not exist in prior cycles. The entry zone that makes sense fundamentally sits between $1,400 and $1,500. Above $1,700 before any new catalyst is where the risk-reward deteriorates. The next major catalyst is Q1 FY27 revenue tracking toward the high end of guidance in the October earnings print.

The second name at approximately $940 remains the cleanest risk-reward in the complex. Forward earnings multiple in the single digits. Quarterly revenue guidance of $50 billion intact. Explicit confirmation from its most important customer that memory supply is constraining demand. The 60% gap to the analyst consensus target has not closed because the market is still applying a discount for sector contagion and China risk, not for any deterioration in the business itself. Entry in the $900 to $950 zone on any pullback driven by macro rather than fundamental news.

The third name at approximately $138 has the weakest near-term setup of the three due to Korean rate sensitivity and cross-market correlation. The structural advantages, global HBM market dominance and next-generation technology delivery ahead of schedule, are unaffected. But September rate hike risk in Korea and the correlation-driven selling that follows every broad risk-off move make this a dip-buyer's name rather than a momentum name. Entry only below $130.

The Honest September Verdict

August taught you which of your positions are thesis trades and which were momentum rides. The thesis trades held their logic through every August test even when the price moved sharply against the position. The momentum rides needed August to be simpler than it was.

September will be volatile. Three separate binary events in a historically weak month, with a hawkish Fed chair, an unresolved China memory story, and CAPE above 41. The discipline is not abandoning conviction on thesis names when the volatility arrives. It is knowing exactly which level the thesis breaks at and respecting that level when you reach it.

The positions that survived August with conviction intact deserve to go into September. The ones that needed August to be simpler deserve a rethink before the historically difficult month begins.

I am not a financial advisor. Trade wisely, Comrades.