To The Moon

Stop calling it a bounce. A bounce is what happens when nothing changed. What happened Thursday is different. Four separate entities, Samsung, Microsoft, Amazon, and Apple, each independently confirmed within 24 hours that the AI memory shortage is real, worsening, and extending further than the market had priced. When the buy-side was selling memory stocks on CXMT IPO fears and AI capex peak anxiety, the sell-side of that trade just had its thesis demolished by the four largest technology companies on Earth.

Here is the full picture. Samsung reported its highest-ever quarterly revenue with operating profit of 89.5 trillion won, ahead of expectations. More importantly, it warned that memory supply constraints may persist into 2028 and signed multi-year supply agreements with major data centre operators. Microsoft's earnings and Azure commentary restored AI spending confidence across the sector. Then after the close, Amazon's CEO Andy Jassy said explicitly that higher memory costs are why Amazon is raising its 2026 capex target from $200 billion to $220 billion. Apple flagged rising memory costs for the September quarter. That is not a squeeze. That is four demand confirmations in a row.

The result: MU gained 18.36% to close at $874, then added another 5.07% in after-hours to $919. SanDisk surged 25.99% during the session and added 4.61% after hours. SKHY gained 17.52%. The DRAM ETF gained 16.70%. SOXS, the 3x inverse semis, collapsed 26.29%. Samsung and SK Hynix jumped 26.81% and 29.95% on Friday morning Korean time as KOSPI recovered nearly 18% from its recent lows.

The Four Confirmations That Changed Everything

Samsung's earnings are the most important piece of this. Not because Samsung reported a 13-fold jump in chip profits, though it did. But because Samsung, which is simultaneously a memory supplier and a consumer electronics manufacturer, explicitly told the market that its customers are not slowing down AI infrastructure spending. When a supplier with full visibility into demand confirms a shortage extending to 2028, that is not a bull-case model. That is a customer order book.

Microsoft's Azure beat matters to memory stocks specifically because Azure growth directly translates to hyperscaler data centre expansion, which directly translates to DRAM and NAND demand. Azure at 40% growth is not an abstract positive. It is a specific order volume signal for every memory supplier in the chain.

Amazon's capex revision from $200 billion to $220 billion, with CEO Andy Jassy explicitly attributing the increase to higher memory costs, is the most direct demand signal of the four. It is a hyperscaler publicly saying that it is spending more money than planned because memory is more expensive and scarcer than expected. That statement, made on an earnings call with full legal accountability, is worth more than any analyst note written this year.

Apple's commentary on rising memory costs for the September quarter adds the consumer electronics layer. AI features on iPhone require more on-device memory. Apple's supply chain is bought months in advance. If Apple is flagging higher memory costs for September, it means the tightness at the enterprise level is now visible in consumer electronics procurement too.

The CXMT Fear Was Overstated

Part of what made this week's initial selloff so sharp was China's CXMT completing its IPO and reports of continued progress in domestic Chinese chipmaking equipment. The bear narrative was that Chinese DRAM capacity coming online would end the shortage.

The Samsung earnings completely changed that framing. DRAM demand is growing 10% faster than global capacity this year according to Deutsche Bank projections, with a 2,261k WSPM demand estimate against 2,051k WSPM capacity. By 2028, that gap widens to 29% as demand reaches 3,563k WSPM against capacity of only 2,769k. Chinese CXMT capacity, even if it ramps faster than expected, does not close a gap of that structural magnitude in any plausible timeline. CXMT was a valid short-term sentiment threat. It is not a fundamental thesis threat.

Specific Predictions for Next Week

This is where I will be direct rather than hedge. Based on the convergence of four independent demand confirmations, the DRAM supply gap data, and the current price levels, here is the directional call for the week of August 4.

MU enters the week at approximately $919 to $970. The most likely range for next week is $950 to $1,050. The bull case for MU touching $1,000 intraday is real and the conditions are in place: 6x forward PE, $50 billion Q4 guidance, Amazon CEO citing higher memory costs as the specific reason for a $20 billion capex increase. The risk to this call is any fresh CXMT headline or BOK rate commentary that triggers Korea-led selling. The base case is consolidation in the $940 to $980 range as the market digests the rally before SanDisk's August 5 earnings become the next binary. A MU close above $1,000 before August 5 is possible but requires another positive catalyst, most likely either a Morgan Stanley or Goldman note raising targets.

SNDK enters the week at approximately $1,589 following Friday's gains. The most likely range for next week is $1,500 to $1,750. The distribution here is wide because August 5 earnings sit at the end of next week and options are pricing a 25% move around that print. The setup is now materially better than it was before Thursday. Samsung confirming the shortage extending to 2028 and Apple flagging higher memory costs directly validates the SanDisk demand thesis. If SNDK trades between $1,550 and $1,650 in the Monday to Wednesday window, that is the consolidation base before the binary. A move above $1,700 before August 5 would imply pre-earnings momentum buying, which historically increases the sell-the-news risk on the print itself.

SKHY enters the week around $154 to $158. The most likely range is $160 to $185. The Korea-specific risk has reduced but not disappeared. The BOK December hike expectation is still live. However, Samsung's 26.81% Friday surge in Seoul reduces the near-term pressure on SKHY significantly. A SKHY close above $170 next week would suggest the July correction has fully reversed in sentiment. The HBM4 thesis is intact and the shortage-to-2028 commentary from Samsung is the single most bullish statement SK Hynix's primary competitor could have made about the entire sector.

DRAM ETF at $54. Expected range: $56 to $64. This is the lowest-conviction single-name entry and the highest-conviction sector expression. If you believe the multi-confirmation thesis but are uncertain which single name captures the most upside, DRAM captures the Samsung, SKHY, MU, and WDC moves in one instrument without leveraged ETF decay risk.

The Risk That Could Break the Prediction

One risk remains live and is underappreciated in the current euphoria. The BOK is still expected to hike to 3.00% by December. The CXMT capacity risk did not disappear, it was temporarily overshadowed. If any headline in the week of August 4 suggests Chinese DRAM capacity is ramping faster than Samsung's outlook assumes, the sentiment will reverse quickly.

The second risk is specific to SNDK. The 25% options-implied move on August 5 means the market is pricing maximum uncertainty into the print. Samsung's shortage confirmation is a positive setup. But if SanDisk's own guidance is cautious, the Samsung confirmation is priced in and the stock trades on what SanDisk specifically says about its own contracted pricing and customer visibility into Q1 FY27. A stock up 25.99% in a single session needs its own earnings to confirm the move, not just the sector tailwind.

The third risk is macro. The Fed held rates at 3.50 to 3.75% on Wednesday but Warsh's language was noted as hawkish in tone. The 30% hike probability did not go away after the hold. If oil stays above $100 on continued Iran tensions and core PCE prints hot in the August data, the September hike probability rises and every high-multiple tech position, including memory stocks, faces the same multiple compression that hit them in July.

Why This Is Not a Repeat of July's Pattern

The July selloffs were driven by: BOK rate hike, KIS earnings miss warning, CXMT IPO fear, Iran oil spike, and Korea-led correlation selling. None of those were fundamental demand signals. All of them were external shocks to valuation sentiment.

This week's rally was driven by: Samsung all-time high revenue, shortage extending to 2028, Azure 40% confirming AI data centre demand, Amazon raising capex explicitly due to memory costs, Apple flagging higher memory costs in consumer supply chain. Every one of those is a demand signal, not a sentiment signal.

The distinction matters because sentiment-driven selloffs reverse when sentiment reverses. Demand-driven rallies reverse when demand reverses. There is zero evidence that AI data centre demand is reversing. There is significant new evidence from this week's earnings season that it is accelerating. The rebound is not a dead cat. The question for next week is whether it continues at this pace or consolidates before SanDisk's earnings binary on August 5.

The answer is almost certainly consolidation with an upward bias, not continuation of 18 to 26% single-session moves. Thursday's move was the snap-back from oversold to fair value. The next leg, from fair value to the bull target, requires SanDisk to deliver on August 5 what Samsung's commentary implies it should.

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