To The Moon

I haven't written an article in a long time, and choosing to pick up the pen today feels rather meaningful—Big Tech's valuations got collectively marked down on this day, because the nature of their business has changed, from asset-light to asset-heavy. Some may remember the HALO (heavy assets, low obsolescence) concept that Goldman Sachs championed in the first half of the year. As it turns out, HALO just means "hard to die"—it doesn't mean "won't fall."

When Big Tech collectively goes HALO, the broad market falls, and when the market falls, all stocks get dragged down with it. But on Thursday, semiconductor stocks—especially memory chips—held up remarkably well, because memory is the party being paid: it absorbs the cash flow coming from the HALO names. So when other stocks get marked down for lack of cash, the stocks holding plenty of cash should naturally be more resilient.

On top of that, the broad market faces a complicated macro backdrop. The US–Iran conflict has pushed oil prices up again; rising oil prices spark inflation; inflation in turn could push the Fed toward a rate-hike decision—not to mention this year brings the presidential midterm elections and a new Fed chair taking office. Weighing it all together, the outlook feels murky, as if missing a single week of the flood of information would leave you out of step.

To reverse a situation like this, all it takes is one widely adopted AI-to-B application. In truth, AI is already an irreversible industrial revolution—it's just that Wall Street is far too skilled at pressing prices down, seizing the moment when sentiment has scattered to aggressively scoop up cheap goods. That's basically been the playbook for the past 100-plus years.

$SPDR S&P 500 ETF Trust(SPY)$

The most basic and least obsolescence-prone bargain is the S&P 500. If you don't know what to buy on a dip, the S&P 500 should certainly be the first thing to consider. The key question is whether the S&P 500 is cheap enough right now.

Based on options positioning, a pullback to 710 over the next month can't be ruled out. Google's earnings were just the appetizer—even the best report card pulled back 7%, so how much will the worst student, Microsoft, pull back after its earnings? The positioning data points to an expectation of 710.

$NVIDIA(NVDA)$

Recent big options orders have gotten a bit hard to read—whether it's because the market is so divided that expectations diverge across institutions, or because memory has minted a batch of newly rich players placing big orders like money's no object. The two big call orders differ wildly.

One is a Sept 18–expiry 190 long call rolled to an Oct 16–expiry 180 call$NVDA 20261016 180.0 CALL$  , with 88,000 contracts opened. Pushing the expiry back and lowering the strike suggests a rather bearish view of the current market.

But this call position runs exactly opposite to another spread order: an Oct 16–expiry 220 call $NVDA 20261016 220.0 CALL$  and a same-day-expiry 250 call $NVDA 20261016 250.0 CALL$  form a bull call spread, expecting NVIDIA's share price to have a shot at returning to its mid-year high before October.

By the analysis I laid out at the start of this article, both big orders can be justified. That said, we can opt for the old method and sell puts on the dip$NVDA 20260729 195.0 PUT$  —although we don't know how the longer-term volatility will play out, it should be able to hold above 200 by month-end July.

$Roundhill Memory ETF(DRAM)$

Memory-semiconductor companies will stabilize and rebound thanks to Big Tech's massive capex.

A big order opened a bull call spread: buying the July 31–expiry 63 call$DRAM 20260731 63.0 CALL$ and selling the July 31–expiry 73 call$DRAM 20260731 73.0 CALL$.

Another very interesting thing: last Friday, July 17, saw a sharp crash. The day before, SMH had a big bearish order opened—buying 100,000 contracts of the July 31–expiry 500 put$SMH 20260731 500.0 PUT$ . But on Monday it didn't fall through. So what happened next? On Monday, July 20, a big bearish order opened—selling 100,000 contracts of the July 31–expiry 520 put$SMH 20260731 520.0 PUT$  —flipping the bearish stance straight into a not-bearish one. Truly jaw-dropping.

The bottom line is: in the face of massive capital expenditure, the semiconductor index is unlikely to fall hard. Memory semiconductors may even rebound.

$Micron Technology(MU)$

Micron had a big order that got shared very widely on social media today—a July 31–expiry 800 call, with more than 5,000 contracts opened and a turnover approaching $100 million.

After reading the analysis above, you'll know this must be a big buy-side call order. Of course, someone digging into the data might ask: why does the direction show as bearish? That's because the buy orders queuing up crowded out the bid price, so the execution price shows as filled at the bid—and as everyone knows, a fill at the bid is generally a sell. That's the logic. In reality it's a big buy order, as can be inferred from the clearly unreasonable bid-ask spread at the time of execution.

There's no need to analyze SK Hynix either—it's all the same.

$SpaceX(SPCX)$

It's downright unreasonable for SPCX to have fallen this hard, especially Wednesday's big bearish order$SPCX 20270319 115.0 PUT$  —the moment the selling was done, the price started to drop. This is absolutely insider stuff.

The unreasonable part is that Musk used to care intensely about his listed companies' share prices, yet now he's letting both SPCX and Tesla crash—it's simply unbelievable. You say some analyst claims the probability of the two companies merging has risen? Oh, a merger—well then, never mind. Indeed, the more it falls, the greater the merger probability.

I'm not sure whether SPCX's share price can hold at 100; if 100 doesn't hold, the only way is down to 80. For now, this downtrend shows no sign whatsoever of stopping.