A $45 Billion Fund Blew Up.
The Market Closed at a Record.

The scoreboard

The S&P 500 rose 3.6% on the week, the Nasdaq Composite 5.2%, and the Dow Jones Industrial Average 3.0% — the biggest weekly gains for the major indexes since mid-April. The S&P closed Friday at a record 7,757.64, the Nasdaq at a record 26,690.62, and the Dow at 54,036.93.^[1]

A note on how those gains get described, because it matters: a 3% week in the Dow is roughly 1,500 points, not 3,000. Percentages and points are not the same thing, and at 54,000 the point number sounds far more dramatic than the move actually is. If you take one habit from this post, make it converting points to percentages before you react to them.

The blowup nobody expected to be survivable

The defining event of the last two weeks was not an earnings report. It was a liquidation.

Situational Awareness, the AI-focused hedge fund run by former OpenAI researcher Leopold Aschenbrenner, told investors its portfolio fell 67% in July. The fund peaked around $45 billion at the start of the month and was down to roughly $10 billion by month's end. Reported leverage of up to 400% brought margin calls from its prime brokers, which forced a distressed sale of its public equity positions — including SK Hynix and CoreWeave — to Citadel at below-market prices.^[2][3]

The structure of the trade explains a lot about how the tape behaved. The fund was long the AI supply chain (chips, memory, data centers, power) and short software companies it believed AI would disrupt. Unwinding that book does two things at once: it forces selling in semis, memory and storage, and it forces buying to close the shorts in software. That is a mechanical flow, not a verdict on any of those businesses.

Earnings did the heavy lifting

Underneath the flow story, the reported numbers were strong.

Company	What was reported
Caterpillar Reported Aug 4	First quarter in company history above $20 billion in sales and revenues, at $20.5B versus roughly $19.2B expected. Adjusted EPS $8.17 against a $6.19 estimate. Backlog hit a record $72B, up about 92% from a year earlier, and full-year sales growth guidance went to mid- to high-teens, with data center demand cited as a driver. Shares rose about 5.7%.[4][5]
Palantir Reported Aug 3	Revenue up 93% year over year to $1.94B, adjusted EPS $0.41 versus $0.35 expected, US commercial revenue up 149% to $764M, and full-year guidance raised to about $8.15B. Shares closed up roughly 29.5% at $162.66 on August 4.[6][7]
Toast Reported Aug 4	Revenue $1.908B, up 23.1%, essentially in line with estimates. Non-GAAP EPS $0.34 against a $0.21 consensus, helped by operating leverage and a roughly $10M one-time tariff benefit. Locations up 22% to about 180,000; gross payment volume up 22% to $60.7B. Through June 30 the company had repurchased 19 million Class A shares for $486M. Shares slipped after hours despite the beat.[8][9]
Wayfair Reported Aug 4	Revenue $3.52B, up 7.5%, EPS $0.95, both above estimates. US revenue up 8.7%, the best domestic growth since 2020. Adjusted EBITDA margin 6.9%, best since 2021; free cash flow $301M. Shares jumped roughly 15-19%, with about 18% of the float sold short going in — so part of that move was short covering, not fresh conviction.[10][11]
Berkshire Hathaway Reported Aug 8	Net income $25.67B, with operating earnings up 16% to $12.98B. Manufacturing, service and retailing earnings up 24%. Cash and Treasury bills fell to $344.1B as Berkshire turned net buyer of equities by nearly $20B and repurchased about $4.5B of its own stock.[12][13]

Opinion: the Toast reaction is the interesting one. A 23% revenue grower that beats on profit and still trades down tells you expectations, not results, were the binding constraint. That distinction — results versus expectations — is most of what separates people who are confused by earnings season from people who aren't. It's the core of Fundamental Analysis, and you can pull the nine metrics that matter for any of these names on the free fundamental pillar.

And the Berkshire line worth sitting with: after years of hoarding cash, the company put roughly $20 billion net into equities in a single quarter. That is not a market call you should trade off. It is, at minimum, a data point about what a very patient buyer thought prices looked like in the second quarter.

The number that changed the story

Here is where a lot of weekend commentary — including my own first draft of this — got it backwards.

Friday's July employment report showed the US economy lost 23,000 nonfarm jobs, against expectations for a gain of roughly 80,000, with the unemployment rate at 4.1%.^[1][14] The market rallied on it, and the reported reason was not "growth is strong." It was that a weak labor print takes a near-term Fed hike off the table.^[1]

That's an important correction to make out loud. It is tempting to build a tidy story where earnings are accelerating and the economy is hot and the Fed is about to tighten into strength. The employment data for July does not support the middle claim. You can hold "corporate profits are growing" and "the labor market softened last month" at the same time — they are different measurements of different things, and they diverge routinely. Economics for Traders spends a whole stage on why.

Rates, gold, oil, dollar, bitcoin

• Treasuries. At around 3:30pm ET Friday: 2-year 4.19%, 10-year 4.65%, 30-year 5.19%.^[15] The curve is already positively sloped, with about a point between the 2-year and the 30-year.
• Gold. Comex December gold rose 2.3% Friday to $4,401, a roughly seven-week high. Worth being precise: that is a breakout from a range, not a record — gold remains about 21% below its late-January peak near $5,600.^[16][17]
• Silver. September silver jumped 3.6% to $63.85, also a seven-week high, and still roughly half its $121.67 peak.^[16] Silver following gold is a real pattern; it is also the more violent of the two in both directions.
• Oil. WTI was around $77-78 a barrel, not $75.^[18] Close enough to sound the same in conversation, far enough apart to matter if you're building a thesis on it.
• Dollar. The dollar index slipped to about 99.60, back below 100, after its worst week in three months.^[19] A softer dollar is part of why metals moved.
• Bitcoin. Around $65,000, up roughly 2.8% on the week — and this is the part I find notable. Jim Cramer announced he was selling his bitcoin over long-term quantum computing concerns, and Strategy sold 1,638 BTC (about $104.7M) in the week ended August 2 at an average of $63,957. Two visible sellers, plenty of bearish headlines, and the price held.^[20][21][22] Opinion: an asset that stops going down on bad news is telling you something about who owns it. That is an observation, not a forecast, and bitcoin remains extremely volatile.

Three claims I could not support — and one I'll defend

Part of doing this honestly is publishing the things that didn't survive a fact check.

1. "The market broadened out." By index, it didn't — not this week. Nasdaq +5.2% beat S&P +3.6% beat Dow +3.0%.^[1] That's still large-cap tech leading. Gold, retail and industrials all worked, which feels like broadening, but the index math says leadership stayed concentrated. If you want to check breadth properly rather than by feel, that's what the 20/50/200-day readings on the technical pillar are for.
2. "Consumer staples is the worst sector in the market." I couldn't verify that. Staples posted some of their strongest relative performance in years in early 2026 as money rotated defensive, and XLP has lagged the S&P over one year (roughly 11% versus 14%) without being the worst group.^[23] Lagging and worst are not the same claim, and the second one needs a sector table, which is on the economic pillar.
3. "SpaceX reported this week." SpaceX is private and does not report earnings. What actually happened is still striking: Starlink subscribers reportedly doubled to about 12 million, the FCC approved SpaceX's roughly $17 billion purchase of about 65 MHz of spectrum from EchoStar, and SpaceX filed plans for an orbital data-center constellation of up to one million satellites.^[24][25][26] None of that is investable in public equity directly, which is exactly why the distinction matters.

The claim I will defend, as opinion: the AI capital-spending fear got smaller this quarter. Combined 2026 capex across Amazon, Alphabet, Meta and Microsoft is running near $725 billion, up about 77% from 2025 — and for the first time all four laid out the same de-risking playbook, committing early only to long-lived assets (land, shells, power) and deferring the chips, the majority of the cost, until demand is visible. Alphabet's cloud revenue grew 63% to $20B with backlog near $462B.^[27][28] Whether that spending earns its cost of capital is still an open question. But "they are burning money into a void" is a weaker argument this month than it was last quarter.

The bear case, stated fairly

The strongest argument against everything above is leverage. It is not a hypothetical: a $45 billion fund just got liquidated by its brokers, and the market absorbed it. That can be read two ways — the system is resilient, or the system hasn't been tested yet at size. Both readings are honest.

Opinion on what would actually change my mind: the shape of the yield curve, and specifically why it's moving. Long rates rising because growth is strong is one thing. Short rates falling hard while long rates climb — the Fed cutting into a slowdown while the long end demands more compensation for inflation and deficits — is a different animal, and historically a less pleasant one. Right now the 2s/30s spread is about 100 basis points.^[15] A move toward, say, a 3.80% 2-year against a 6% 30-year would be worth taking seriously. That is a scenario to watch, not a prediction, and I have no idea if it happens.

One more piece of hygiene: "a stock is far below its all-time high" is not a valuation. It's a distance from a past price. Valuation compares price to earnings, cash flow or assets. Plenty of things are 60% off a high and still expensive. The habit of treating drawdown as discount is one of the most reliable ways retail investors lose money, and it's the first thing Fundamental Analysis takes apart.

The week ahead

The main event is the July CPI report midweek. Consensus looks for headline inflation around 3.4% year over year and core around 2.5%. A hot core print — something near 2.9% — would put rate-hike odds back on the table ahead of the September Fed meeting, which is the opposite of what Friday's jobs number did. Earnings slow down considerably, with Super Micro Computer, CoreWeave and JD.com among the remaining names.^[29]

Two things worth watching that aren't on any calendar: whether silver confirms gold's move, and whether the semiconductor complex stabilizes now that the forced selling from the fund liquidation is reportedly done. If chips can't rally without a bid, the flow story was never the whole story.