The $2 Trillion Week

For most of this year the Magnificent Seven was dead money. Bloomberg's Mag 7 index lost 1.9% in the first half while the S&P 500 gained 9.3%. The group trailed roughly 300 stocks in the index. As recently as July 21, Microsoft was down -18.6% on the year and Tesla was down -15.3%. The one name working was Apple, up about 23%, largely because it wasn't spending on AI.

Then, in the last week of July, the hyperscalers reported.

What earnings said

Microsoft put up Azure's first $100 billion year. The quarter grew 43%, management guided to acceleration, and capex guidance didn't move. The stock jumped +15.5% on Thursday, the biggest single-day gain in market value ever recorded, and added over $600 billion for the week. By Monday's close it had clawed back basically all of its 2026 decline.

Amazon raised capex from $200 billion to $220 billion and ran free cash flow to -$7.6B, and got rewarded anyway because AWS growth was that strong. The stock crossed $3 trillion on Monday. It came into the week up about 4% on the year.

Alphabet had the strangest week of the three. Google Cloud grew 82%, its best quarter ever, with a big backlog build. But capex guidance went up again to $205 billion, quarterly spending ran above operating cash flow, and free cash flow went negative. The stock sold off on the print, then added more than $400 billion by Friday.

The real question is return on capital

Alphabet is the cleanest way to see what the market is actually arguing about.

Return on invested capital measures what a business earns on the money put into it. Alphabet has been one of the great ROIC stories in the market, and not just lately. Its 20-year average runs around 30%, depending on how you measure, against a cost of capital around 10. Basically since the IPO, every dollar reinvested in search and ads has earned roughly three times what it cost. That's why the stock compounded the way it did.

This capex cycle changes the math. Alphabet's invested capital grew from roughly $228 billion in 2024 to about $311 billion in 2025, and this year's $205 billion budget keeps piling into the denominator. The numerator on those new dollars doesn't exist yet. A data center earns nothing while it's being built, and no one can say whether it will eventually earn 25% like search or 8% like a utility.

That's the doubt in the stock. Not whether people use the products. Whether the new dollars earn what the old dollars did. If the incremental capital comes anywhere near Alphabet's historical rate, today's multiple is too low. If it earns cost of capital, the company is growing revenue while destroying value. Record cloud growth and a selloff can show up in the same print when the market can't tell which of those it's buying.

Microsoft got a different verdict because it offered an answer. Azure crossed $100 billion and accelerated on a capex budget that didn't grow. More return, same denominator. That is an ROIC statement, and the market paid for it the same day.

The winner sold off

Apple, the best performer in the group all year, fell -10.8% over the same five sessions. Tim Cook warned that memory prices and chip capacity constraints from the data center buildout will weigh on guidance. The company that got paid a premium for skipping the capex race is paying for it anyway, through its supply chain.

Meta dropped 9% on a messy print (legal charges, severance, a higher capex floor) and got it all back by Monday, including +6.0% on Monday alone. Tesla's free cash flow went negative in Q2 and the stock is still down 27% over six months. Nvidia reports August 26, now with more riding on it.

What we take from it

The group doesn't trade as a bloc anymore. About $2 trillion moved in and out of six megacaps in one week, and the spread between the best name and the worst was 35 points. The verdicts weren't even consistent. Alphabet got punished for negative free cash flow on Wednesday and forgiven by Friday. Amazon posted the same thing plus a capex hike and got rewarded immediately.

These seven stocks are about a third of the S&P 500. Own the index and you own all of this at full size, whether you meant to or not. We'd rather underwrite each name on its own capital returns than hold a basket because it sits at the top of the index.