Sales rose 26 percent. The quarter’s building bill was larger than the cash the business produced. The Nasdaq fell. Gold fell with it. The test was cash, not the story.
This is not a recommendation to buy or sell Meta, gold, silver, bitcoin, oil, or any stock. It is not a forecast. Figures are from Meta’s October 21, 2026 release, its quarterly filing, and market screens on October 22. Screens differ by a little. Read the filing.
The one idea is simple. A bigger top line is not cash left after the build. Meta proved it in one quarter. The market then sorted the day by who still had cash to spare, and who had to keep writing checks.
What the company said
Meta reported third-quarter results on October 21. Revenue was $64.601 billion. That was up 26 percent from $51.242 billion a year earlier, and ahead of the $63.37 billion analysts had expected on CNBC’s tally. The ad business is not shrinking.
The spare cash is. Capital spending in the quarter, including principal payments on finance leases, was $34.91 billion. Most of it was for servers, data centers, and networks. Cash from operations was $32.79 billion. Free cash flow was negative $2.12 billion. A year earlier, the same quarter had thrown off $10.63 billion of free cash flow. Reuters and CNBC reported this as the first negative free-cash-flow quarter since the firm went public in 2012. That history is their report, not a line in the release.
Free cash flow is what is left after the business pays its costs and its building bills. It is the number a miner already knows. You can ship more ore and still have to raise money if the mill, the power line, and the camp cost more than the metal brings in that quarter. Meta shipped more ads. The data centers cost more than the quarter’s operating cash.
That is not an empty treasury. At September 30, cash, cash equivalents, and marketable securities were $82.45 billion. One bad quarter of free cash flow did not drain the pile. It did show that the build, this quarter, was larger than the cash the ads produced.
The firm also raised the floor on this year’s building budget. Full-year capital spending is now put at $135 billion to $145 billion, narrowed from $130 billion to $145 billion. The ceiling did not move. The company did not give a 2027 capital-spending number. It did say it is still drawing next year’s budget, that it expects to keep investing hard, and that it expects strong revenue growth through 2027. It also said Reality Labs losses should narrow to about $15 billion next year, from an expected $18 billion to $19 billion in 2026.
A separate tally is easy to misread. It is not in the earnings release. Bloomberg added three items in the quarterly filing and got about $825 billion of future leases and contracts. The parts are $454.5 billion of contracts that cannot be canceled, $355.7 billion of leases not yet on the balance sheet, and up to $14.7 billion of cloud capacity. In July, Bloomberg’s sum was almost $700 billion. This is not cash that left the building this week. It is a stack of promises, many of them for computing, data centers, and power, with some leases running out toward 2036. A promise that large changes the question from “is the product good?” to “who pays?”
There was a profit miss too, and it is the smaller fact. A charge of $9.89 billion, tied to a multistate attorneys general agreement, pulled reported earnings to $3.83 a share. The outlook rounds that charge to $9.9 billion. Take the charge and its tax effect out, and diluted earnings were $7.40 a share, which beat the Street. The quarter was not a collapse of the business. It was a choice to spend past the cash the business made in the quarter.
Muse is already out. Mark Zuckerberg told analysts that more than 10 million people in the United States and Canada have set up accounts, and that about a quarter of them use it on a given day. He also said the agent uses a lot of computing power. He expects the cost to run it to fall by more than ten times over the next year, which he said would let it turn a profit once the revenue is large enough. That is a builder’s sentence. It is also a bill.
Why the stock and the insurance moved together
The first read of the report looked like a win. The shares had closed Wednesday at $697.52. CNBC said they rose about 2.5 percent in extended trade on the sales beat and a fourth-quarter revenue range of $73 billion to $76 billion. Another report put the extended price at $722.42, up 3.57 percent. A claim that they jumped 5 percent to $748.80 does not fit that close. Five percent on $697.52 is about $732. That higher print came from a paywalled wrap and is not used here.
Then readers got to the cash, the higher floor on spending, the long list of future commitments, and a plan for a new Class C share.
Class C stock would not vote. The board has approved it, with an effective date in January 2027. The company said the point is to keep a founder-led firm and to raise money for growth without handing out votes. The California line is narrower than it first sounds. CNBC, citing the filing, said Zuckerberg could face a multi-billion-dollar tax if a state wealth tax passes, and could leave. The filing described the new stock plan as what convinced him to stay. The market read a plainer line. New stock is a way to pay for a build the quarter’s cash no longer covers. New stock dilutes the people who already own the old stock.
Thursday’s close was $648.85. That is down $48.67, or 6.98 percent, from $697.52. A Reuters screen during the day had the shares at $655, down 6.1 percent. That was not the close. At about $1.77 trillion of market value before the drop, a fall of nearly 7 percent is on the order of $120 billion. Earlier on Thursday, Reuters used about $114 billion if the losses then on the screen held. “About $100 billion” is too low for the close.
The credit market said the same thing in a quieter voice. A credit default swap is insurance on a company’s debt. If the firm fails to pay, the insurance pays. Meta’s five-year insurance hit a record. A Markit screen cited by Reuters put the cost at 114.79 basis points, up 7.6 on the day. A basis point is one hundredth of a percent. Other screens were above 119. One note said these contracts have traded for about a year, and that this is the high since they started. A claim that the spread is more than double last year’s level comes from the wrap alone. It is not used here as fact.
A record insurance price is not a forecast of default. Meta is not a weak firm. It still holds about $82 billion of cash and a huge ad business. Shay Boloor wrote on October 22 that the jump could have more to do with banks hedging loans tied to the build than with a real fear that Meta stops paying. Both things can be true. The insurance got more expensive because more people wanted it. They wanted it on a day the firm said it will spend more, and showed that this quarter the spending already exceeded operating cash.
That is the crack the wrap was pointing at. For a year the argument was about demand for cheaper computing. October 22 was about supply of money. You can want all the chips in the world. Someone still has to pay the invoice.
The tape split on purpose
The day was not a crash of everything. It was a hit to the expensive growth that must keep spending.
The Nasdaq Composite fell 2.60 percent, to 26,008.86. The S&P 500 fell 1.21 percent, to 7,663.43. The Dow rose 0.02 percent, to 52,867.98. The Russell 2000, the small-stock index, rose 0.11 percent, to 2,784.03. Big industrial names and small firms held. The index full of companies priced for a long build did not.
Bonds were bought. The 10-year Treasury yield was 4.957 percent on the Barron’s screen, down 0.172, or about 17 basis points. A lower yield is a higher bond price. The wrap called it a panic bid, and gave four reasons that can sit together. Less fear of a flood of new tech debt. A scare that growth is slowing. Money leaving a shaky Europe. Traders who had bet against bonds being forced to buy. You do not need to pick one. The yield fell while the Nasdaq fell. That is not the pattern of a simple risk-on or risk-off day. It is the pattern of a market that suddenly cared about who has to borrow.
Oil did not follow the stock market. On the Barron’s futures screen, U.S. crude was $99.10, up 2.01 percent. A later settlement sheet had crude nearer $99.47, up about 2.4 percent. Same direction. The wrap tied the jump to fresh tension in the Middle East, including more attacks out of Yemen and talk of new U.S. and Israeli strikes. Treat that as the wrap’s cause, not as a finding from this desk. The price is the fact. Oil up, tech down.
The dollar eased. Gold did not catch a safe-haven bid. On that same Barron’s screen, gold futures were $4,039.80, down 2.66 percent, or about $110. Silver futures were $54.63, down 5.34 percent. A later sheet had gold near $4,043 and silver near $54.71. Bitcoin, which the wrap also called battered, traded near $81,200 on one daily history, down about 3 percent from a prior close near $84,100. Another wrap said bitcoin fell 3.7 percent and did not print a price. The direction was down.
Hold that set in one hand. A firm with a real business spent, this quarter, more than its operations produced. Its stock fell. The cost of insuring its debt hit a record. The growth index fell. The old industrial index did not. Bonds rose. Oil rose. Gold, silver, and bitcoin fell.
If your model says any scare buys gold, this day broke the model. The scare was not a war bid and it was not a money-printing bid. It was a cash bid. People wanted the asset that pays them to wait, which on this day was the Treasury bond, and they sold the assets that pay nothing while they wait. Gold can be a store of value over decades and still lose about $110 in an afternoon when the question in the room is who has promised to spend more than this quarter’s cash.
This did not start with one earnings call
Meta is the clean example because the business is so strong. The worry is older than Wednesday night.
For two weeks the weak point has been the layer that borrows to buy chips and rent power, not the firms that already sit on huge profits. CoreWeave, a renter of computing, has been the name in the middle of that storm. Reuters said on October 22 that it has lost more than 34 percent of its market value this month. Its bonds and its default insurance had already been telling a harsher story than the stock headlines. When a lender starts charging a high rate to a firm whose assets are chips that age in a few years, the theme is no longer whether there is demand. The theme is whether the borrower can roll the debt.
Meta is not CoreWeave. Meta’s ads still throw off enormous cash before the building bill, and the firm still holds $82.45 billion. That is why the negative free-cash-flow print matters more, not less. If a firm this rich can spend through a quarter’s operating cash, the firms with no ad machine have a thinner margin for error. The market did not wait to sort them. It sold the expensive growth and left the Dow alone.
Asia had already flinched. Japan’s Nikkei fell 2.65 percent. The wrap also described a bad day in Paris and a wider European scare. Those are the wrap’s words. The U.S. split is the part you can check on a closing sheet.
What a mining investor can actually use
You do not need a view on Muse, or on whether the Class C share is clever. You need the test the quarter just ran.
A project is not proved by the size of the plan. It is proved by what is left after the plan is paid for. Meta’s plan got bigger. Free cash flow went from a surplus of $10.63 billion in the year-ago quarter to a hole of $2.12 billion. The stock fell even though sales grew, and even though $82.45 billion still sat in cash and securities. That is the whole lesson, and it travels. A large pile can fund a hole. It does not make the hole disappear, and it does not stop the lender from charging more to insure the debt.
Look at a mine the same way. A feasibility study can show a bigger plant, a longer life, and a pretty net present value at a gold price the author chose. None of that is cash. Cash is metal sold, minus the build, the power, the labor, and the interest. If the build in a quarter is larger than the operating cash, the owner must sell shares, borrow, slow down, or draw down a pile. Meta chose a mix of those paths in public. A new class of stock. A credit market that immediately charged more to insure the debt. And a filing that lists hundreds of billions of dollars of leases and contracts still to be paid. A junior miner does the same thing with a smaller press release. The market’s answer does not get kinder because the commodity is gold instead of ads.
There is a second cut, and it is the one this day made obvious. Not every real asset moved together. Oil rose. Gold and silver fell. Copper and the power inside that stack of commitments are a different bet from a bar of gold. The build still wants metal and electricity. The stock market, on this day, did not pay up for the firms doing the spending. It paid up for bonds, and it left alone the companies that were not asking for fresh money.
That is not a signal to sell every gold miner, and it is not a signal to buy the copper name that supplies a data center. It is a filter. Prefer the claim that still produces cash if the build slows, or if the cost of money stays high. Be wary of the claim that only works if someone else keeps writing larger checks. Meta is the someone else. On October 22 it showed that even a firm with $82 billion in cash can spend past a quarter, and that the stock and the debt insurance will notice.
The wrap’s headline treated the day as a tech wreck, a bond squeeze, and a hit to bitcoin and gold. All of that happened. The cause underneath is smaller than the headline. One of the richest firms in the world grew, spent past the quarter’s operating cash, and watched the price of its shares and the price of its debt insurance move apart. Sales were the story they wanted. Cash was the story the market took.
A mine, a mill, or a data center is the same shape. The plan can be right and the quarter’s treasury can still be short. October 22 was the day the treasury won the argument.