Last checked
Short answer
The 10-year yield fell 19.3bp to 4.953% on Oct 22 as growth fears sent money from AI stocks into Treasuries, after months of both falling together. 5% matters as a marker, not a rule: Oct 22 was the first close below it since Sept 23, and the Treasury par curve printed exactly 5.00% on Oct 23, so the line is still in play into the Oct 28 FOMC decision.
- What happened The 10-year yield closed at 4.953% on Oct 22, down 19.3bp, a day after Meta reported negative free cash flow and a 2026 capex plan of $135-145 billion.
- Why it matters For most of 2026 bonds and stocks fell together. On Oct 22 the old hedge returned: growth fears, not inflation fears, drove the selling.
- What is priced CME FedWatch put a 25bp-or-more hike on Oct 27-28 at 9.4% on Oct 23; Polymarket showed 'No change' at 94% as of Oct 26.
- What to watch FOMC decision Oct 28 (no dot plot), Microsoft earnings Oct 28, Q3 GDP and September PCE at 8:30 AM Oct 29, Amazon call Oct 29.
Key facts
| 10-year yield, Oct 22 close (Reuters) | 4.953%, down 19.3bp, biggest drop since March 10, 2023 |
| 10-year par rate, Oct 21 to Oct 22 (Treasury) | 5.13% to 4.96% |
| 10-year intraday high, Oct 7 (NBC News) | 5.36%, highest since 2002 |
| Meta Q3 2026 free cash flow (8-K) | -$2,120 million; 2026 capex guidance $135-145 billion |
| Fed funds target range (Sept 16, 2026) | 3.75%-4.00% after a 25bp hike, 12-0 vote |
| September CPI (BLS, Oct 14) | 3.7% y/y all items, 2.6% core, energy +20.4% y/y |
What happened to Treasury yields on Oct 22?
On Thursday, Oct 22, the US 10-year Treasury yield fell 19.3 basis points to 4.953%, according to Reuters. That is the biggest one-day decline since March 10, 2023, the day regulators closed SVB. It was also the first close under 5% since Sept 23.
The move was not confined to one maturity. Reuters put the 2-year yield down 18.5bp at 4.638% and the 30-year down 15bp at 5.287%. The Treasury's official daily par yield curve tells the same story with slightly different decimals. On that measure the 2-year went from 4.81% to 4.65%, the 5-year from 4.95% to 4.78%, the 10-year from 5.13% to 4.96%, and the 30-year from 5.42% to 5.29%. The par curve is a 3:30 PM bid-side snapshot, not a closing quote. Neither is wrong; they are measured at different moments.
Stocks went the other way. Yahoo Finance recorded the Nasdaq Composite down 2.6% at 26,008.86, the S&P 500 down 1.2% at 7,663.43, and the Philadelphia semiconductor index down 6.7%, its worst day since June 23. Meta shares lost nearly 8% during the session, according to Reuters.
Friday brought a partial retracement. Reuters reported the 10-year up 3.7bp to 4.99%, the 2-year up 5.5bp to 4.69% and the 30-year up 2.9bp to 5.316%. The Treasury par curve shows the 10-year at exactly 5.00% on Oct 23. So the line held, but only just.
| Date | 2-year | 10-year | 30-year |
|---|---|---|---|
| Oct 15 | 4.77 | 5.10 | 5.43 |
| Oct 16 | 4.74 | 5.09 | 5.43 |
| Oct 19 | 4.77 | 5.15 | 5.48 |
| Oct 20 | 4.79 | 5.12 | 5.42 |
| Oct 21 | 4.81 | 5.13 | 5.42 |
| Oct 22 | 4.65 | 4.96 | 5.29 |
| Oct 23 | 4.69 | 5.00 | 5.33 |
Why did bonds rally when stocks fell, when they had been falling together for most of 2026?
For most of 2026, when stocks fell, bonds fell too. On Oct 22 they parted ways. The reason comes down to what frightened the seller.
A Federal Reserve FEDS Note published Oct 16 lays out the mechanism. When inflation is near target and expectations are anchored, the authors write, central banks can ease in response to adverse shocks, so bond prices rise when stock prices fall. Bonds then hedge stocks. The note argues that the loss of this property after 2021 reflects the inflation environment, not any change in bonds themselves.
On Oct 7, the 10-year yield rose as high as 5.36% and the 30-year to 5.73%, the highest for each since 2002, according to NBC News. That day's selling was an inflation and supply story: a global bond sell-off, oil above $100, heavy AI-related corporate borrowing, a 10-year auction that stopped at 5.3%, and September CPI that would print a week later at 3.7% year on year with energy up 20.4%. In that regime, bad news for stocks is bad news for bonds: both get repriced for higher rates.
Oct 22 was different. The trigger was Meta's results the evening before: negative free cash flow and roughly $825 billion of contracts and leases running out to 2036 (see below). That raises a growth question, not an inflation question. If the AI buildout slows, the economy slows, and a Fed that has been hiking has less reason to hike again. So the money that left chips and software went into Treasuries.
Reuters also noted French fiscal worries and Middle East tensions adding to the bid, so the AI scare was the largest driver, not the only one. Thomas Urano of Sage Advisory told Reuters: "The flight-to-quality trade is winning in the Treasury space." Joe Saluzzi of Themis Trading, quoted by Yahoo Finance, said investors were "rotating out of techs, out of the AIs, and into the more safety trades."
The switch can be seen in the shape of the move. The 2-year, the maturity most sensitive to the next few Fed decisions, fell almost as much as the 10-year (18.5bp against 19.3bp on Reuters figures). That is the market trimming its hike expectations, not just seeking a safe place to park money.
What did Meta actually report, and what is the $825 billion?
Meta's Form 8-K for the third quarter, filed Oct 21, shows revenue of $64,601 million, up 26% year on year, and net income of $9,860 million. Operating cash flow was $32,786 million. Capital expenditure was $33,942 million. Subtract capex ($33,942 million) from operating cash flow ($32,786 million) and free cash flow was -$2,120 million. The company guided 2026 capex to a range of $135-145 billion.
The 10-Q filed the same day discloses the longer tail. As of Sept 30, 2026, Meta had $454.52 billion of non-cancelable contractual commitments, about $355.73 billion of leases not yet commenced running through 2036, and contingent cloud obligations of up to $14.72 billion. Press and analysts added those three together to get roughly $825 billion. Meta's filing does not print that total, and it is not a debt figure. The actual long-term debt disclosed is $84.0 billion at face value. The rest is contracts and leases the company has promised to pay over a decade.
Meta also announced a new non-voting Class C share class to be distributed one-for-one as a dividend, which TechCrunch described as a way to fund AI infrastructure without diluting founder control. Reuters cited the share-class plan alongside the earnings as the trigger for the stock's drop.
Why does one company's cash flow move the whole bond market? Because the September FOMC minutes, released Oct 7, show several participants flagging the AI buildout as surprising to the upside, and the same minutes say most participants expected another rate increase by year-end. The buildout has been part of the case for higher rates. The first sign that it strains a balance sheet weakens that case.
What do markets price for the Fed this week?
The Fed is in a hiking cycle. On Sept 16, 2026, the FOMC raised the target range by 25 basis points to 3.75%-4.00% by a 12-0 vote and said inflation remains elevated. The next meeting runs Oct 27-28 and, per the Fed's calendar, does not include a Summary of Economic Projections, so there is no new dot plot.
Market odds of a move at that meeting have collapsed over the month. CME FedWatch pricing of a 25bp-or-more hike stood at 55.4% a month before Oct 22, 27.7% a week before, 7.2% on Oct 22 itself and 9.4% on Oct 23, according to Reuters. Yahoo Finance put the odds of a December hike at about 78% after Thursday's session. Earlier in October, Bloomberg reported swaps fully pricing a hike by year-end, so the market has moved the expected hike back rather than erased it.
Prediction markets agree on October. Polymarket's October Fed decision market showed 'No change' at 94% and a 25bp increase at 6% as of Oct 26. These are market odds at a stated time, not forecasts, and they move continuously.
Two other data points from Oct 22 cut against the growth-scare reading. Weekly jobless claims fell 6,000 to 194,000, which is not a weak labour print. And the 10-year TIPS breakeven, a market gauge of expected inflation, was 2.333% according to Reuters. The bond market rallied on Thursday despite, not because of, that claims number.
| Measure | Reading | Date | Source |
|---|---|---|---|
| Odds of 25bp+ hike at Oct 27-28 FOMC | 55.4% | one month before Oct 22 | CME FedWatch via Reuters |
| Odds of 25bp+ hike at Oct 27-28 FOMC | 27.7% | one week before Oct 22 | CME FedWatch via Reuters |
| Odds of 25bp+ hike at Oct 27-28 FOMC | 7.2% | Oct 22 | CME FedWatch via Reuters |
| Odds of 25bp+ hike at Oct 27-28 FOMC | 9.4% | Oct 23 | CME FedWatch via Reuters |
| October decision: No change / 25bp increase | 94% / 6% | Oct 26 | Polymarket (polymarket.us) |
What could push the 10-year back over 5% this week?
The calendar is unusually dense. On Wednesday, Oct 28, the FOMC announces its decision. With no dot plot, the statement wording and press conference carry the weight. A statement that repeats 'inflation remains elevated' and leaves the door open to a December move would speak to the 2-year first, and the 10-year would follow.
Also on Oct 28, Microsoft reports its fiscal first quarter, per its investor relations site. On Thursday, Oct 29, at 8:30 AM, the Bureau of Economic Analysis publishes the advance estimate of Q3 2026 GDP and September Personal Income and Outlays, which includes PCE inflation, in a single release. Amazon's Q3 call follows at 2:00 p.m. PT / 5:00 p.m. ET.
Here is how the pieces fit. If the next big AI spenders show the same cash-flow strain as Meta, the growth-fear channel stays open and the bid for Treasuries persists. If GDP and PCE come in hot, the inflation channel reopens and the Oct 7 pattern, where bonds and stocks sold off together, can return. JoAnne Bianco of BondBloxx Investment Management told Reuters: "We're going to see yields over 5 percent for the 10-year again. We're not that far away." That is one strategist's view. The par curve's Oct 23 print of exactly 5.00% shows how little separates the two readings; nothing in the data says which way it resolves.
5% is a reference point, not a rule. The 10-year had closed above it every day since Sept 23 until Thursday, and the par curve printed exactly 5.00% on Friday. To check where it stands, read the Treasury daily par curve (posted each evening) against the Reuters closing quote; as noted above, the two differ by a few basis points.
The level matters outside the US because the 10-year anchors mortgage rates, corporate borrowing costs and the discount rate applied to long-dated assets. The Treasury par curve shows the 10-year sat between 5.09% and 5.15% for the whole week before Thursday. A sustained move below 5% is a change in the cost of money for everyone, not a US-only story.
- FOMC meeting Oct 27-28; decision Oct 28, no Summary of Economic Projections
- Microsoft FY27 Q1 earnings, Oct 28 (per Microsoft IR)
- BEA advance Q3 2026 GDP and September PCE inflation, Oct 29 at 8:30 AM
- Amazon Q3 2026 earnings call, Oct 29 at 2:00 p.m. PT / 5:00 p.m. ET
What is not confirmed
Several numbers in circulation this week do not survive a check against primary sources, or conflict between outlets. We flag them rather than use them.
The size of Meta's share drop on Oct 22 varies by outlet and by measurement point. We use Reuters' 'nearly 8%', which refers to the session, and do not cite a closing figure because we could not confirm one from a primary source.
Alphabet's Oct 28 call time comes from secondary calendars rather than its investor page, and Apple's date is unclear across calendars, so neither appears in our schedule above. The FEDS Note's correlation statistics are omitted because the sign convention was not confirmed; only its verbatim reasoning is quoted.
FAQ
Why did the 10-year Treasury yield fall below 5%?
On Oct 22 investors sold AI-linked stocks after Meta reported negative free cash flow and large future commitments, and moved into Treasuries. Reuters recorded a 19.3bp fall to 4.953%, the biggest one-day drop since March 10, 2023.
Why do bond yields usually rise when stocks fall in 2026?
Because the selling has been driven by inflation. A Fed FEDS Note from Oct 16 explains that bonds hedge stocks only when inflation is anchored and the central bank can ease into shocks; with September CPI at 3.7% and the Fed hiking, that condition has not held.
What is the $825 billion figure for Meta?
It is the sum of three 10-Q disclosures: $454.52 billion of non-cancelable commitments, about $355.73 billion of leases not yet commenced through 2036, and up to $14.72 billion of contingent cloud obligations. Meta's filing does not print the total, and it is not debt; long-term debt is $84.0 billion at face value.
Will the Fed raise rates on Oct 28?
Markets price it as unlikely. CME FedWatch put the odds of a 25bp-or-more hike at 9.4% on Oct 23, and Polymarket showed 'No change' at 94% as of Oct 26. The September minutes show most participants expected another hike by year-end, and December odds were about 78% after Oct 22 per Yahoo Finance.
When are the next US data releases that could move yields?
The FOMC decision comes Oct 28. The BEA publishes advance Q3 2026 GDP and September PCE inflation together at 8:30 AM on Oct 29.
Sources
- U.S. Treasury – Daily Par Yield Curve Rates (Oct 2026)
- Meta Platforms 8-K Ex. 99.1 Q3 2026 results (SEC EDGAR)
- Meta Platforms 10-Q for quarter ended Sept 30, 2026 (SEC EDGAR)
- Federal Reserve – FOMC statement Sept 16, 2026
- Federal Reserve – FEDS Note: Government Bonds' Safe-Haven Property and Inflationary Pressures
- BEA – News release schedule
- Reuters via The Globe and Mail – US Treasury yields plummet as fears about AI boom sustainability flare
- Reuters via Business Recorder – US yields climb, but 10-year holds under 5pc
Figures change. Check the primary sources above.