When the 10-year yield jumps, the stocks that fall hardest are often the ones whose profits are supposed to arrive later. People say duration and then stop, as if the word had finished the job. It hasn't. Duration is a statement about timing. A dollar promised in thirty years changes value, when the discount rate changes, by more than a dollar promised in two years. The arithmetic is ordinary. Pretending a share of stock is that thirty-year dollar is the sloppy part.
I want the arithmetic in public numbers, with a filing next to it, so I can see how much of a real company is actually dated. The yields are the Treasury constant-maturity series on FRED, from the Board of Governors' H.15 release, quoted on an investment basis, in percent, not seasonally adjusted. The cash flows in the example are invented. One hundred dollars, paid once, on a date I choose. Nothing in the example is a price target.
The yields on those two days
I pinned the example to November 5, 2026. Series DGS2, the 2-year constant maturity yield, finished that day at 4.46 percent. The 10-year, DGS10, was 4.84 percent, with the 30-year, DGS30, at 5.25 percent. DFII10, the 10-year inflation-indexed yield, the TIPS yield, finished at 2.60 percent. On November 5, 2025 the same four closes had been 3.63, 4.17, 4.74 and 1.87 percent.
Over that year the 10-year nominal yield rose 0.67 percentage points, and the 10-year real yield rose 0.73 percentage points. The gap between the nominal 10-year and the TIPS 10-year was 2.24 percentage points on the later date, 2.30 on the earlier one, so most of the year's move in the nominal yield was the real yield. Inflation compensation sitting between the two series barely changed. I wouldn't treat the gap as a clean inflation forecast anyway. TIPS yields also pick up how easy the bonds are to trade. Cash far in the future is still when I look at DFII10. A higher real yield is a higher hurdle after a rough allowance for inflation, and the nominal yield mixes that hurdle with the inflation piece.
| FRED series | Nov 5, 2025 | Nov 5, 2026 | Change, points |
|---|---|---|---|
| DGS2, 2-year | 3.63 | 4.46 | +0.83 |
| DGS10, 10-year | 4.17 | 4.84 | +0.67 |
| DGS30, 30-year | 4.74 | 5.25 | +0.51 |
| DFII10, 10-year real | 1.87 | 2.60 | +0.73 |
Figures are percent, daily closes, not seasonally adjusted. FRED attributes them to the Board of Governors. A constant-maturity yield is a constructed point on the curve. It isn't the coupon on a bond you can buy with that exact maturity today. The method sits in the H.15 notes and in the Treasury's yield-curve description, which is where I stop. Discounting a made-up cash flow at a published yield doesn't need the fitting procedure, and I don't pretend it does.
One hundred dollars, three dates
Take $100 paid once, with no coupons along the way. That's a zero. I discount it with the 10-year yield alone, first at 4.17 percent and then at 4.84 percent. One yield for every maturity is a flat curve, and the real curve on November 5, 2026 was not flat: the 2-year was 4.46 and the 30-year was 5.25. I flatten it on purpose. Change the yield as the maturity changes and you lose the pure effect of waiting. Present value equals 100 divided by (1 plus the yield) raised to the number of years. The yields go in as decimals, 0.0417 and 0.0484.
At 4.17 percent, $100 due in two years is worth $92.15 today. Give that claim ten years and it's worth $66.46. At thirty years you're looking at $29.36. Repriced at 4.84 percent, the three figures are $90.98, $62.33 and $24.22. The two-year claim lost $1.17, which is 1.3 percent, a little more than a dollar on a ninety-two dollar starting value. Ten years out the decline is $4.13, or 6.2 percent. Thirty years out it's $5.14, and that is 17.5 percent, off the same 0.67 point move in the yield.
| When the $100 arrives | PV at 4.17% | PV at 4.84% | Change |
|---|---|---|---|
| 2 years | 92.15 | 90.98 | -1.3% |
| 10 years | 66.46 | 62.33 | -6.2% |
| 30 years | 29.36 | 24.22 | -17.5% |
Present values are dollars per $100, rounded to the nearest cent. The percent change is the ratio of the unrounded values, then rounded to one decimal. You can rebuild the cent from 100 divided by 1.0417 or 1.0484 to the power of the year. I use the November 5 closes because they are a real move over a real year. A round hundred basis points would have divided more cleanly, and it would have been a number I picked.
The long one moves more because the multiplication repeats. Each extra year multiplies the present value by 1/(1+y) again. Raise y and that factor shrinks, and then you apply the smaller factor more times. Traders shortcut this with modified duration. For a zero it's the number of years divided by (1+y). At the starting yield of 4.17 percent, the thirty-year zero has a modified duration of 30/1.0417, about 28.8. Multiply 28.8 by 0.67 percentage points and you get about 19.3 percent, against a full discount that fell 17.5 percent. The shortcut is nearly two percentage points too large on the long cash flow. Duration is a straight-line estimate, and 0.67 points is big enough for the curve of the formula to matter. On the two-year zero the shortcut and the full discount almost agree. I trust the table when I want the dollars. Duration is a reason the rows look like that, and I drop it once the question is the dollars.
A Treasury yield is not the stock's discount rate
An equity investor doesn't discount a company's cash at the 10-year Treasury yield. The Treasury yield is the part of the hurdle that comes from time and from the government's credit. The stock adds a premium for the chance the cash never arrives, and for the fact that you can't put the cash flow in a museum and hold it to the date. I'm not going to invent that premium. If I did, every present value in the table would be smaller, and the far row would shrink by more than the near row, which only sharpens the shape. The table would also become a fake valuation. What I actually have is a comparison of dates at a yield I can cite.
Ask the same question with the real yield and you get a different answer. Discount the same $100 due in thirty years at the November 5, 2026 TIPS yield of 2.60 percent and the present value is $46.30. At the nominal 4.84 percent it was $24.22. The larger number is a present value after stripping out the inflation yield. It isn't 'what the cash is worth in today's dollars' unless you also believe the TIPS yield is a pure real rate, which I don't, for the liquidity reason above. I keep both. A year when DFII10 rose 0.73 points and the nominal-minus-TIPS gap went from 2.30 to 2.24 is one I describe as a higher real hurdle. If only the gap had widened, that would be an inflation story. This was the first kind.
What the companies have actually dated
Microsoft's Form 10-K for the year ended June 30, 2026 says revenue allocated to remaining performance obligations was $684 billion. The commercial portion was $678 billion, with a weighted average duration of approximately 2.3 years. The company expects to recognize approximately 30 percent of that revenue over the next 12 months, and the rest later. Fiscal 2026 revenue was $331,839 million. Put the commercial book next to that revenue and it's roughly two years of it, on a clock of about 2.3 years. Thirty it is not. See Microsoft.
That 2.3 years is the contracted book, unearned revenue plus amounts still to be invoiced. It isn't the stock. The share price also capitalizes revenue nobody has contracted, products that don't exist yet, and the margins on both. Equity duration actually lives in that uncontracted tail, and Microsoft does not publish a number for it. The disclosed book would sit nearer the two-year row of my table than the thirty-year row. I can't say that about the stock. A 0.67 point rise in DGS10 is not a license to knock 17.5 percent off Microsoft because a thirty-year zero did that.
The same 10-K has a rate sensitivity that people grab when they want a dollar figure. In the market-risk table, a 100 basis point increase in U.S. Treasury interest rates is a $1,201 million loss of fair value. Read the header before you borrow the number. That column is fair value, and the row is the interest-rate risk of the instruments in the sensitivity analysis, alongside a foreign-exchange row and a credit row. Bond book. Not the duration of the common stock. I've watched a writeup paste the $1,201 million onto the equity story. The filing went to the trouble of naming the object, so I try to name it the same way.
Nvidia's Form 10-K for the year ended January 25, 2026 is the other end of the disclosed clock. Revenue was $215,938 million. Revenue related to remaining performance obligations from contracts greater than one year was $2.3 billion, of which $1.9 billion was already in deferred revenue and $390 million had not yet been billed. About 42 percent of that long-contract revenue is expected over the following twelve months. Set $2.3 billion next to $215,938 million and the long contracts are a small slice of a single year of sales. Nvidia's customers are largely paying for what is being delivered now, on orders that can slip. That's a short cash-flow fact. The stock, at the multiples people pay, is a claim on shipments that aren't in that footnote. The duration is in the hope, not in the performance-obligation note. See Nvidia.
I keep a company whose earnings are mostly this year's product next to those two, so the word 'growth' has something to scrape against. Coca-Cola's 2025 net operating revenues were $47,941 million, and operating income was $13,762 million, built from concentrate sold to bottlers and from the finished-product businesses it still owns. You can see that stream in the current income statement. It can still be a long-lived franchise. The cash just recorded isn't a thirty-year zero. A yield backup hits it through the valuation people put on the franchise that continues, which is a judgment, and through whatever debt reprices. The second of those is a cash-interest question, and it has its own clock. See Coca-Cola.
The interest bill is a different clock
A higher Treasury yield can reprice the share this afternoon and leave the coupon unchanged for years. Floating debt, commercial paper and the next maturity are how the yield eventually becomes cash interest. I separated those dates in when a rate move reaches the cash. This piece is the other channel, the one that reprices distant profits whether or not a single coupon changes. It's the price-of-money channel in how world events reach your portfolio.
Currency translation runs on another clock entirely. A strong dollar changing reported revenue is about sales that already happened in another currency, which I went through with the broad dollar index in a strong dollar and reported earnings. Microsoft shows up in both notes for a reason. The company has foreign revenue, and it has a long uncontracted cloud story. One headline about 'rates and the dollar' is two positions. I write them on two lines.
There's a further way the yield hurts a growth company, and the zero never captures it. If customers finance equipment, or delay a cloud contract because their own hurdle rate moved, the cash flow itself shrinks. My table holds the $100 fixed and only changes the discount. A business where the $100 was never contracted can lose the $100. Nvidia's long-contract footnote being small is why that second effect matters there. Microsoft's $678 billion is contracted and still has a 2.3 year average life, so a chunk of it arrives even if new bookings slow. I'd rather know which of those two I own before I call the stock 'long duration' and go to lunch.
What I check before I call it duration
- Date the yield you are using. The 10-year close was 4.17 percent on November 5, 2025 and 4.84 percent on November 5, 2026. A 'yields are up' headline without those two numbers is a mood.
- Nominal and real have to be split before you name the year. Over that year DGS10 rose 0.67 points and DFII10 rose 0.73. The gap between them went from 2.30 points to 2.24. I call that a real-yield year.
- Rank the stocks only after a near cash flow and a far one have been run at the same yield. On these closes the two-year zero fell 1.3 percent and the thirty-year zero fell 17.5 percent. The duration shortcut on the long zero, about 19.3 percent, overshot the full discount.
- The performance-obligation note, when the company prints one, is worth opening. Microsoft's commercial book was $678 billion with a weighted average duration of about 2.3 years. Nvidia's contracts longer than a year were $2.3 billion against $215,938 million of revenue. The stock can still be long-dated. The footnote tells you whether the length is contracted.
- Microsoft's $1,201 million figure is a 100 basis point Treasury move against fair value in that table. Leave it on the bond portfolio. It isn't an earnings guide for the share.
- Keep the coupon on its own line. Fixed debt makes this yield move a valuation fact until a maturity or a reset, and the cash date is a different guide.
I still react when the 10-year moves. Less than I used to. I got tired of explaining a 17 percent zero with a stock that had a 2.3 year contracted book and a completely unpublished tail. The tail is the bet. The table is the shape of that bet, and the filing is how much of it the company has already been paid for.
Educational reading of yields and the timing of cash flows. Not investment advice, and not a valuation of any company. Yields are daily closes from FRED, Board of Governors H.15, on November 5, 2025 and November 5, 2026: DGS10, DGS2, DGS30, DFII10. Present values are the author's discounting of a hypothetical $100 zero at those 10-year nominal closes, and one figure at the 2.60 percent TIPS close. Company figures are from Microsoft's Form 10-K for the year ended June 30, 2026 (filing), Nvidia's Form 10-K for the year ended January 25, 2026 (filing), and Coca-Cola's Form 10-K for the year ended December 31, 2025 (filing).