# Same River, Different Water

> Notes on The Compound and Friends, September 11, 2026: Jens Nordvig on how the force behind Korea's stock market changed hands, why hyperscaler bonds and US Treasuries now compete for the same buyers, Bessent's 'I am the house,' and which data to check when retail stocks collapse. Educational, not investment advice.

Published: 2026-09-11
Locale: en
Tags: macroeconomics, Treasury yields, Korean stocks, capital flows, investor psychology


![At dusk on a stone riverside quay, paper lanterns drift toward the viewer on a wide river, while far down the bank enormous data-center buildings glow cold white among construction cranes](/covers/compound-2026-09-11-the-most-interesting-macro-moment-of-my-lifetime-w-cover.png)

> Heraclitus says that everything moves and nothing stands still; comparing the things that exist to the flow of a river, he says you could not step into the same river twice.
>
> —— Plato, *Cratylus*, quoting Heraclitus (c. 360 BC; translation mine)

The river keeps its name. The water passing your feet is a new batch.

That line kept running through my head while I listened to this episode of The Compound and Friends (September 11, 2026 — Josh Brown and Michael Batnick hosting, with guest Jens Nordvig). Jens opened by saying that in all his years doing macro, he has never seen markets more interesting than they are now. By the end, his "interesting" had a shape: many of the market patterns we are used to still carry the same names, while the forces underneath them have been swapped out.

## What the episode is about

Jens Nordvig founded the macro data firm Exante Data and this year merged it into Vanda, which tracks capital flows and positioning; he is now its president. He used to be head of research at Nomura, was ranked the top currency strategist by Institutional Investor five years running, and has known Treasury Secretary Scott Bessent for years.

The first stretch covers his business and how AI is changing research. The middle is the main event: long-end Treasury yields, hyperscaler bond issuance, Bessent's yen intervention and Treasury buybacks, and the inflation coming from energy and AI. In the back half Michael brings out a stack of consumer data and argues about what the collapse in retail stocks means. There's a light opener too: Josh complains about a friend who answers the phone with "living the dream" — sincerely — and gets teased for not being able to stand his friends doing well.

## The main points

**1. Data alone doesn't sell; clients want someone to vouch for it.** Exante Data tried a data-only model early on and found that portfolio managers are too busy — they need someone to tell them which number matters right now. The more interesting part: the highest-end clients want to get him on the phone more than they did a few years ago, because so much unvetted information is floating around. Josh added his own heuristic: rich people don't talk to robots. They pay someone else to talk to the robots and bring the answer back.

**2. Korea: the same rally, a different pusher.** Jens thinks that relative to its size, Korea is the country where the AI boom hits hardest: half the stock market is memory, the trade surplus is exploding, and the won just had its best three-month run ever. Korean stocks used to be driven by foreigners — foreigners buy, the market rises. This summer local retail investors piled in, the market rose so much that foreigners had to sell to rebalance, and the relationship between foreign flows and the index flipped. Anyone reading foreign flows with the old model would get the direction backwards.

![Two side-by-side panels: in the left "Past" panel, the foreign-investor arrow and the index arrow both point up in the same direction; in the right "This summer" panel, local retail investors and the index point up while the foreign-investor arrow points down, so the two now move in opposite directions.](/figures/korea-foreign-flow-flip-en.svg)

**3. Hyperscalers are issuing as much long-dated debt as the US government.** By his team's estimate, at the current pace a handful of hyperscalers are issuing as much at the long end of the curve as the US Treasury. The US deficit is around 6% of GDP, on par with Reagan's biggest, and a few private companies are still matching it in the same part of the curve. He stressed that this is nonlinear: early capex was paid out of free cash flow, but now most of every extra $10 billion turns into new bonds, and next year will be bigger.

![Four bars rise year by year, with a dashed line marking the ceiling that cash flow can cover; the first two years stay below the line, while in the last two years the portion above the line is shaded blue to represent newly issued debt, and that portion grows larger each year.](/figures/bigtech-capex-debt-overflow-en.svg)

**4. The same 5%, on a different debt base.** The 10-year was around 4.92% at the time; the 30-year is back at 2007-era highs. People say "we've been here before." Jens's answer: the last time yields sat at these levels for long was before the financial crisis, when debt levels were nowhere near today's, and the Congressional Budget Office projects US debt-to-GDP going from 100% to 200% within a couple of decades. He sees three signs: long-end yields have decoupled from economic data for three or four months, globally; currencies of low-debt countries (Switzerland, Singapore, Australia) are outperforming; and the dollar shows a residual that rate spreads and risk sentiment can't explain, "leaking" weaker. Among the top 30 economies, the worst debt metrics now belong to the US rather than Japan.

![A horizontal blue line represents the long-term yield and sits at the same height in 2007 and today; below the line, the debt bar is short in 2007 and reaches 100% today, with a dashed box stacked on top marking the projected 200% twenty years from now.](/figures/same-yield-heavier-debt-en.svg)

**5. Bessent says, "I am the house now."** After the yen intervention he said he has asymmetric information and anyone who wants to bet against him is welcome to. Jens's read: when the intervention happened, the Bank of Japan had just passed in July and a September hike was priced at three or four basis points; afterward it rose above 90%. Bessent was right. But the policy goals conflict: the administration wants a weaker dollar against Asian currencies to help manufacturing while it has an inflation problem, and it wants the AI sector to carry growth without letting long-end yields run away — hence the buybacks.

**6. AI may push prices down in the long run; the build-out is pushing them up now.** Data-center demand is driving up memory prices, which is part of why Korea benefits. Jens cited Elon Musk: even stepping on the gas, supply might rise about 20% in a year or two, when what's needed is 100%. Add oil, a conflict with Iran now more than six months old with the Strait of Hormuz still blocked, and 5-year breakevens moving from 2.2% to 2.4–2.5%, and it's hard for the Fed to relax.

![Three horizontal bars run from top to bottom: today is the baseline of 100, while in one to two years supply grows only 20% and demand needs to double; a gap is marked between the end of the supply bar and the end of the demand bar, representing memory prices being pushed higher.](/figures/memory-supply-demand-gap-en.svg)

**7. Gold has run on a different logic since 2022.** Jens used to find gold boring: if you knew what the dollar and real rates were doing, you knew roughly what gold was doing. After 2022, and especially from mid-2023, that link broke. The buying came in three waves: China accumulating heavily, broader central bank buying, and last year's speculative wave into January (ETFs and GLD options). He doesn't expect the options frenzy to repeat, so targets should be tempered, but ETF inflows — especially from Europe — are steady, because people are uncomfortable owning long bonds and want gold as a hedge.

![An illustrative line chart: on the left, the actual gold price and a dashed line estimated from the dollar plus real interest rates almost overlap; after mid-2023 the dashed line flattens and drifts lower while the actual gold price keeps climbing and splits away, with three colored bands below marking the three waves of drivers in order China, central banks, and a speculative wave and the last band labeled as unlikely to repeat.](/figures/gold-old-formula-broke-en.svg)

## Going further

### Yields keep climbing and stocks don't care. Am I too nervous?

The feeling I've had most often these past weeks is a split: oil up, mortgage rates near 7%, the 10-year heading for 5%, and yet the VIX sits at 18, the equal-weight S&P is about 4% off its highs, and we just had one of the nastiest tech momentum crashes on record last month. The headlines all scream danger, the index barely reacts, and you start wondering whether you're overthinking or the market is being naive.

After this episode I broke the question into three layers.

First, the calm has reasons. Josh put it plainly: the market isn't especially expensive, earnings are growing 20%-plus, and every big tech company says capex won't slow. Until someone changes the story, the market will act on it. His example was Meta, at 17 times forward earnings two weeks ago with nobody wanting it.

Second, where the pressure is building. Jens's worry sits in the long end: another 20 to 35 basis points in a short time could push into territory where equities can't take it, and we're close to that line. This kind of risk is nonlinear — nothing happens until a threshold is crossed, then the reaction comes all at once — so "fine so far" offers limited comfort.

![A curve stays almost flat on its left side, then drops steeply to the bottom after crossing an orange zone; a blue dot labeled "Now" sits right at the end of the flat section, with a note below saying another 20 to 35 basis points would push it into the steep-drop zone.](/figures/long-yield-cliff-threshold-en.svg)

Third, rather than guessing whether it will crash, I'd rather know which concrete thing to watch. Josh offered a signal I love: the day Alphabet or Meta comes to market with a bond deal and instead of being two or three times oversubscribed it underwhelms, the rate has to go up, or the deal gets pulled — that's when the party's over. He said it would be poetic if this bull market ended because the bond market finally said no to Amazon: companies that were the greatest cash generators in history choosing to become the largest debtors in history. In noise-versus-structure terms, a day's jump in oil is noise; hyperscaler issuance accelerating year after year is structure, and the second is what I want to watch. My own habit is to write a card: which events would change my mind — say, weaker demand for a hyperscaler bond deal, or 5-year breakevens continuing to climb. The daily news only gets checked against that card.

### Retail stocks are getting crushed. Is the consumer cracking?

This is what friends ask me most: Dick's Sporting Goods down 38% in a month, American Eagle down 15% in a day, headlines about consumers pulling back — should we worry about the economy?

Michael did something on the show I think is worth copying: he set the stock prices aside and went to the source. He built a composite of the dozen-plus retailers that have reported same-store sales since 2001, and it's up 5.1% year over year; Bank of America's credit and debit card spending per household is up 5%, inflation-adjusted household savings are still above 2019; and Bank of America CEO Brian Moynihan said August spending ran about 4% above last August. His conclusion: consumer discretionary stocks have been falling relative to the market for a decade and that line has predicted nothing. The stock price mixes in valuation, company-specific risk, and fund managers simply not wanting to own the names — too much noise.

Josh then cut it a different way: split discretionary into experiences and things. The basket of golf courses, airlines, hotels, and concerts moves with the card data; the basket of stuff that fills garages is the one falling. Delta is expanding business class.

![An illustrative line chart: on the left, a single combined "discretionary spending" line splits in the middle, where the blue experiences basket rises and tracks the dashed credit-card spending line while the orange goods basket keeps falling.](/figures/discretionary-two-baskets-en.svg)

Then Jens added a caution that took back half my confidence: this year brought big tax refunds (no tax on overtime, no tax on Social Security and similar measures, paid out only after people filed this year), many people got the money in May, and part of the spending data is that money. The question is how long it lasts: spent in two or three months, there could be a gap afterward; spread over six, there are a few months left.

![Two side-by-side monthly bar charts use gray for normal spending and blue for the added tax-refund portion, and the blue areas on both sides are equal in size; on the left the blue is tall and packed into three months before dropping back to baseline, which is marked as a cliff, while on the right the blue is short and spread across six months.](/figures/tax-refund-spending-pace-en.svg)

Stacked together, the order I take away is: don't treat a falling stock price as the conclusion; go find the data closest to the facts; and when that data looks good, ask whether something one-off is propping it up.

## Worth a look

- Jens Nordvig on X at @jnordvig, and Vanda's website (he mentioned the research arm will go by Vanda Macro Research from October)
- The Congressional Budget Office's long-term budget outlook, for the projected path of US debt-to-GDP
- Bank of America Institute's monthly consumer spending tracker
- The St. Louis Fed's FRED database: 5-year breakeven inflation (T5YIE) and the 10-year Treasury yield (DGS10)

## The one thing to take away

The reason a rule works can change while the rule itself stays in your head and keeps getting used. This episode had three of them: read Korean stocks through foreign flows, gold follows real rates, retail stocks tell you about the consumer. Each was right once; the force behind it changed, and the people using it didn't notice.

One thing I've tried: pick a rule of thumb you've relied on for three years or more, from any part of life — "this route is the fastest to work," "when the baby cries, she's hungry," "this shop is emptiest on weekends," "when this colleague replies slowly, it means no." Write two lines on paper. The first says why it was true when you learned it — the construction site wasn't on that road yet, the baby wasn't teething yet. The second says whether that reason still holds today.