We're trying a new format for our monthly note on markets. Please let us know whether you like it or not.

This edition carries data through 21 September and covers the four weeks from 24 August. It replaces the version we circulated on 10 September. A great deal happened in the eleven days between the two, and we would rather reissue than let a stale note stand.

TL;DR

A war premium arrived in the oil market. Escalating geopolitical tensions in the Middle East took Brent from $92 to a $130 in three weeks. It's come down a bit but crack spreads and consumer prices for gas and diesel don't seem to be coming down regardless of what oil prices are doing. This tells me that the price of oil and the price of gas/diesel have decoupled. The most likely reason is getting oil out of the middle east is not the limiting factor in the oil supply chain. It's refining. Much of the worlds refining capacity is either stuck in the middle east along with the oil or coming under pressure due to Ukraine's counter offensive into russia. The remaining refiners are running at full capacity but can't make up the shortfall created by these two areas. Buffer stock is largely depleted and I don't see the price coming down in a real way any time soon. This has implications for CPI/PCE and interest rates.

Headline consumer prices rose 3.35% in the year to August, up from 3.30%, with gasoline up 27.4% and reaccelerating even though crude spent August roughly 30% below its March high. That is the refining bottleneck above showing up in the official data. Core PCE, the Fed's preferred inflation metric, has been stuck at 3.34% for two months and has been above 2% for 64 consecutive months. And all of this is happening with diesel, which is used to transport just about everything, is at an all time high. I don't expect PCE to get closer to goal any time soon. In response,the Federal Reserve raised interest rates. Fed increased its policy rate by a quarter point to 3.75-4.00%, its first increase since 2023, on a unanimous 12-0 vote. Sixteen of eighteen participants expect at least one more this year. Chair Warsh said inflation "is too high, and has been for too long." This, in turn, had an effect on the bond market which repriced. How it repriced is the interesting part.The 2-year yield rose 43 basis points, the 10-year 24, the 30-year only 6. Essentially all of the 10-year move was in the "real" (after-inflation) component; the inflation-expectation component did not move. At five years, expected inflation actually fell. So the bond market watched oil spike, watched the Fed tighten, and concluded that inflation is less of a long-run problem, not more. It is pricing a central bank that will do what it takes. It reamins to be seen if this is a good bet. No matter what the better angels of the Fed governers are telling them, aggressively raising rates to battle inflation could affect the other side of its mandate, full employment. It also makes interest payments on the federal deficit more expensive.

On the equities side of things. The market had some big moves this summer but coming into fall, its within a few percent of all time highs. While prices of AI tokens used stabalized at the end of the summer the capex to increase AI capacity just keeps going up. As does earnings. Q3 earnings are expected to be up 24%. While the market narrative around AI is bubbly, the valuations of the big AI players like NVDA and the hyperscalers are actually becoming more reasonable, not less. Not very bubbly behaviour. While the capex train may crash as some point, it doesn't look like valuations will be the cause. Of course this is a new, world changing technology who's biggest boosters are asking for more regulation because it may be a danger to humanity. So there's still plenty that could run us off the rails. It just won't be the same thing that cause the dot com crash.

Part 1: The data

These are all the numbers we're keeping an eye on here at BFG.

Interest rates and inflation

Source: US Treasury daily yield curve rates (2-, 10- and 30-year Treasury yields; 10-year inflation-protected yield); Federal Reserve Bank of St. Louis FRED database (Federal Reserve Board broad dollar index; market-implied inflation; 2-year to 10-year spread; ICE BofA corporate bond spreads; Cboe VIX; US Department of Labor weekly jobless claims). Data as of the dates shown, latest September 18, 2026.

The Fed's own published projections now show a median expectation of one further increase this year, with sixteen of eighteen participants penciling in at least one and four seeing room for two. The projections also pushed expected inflation for 2026 up a tenth, to 3.7% headline and 3.4% core, and do not have inflation back at 2% until 2029. The market read it as hawkish. The 10-year yield initially fell on the announcement, then reversed during the press conference and closed the day back at 5.01%, its highest in twelve months. It has since eased to 4.94%.

Over the four weeks the 2-year rose 43 basis points against the 30-year's 6. That is a classic "bear flattener": the front end selling off because the policy path has been repriced, while the long end barely moves. The gap between 2-year and 10-year yields compressed to 25 basis points, the narrowest in twelve months.

Of the 10-year's 24 basis point rise, 23 came from the after-inflation component and 1 from expected inflation.

The bond market's inflation expectations, at 17 September, with 1 September for comparison:

Source: US Treasury daily par yield curve and par real yield curve rates, September 1 and September 17, 2026. Implied inflation is the nominal yield less the after-inflation yield, calculated by Bergenn Financial Group.

Expected inflation fell at every maturity. Two potential readings of this data. The first is that investors believe a tightening Fed will not let an energy shock transmit to inflation so a supply disruption is a tax on growth rather than an inflationary impulse. The second is that the oil move was always expected to be temporary. Either way, the asymmetry we flagged last month has widened rather than closed. After-inflation yields sit at the very top of their range while expected inflation sits in the middle of its range at ten years and in the bottom quartile at five. Investors are charging a great deal to lend long and close to nothing for inflation risk.

August consumer price data arrived during the window. Headline inflation ticked up to 3.35% from 3.30%, which is energy showing through. Core inflation, which strips out food and energy, edged down to 2.45% from 2.47% and is now at its lowest in a year. The Fed's preferred measure, core PCE, has not updated since July and remains 3.34%. I think this is why there's no inflation premium in the yields being asked for by investors. As the increased diesel price propagates through the economy (Fuel cost for goods transport is +13.9% y/y in the latest data ) we may see that inflation premium increase with will put a further pressure on yields.

Core PCE now runs 0.88 percentage points hotter than core CPI. Since 2000 core PCE has run below core CPI in the typical month, by about a quarter point. The inversion has been widening all year, from 0.23 points last November. Core CPI at 2.45% and falling says inflation is nearly back to target. Core PCE at 3.34%, unchanged for two months and above 2% for 64 consecutive months, says it has not improved since June. The Fed targets PCE, which is a large part of why it raised rates while the more widely quoted measure was at a twelve-month low.

Gasoline prices were 27.4% higher than a year earlier in August, up from 24.6% in July, at a time when crude oil spent the month roughly 30% below its March level. Pump prices have decoupled from the crude price, which points to refining capacity rather than crude supply as the binding constraint.

August's consumer price data covers August. September's crude spike, which peaked on 15 September, is in no inflation release yet. The first reading that captures it is the September consumer price report in mid-October, and the first on the Fed's preferred measure lands later still. So the inflation data available to the Fed when it raised rates on 16 September contained none of the energy shock.

Interest rates and inflation, twelve-month trajectory. The first panel is the one that matters: the red line (real yields) is doing the work while the green line (expected inflation) sits flat.

Source: Federal Reserve Bank of St. Louis FRED database (10-year Treasury, inflation-protected and market-implied inflation yields; 2-year to 10-year spread; Federal Reserve Board broad dollar index; ICE BofA US High Yield Index spread; Cboe VIX; US Department of Labor initial jobless claims). Month-end values September 2025 to August 2026; the September point is the latest daily reading, September 17–18, 2026.

Credit

High-yield corporate spreads are 2.68%, one basis point tighter than four weeks ago. Investment grade tightened to 0.77%, its lowest in twelve months. Emerging-market corporate debt tightened to 1.34%, essentially its tightest in a year. Within high yield, the lowest-rated tier (CCC) now yields 10.83% over Treasuries while the highest-rated tier (BB) yields 1.55%. That gap of 9.28 percentage points is the widest of this cycle and the widest in twelve months. Four weeks ago it was 8.79.

A 9.28 point gap between weak and strong borrowers means the market is discriminating based on individual factors and not painting everything with the same brush. Add the policy rate moving up rather than down, and the weakest borrowers face a higher hurdle to borrow than before. Something to keep an eye on.

In the new-issue market, financing remains available to companies with good fundamentals but the terms are getting creative. CoreWeave closed an $8.5 billion facility that is the first investment-grade rated financing secured by graphics processing units, rated A3 by Moody's, and has raised more than $30 billion of debt and equity this year. That is a genuine milestone: rating agencies have accepted rapidly depreciating chips as investment-grade collateral, which sets a template others will use. Some of you are old enough to remember that MBS (mortgage backed securities) became an impolite phrase for a while after 2008. Welcome to the era of CBS (Chip Backed Securities). What could go wrong? As long as everything is copacetic this isn't a problem. But if anything in this circular financed world has a hiccup, remember the wisdom of the Wu-Tang. Protect ya neck.

Commodities

Source: Yahoo Finance (ICE Brent front-month futures; COMEX gold futures); Federal Reserve Bank of St. Louis FRED database (US Energy Information Administration Europe Brent spot price; US Census Bureau manufacturing construction spending; ICE BofA Emerging Markets Corporate Plus Index spread; Federal Reserve Bank of New York and Federal Reserve Bank of Philadelphia manufacturing survey new orders, averaged for the proxy); Statistics Bureau of Japan (Japan inflation). Data as of the dates shown, latest September 21, 2026.

Escalating US and Iranian strikes around the Strait of Hormuz drove the move. US forces destroyed five Iranian tankers, Houthi attacks hit Saudi facilities at Abha, Jazan, Najran and Khamis Mushait, and a drone strike on Saudi Arabia's Petroline coincided with a Libyan valve closure. Front-month futures rose from $92 to $108.75 by 15 September. But the physical market went much further: the European spot assessment reached $130.80 on the same day, roughly $22 above the futures price.

When physical barrels trade far above the futures curve it means buyers need oil now and cannot wait, which is the signature of a genuine supply disruption rather than a speculative move. A futures market pricing $108 while physical cargoes clear at $130 is a market that believes the disruption is real but temporary.

Front-month Brent closed at $100.02 on 21 September, down from the peak, on reports that the United States may meet Iran at the United Nations General Assembly and on satellite data showing Saudi Arabia moved 2.8 million barrels a day through the Strait over six days against just 700,000 a day in August. The rerouting worked better than feared. We would keep this in proportion: Brent was $126 in March and $70 in June. A violent round trip is normal for this market, which is precisely why the bond market declined to reprice inflation for it.

Manufacturing construction spending was $169.8 billion in July, down 21.2% on the year, the sixth consecutive monthly decline. This is the 2022-24 semiconductor fabrication boom rolling off, a separate cycle from AI capital spending.

Part 2

“The second vice is lying, the first is running in debt.”
-Benjamin Franklin

It shouldn't surprise anyone reading an obscure investing blog that a lot of countries are borrowing a lot of money to keep the lights on. Most G10 countries have debt-to-GDP ratios above 100%. But anyone who's carried credit card debt knows that the amount you owe isn't what gets you in trouble. Not being able to make the monthly payments is. Governments have one big advantage: they can print money, so they never have to default. The catch is that more money chasing the same amount of stuff makes each dollar worth less.

This section works through four questions: how the US borrows, who lends to it, why the current path is unsustainable, and what can be done about it.

How the US funds its debt

The government's credit card is bonds. A bond is a contract to pay interest for a set period and then pay back what was borrowed. When bonds mature, the US does what a broke college kid does with a balance transfer: it issues new bonds to pay off the old ones. And like the college kid, it doesn't get to pick its rate. Buyers do, and if they expect prices to rise, they demand more. The Treasury has to balance getting a good rate against making sure every auction sells out.

It manages that by borrowing across a range of maturities. At the end of August the US had $31.8 trillion of debt outstanding:

Source: US Treasury, Monthly Statement of the Public Debt and average interest rates on Treasury securities, August 31, 2026.

A few things to pay attention to.

The bill share is high. Treasury's own advisory committee of dealers and investors has long recommended keeping bills at 15-20% of the debt. We are at 22.8%. Funding short was a reasonable bet while the Fed was cutting: bills were cheaper than long bonds, and money market funds will take almost unlimited amounts of them.

A third of the debt comes due within a year. Bills plus the notes and bonds maturing in the next twelve months add up to $10.8 trillion, 34% of the total. All of it gets refinanced at whatever the best rates are available when it rolls. As cheap debt issued in 2020 and 2021 matures it gets replaced at today's rates.

Whether the debt is growing relative to the economy comes down to three numbers:

1.      The interest rate on the debt: 3.48% on average.

2.      How fast the economy is growing in dollar terms, inflation included: 6.56% over the past year.

3.      The primary deficit, which is the deficit before interest: about $0.95 trillion over eleven months, roughly 3% of GDP.

Today the debt is roughly stable for one reason. The economy is growing faster in dollar terms (6.56%) than the average rate on the debt (3.48%), and that three-point cushion about cancels out a primary deficit of roughly 3% of GDP. The cushion is shrinking from both ends. A third of the debt rolls over within a year, and every cheap 2020-21 bond that matures gets replaced at 4 to 5%. And about half of that 6.56% is inflation, so if the Fed gets inflation back to 2%, the cushion goes with it. Congressional Budget Office projections assume the Fed wins, with inflation back at 2% by 2030. Over the next thirty years they have the average rate on the debt at 4.0% and the economy growing 3.8% a year in dollar terms. The cushion doesn't just shrink, it flips. On those numbers debt goes from 101% of GDP this year to a record 120% by 2036 and 175% by 2056, and the annual interest bill doubles from $1.0 trillion to $2.1 trillion within ten years. Holding debt at today's level would take a primary deficit 1.9 points of GDP smaller. That's roughly $600 billion a year of tax increases or spending cuts at today's size of economy.

And that's the optimistic version. CBO built its forecast last December assuming the Fed would cut this year and the 10-year yield would settle around 4.3%. Instead the Fed hiked, and the 10-year closed at 5.10% on 23 September, its highest since 2007. On 24 September CBO published a scenario where rates end up one point above its baseline: debt reaches 222% of GDP by 2056 instead of 175%. Five years out, the path where the Fed gets inflation back to target leaves debt at about 113% of GDP. The path where it stops early and lets inflation run near 4% leaves it at about 100%.

There are a handful of ways out, and history has tried all of them.

1.      Grow out of it. Real growth that outpaces interest rates for years. The only painless option, and the one nobody can order up. The hope today is an AI productivity boom. Hope is not a plan.

2.      Austerity: tax more, spend less. Close the primary deficit directly. The US did it in the 1990s, cutting the deficit by about 5% of GDP between 1991 and 1998. It's harder now: Social Security, Medicare, other health programs, income support and interest together make up about three-quarters of federal spending. And cutting too fast slows the economy, which shrinks the growth side of the equation.

3.      Default or restructure. Refuse to pay, or change the terms: stretch out maturities, cut coupons. Not realistic for a country that borrows in its own currency, and nobody can force the US there. Softer versions, like "voluntary" swaps into very long, low-coupon bonds, get floated from time to time.

4.      Inflate it away. Let prices rise faster than the interest rate so the debt shrinks in real terms. Two catches. It only works on debt with a locked-in rate: bills reprice within months, so a third of US debt can't be inflated away unless short rates are also held below inflation. And it only works as a surprise. Once lenders expect it, they demand higher rates.

5.      Financial repression. Hold interest rates on government debt below inflation using rules rather than free markets: require banks, pensions and insurers to hold government bonds, cap rates, limit the alternatives. The economists Carmen Reinhart and Belen Sbrancia define it as regulation that keeps rates below what a free market would set and creates a "captive audience" of buyers. It is how the US worked down its Second World War debt. From 1945 to 1980 the real interest rate on US government debt was negative in half of all years, saving the government an estimated 1 to 2% of GDP a year. Nobody defaults. Savers just slowly lose purchasing power.

6.      Print it. The central bank creates money and buys the debt, so the government doesn't have to find willing buyers at all. That solves the auction problem. At scale, it becomes an inflation and currency problem.

So where are we now?

Source: Federal Reserve (FOMC statements and implementation notes, including the September 16, 2026 decision); Federal Reserve Bank of New York (System Open Market Account holdings and Treasury bill purchases); US Treasury (Monthly Statement of the Public Debt, August 31, 2026, and buyback announcements); GENIUS Act, signed July 2025; press reports on the yen intervention. The match to each option is Bergenn Financial Group's assessment.

The Fed's bill purchases are printing in spirit, if not in name. Since the end of 2025 the Fed has been buying Treasury bills outright, which it describes as keeping bank reserves "ample". Its bill holdings have gone from $195 billion to $548 billion, partly by swapping out of mortgage bonds and partly with newly created money, and its overall balance sheet has grown by about $220 billion. Its recent operations have bought $112.6 billion of bills and no longer-dated bonds at all. It has also stopped shrinking: maturing Treasuries are rolled into new ones at auction. Whatever the stated purpose, the practical result is a central bank creating money to buy government debt. Two limits keep this short of the full version of option 6. The Fed is buying bills, not long bonds: its holdings of 20-year-plus bonds are down $30.7 billion over the year, to $490.7 billion. And the scale is modest next to a $31.8 trillion market. It is printing at the short end, not capping the long end.

Why would it make sense to fund your debt through the short end? Two reasons. Either you think long end rates are going to come down and you'll be able to lock in debt for the long term at good rates, or that the short end rates will stay low. So what are "good" rates? In this case, they are rates that are lower than nominal growth and/or low relative to inflation. If it gets these rates the debt is quietly eroded every year without anyone voting for it. Ray Dalio lists a Treasury shortening the maturity of its borrowing, because long-term demand is thin, as one of the late-cycle signs of a debt problem. With the Fed hiking, the same choice leaves the government as exposed to the hike as it could be.

Buybacks do what the Fed's bond buying used to do. Treasury's long-bond buybacks, now up to $6 billion per operation from $2 billion, take older 10- to 30-year bonds out of the market and pay for them with cash or newly issued bills. Taking long-dated bonds off the market to hold down long-term yields is precisely what the Fed's quantitative easing did, and what its 2011 "Operation Twist" (sell short-term bonds, buy long-term ones) did without creating any new money. When Treasury pays from its cash account, the money lands in the banking system, which adds a small dose of printing. Buybacks don't increase or reduce the debt. They swap long debt for short, which adds to the exposure described in the first section. So far they are small, and demand to sell into them is shallower than expected.

Stablecoins!? Under the GENIUS Act, signed in July 2025, dollar stablecoins have to be backed one for one by cash, short-term Treasury bills and similar assets, and issuers may not pay interest to the people holding the tokens. The practical result: every new stablecoin dollar is a legally required purchase of short-term government debt, and the holder earns nothing while the issuer collects the bill rate. That is repression by Reinhart and Sbrancia's definition, a rule that creates a captive audience. In 2015 they speculated that households might be the next audience targeted. A zero-interest digital dollar backed by Treasury bills is one way to get there. It is small next to a $7.25 trillion bill market today, but it grows every time someone chooses to hold one.

The yen intervention manages the buyer base. As covered above, the US joined Japan in defending the yen, reportedly so that Japan wouldn't have to sell Treasuries to pay for it. Using official action rather than price to keep a big creditor from selling is the demand side of repression. It's also given Japan access to FIMA. A Fed facility that let's countries exchange US Treasuries for cash so the don't have to be sold. Limit is $60 Billion

The exception is the rate hike, and it's a big one. Repression and printing only shrink the debt burden if the central bank holds its rate below inflation. On 16 September the Fed did the opposite: it raised rates, unanimously, into an energy shock, with Chair Warsh saying inflation "is too high, and has been for too long." That shuts options 4 and 5 for now. And by making the bill stack more expensive, it leaves the hard options (growth and austerity) or a recession that forces a U-turn.

To sum up. The government is funding more of itself at the short end, where the central bank sets the price. The central bank is creating money to buy that short end. A new law is creating captive buyers for it. The Treasury is buying back the long bonds the market is least keen to hold, and helping its largest foreign creditor avoid having to sell. Call it whatever you like, but the result is a slow but steady building of the infrastructure needed for financial repression and debt monetisation. What's missing is a Fed willing to keep rates below inflation. See the chart below.

Source: Federal Reserve (FOMC statement and Summary of Economic Projections, September 16, 2026); Federal Reserve Bank of New York (System Open Market Account holdings, open market operations and the ACM 10-year term premium); US Treasury (Monthly Statement of the Public Debt, August 31, 2026; buyback operation results, September 10, 2026; daily par real yield curve rates); Treasury Borrowing Advisory Committee guidance on the bill share. Data as of September 2026. Verdicts are Bergenn Financial Group's assessment.

So we have a few data points consistent with financial repression and a lot that argue against it. Awesome. Then why are we talking about this? Because if things keep going the way they are our base case is that, over time, a lot of the levers that are reading "against" right now will slowly flip. And we hold that view simply because of math. So how do we get from here to there?

Let me preface this by saying that the option not included is the US could default on it's debt. Our base case is option 3 below.

1.      Suppression quickly, from a worse starting point. The Fed tightens, something breaks, and the response is larger and faster than it would have been. The tell: the Fed's 20-year-plus holdings rising, or an outright long-bond purchase appearing in its published operations. We think this month made this path less likely in the near term and more violent if it comes.

2.      A bond-market revolt forces fiscal consolidation. Enter the "bond vigilantes". The market wins, yields stay high, fiscal policy adjusts. The tell: consolidation enacted and visible in issuance, not merely announced. This is now the path current price signals most resemble.

3.      Muddle along for awhile and suppression arrives gradually. Nothing breaks. The Fed keeps rates high for a while, inflation drifts down but stalls around 3%, and yields stay elevated. Every month more cheap debt rolls into expensive debt, so the interest bill grows and the pressure on the Fed builds. Eventually the Fed stops tightening with inflation still above target. It isn't responding to a crisis; it's responding to the rising cost of holding. The levers in our table then flip one at a time, each under a technical name, and nobody ever announces financial repression.

This note is for information only. It reflects our views as of the date shown, is based on data believed reliable but not guaranteed, and is not a recommendation to buy or sell any security or a substitute for advice specific to your circumstances. Forecasts and market-implied probabilities are estimates, not facts. Past performance does not indicate future results.