1Nothing in the long bond's rise is inflation — it is all real rates, and AI borrowing is part of it
Nick Colas, co-founder of DataTrek Research, decomposed the 30-year Treasury yield on The Compound's What Did We Learn and the split is stark. Thirty-year inflation expectations have sat between roughly 1.5% and 2.5% since 2010 and have not moved. Real rates did everything. They have gone from around 2% to 2.5% to pushing 3%, and that is the whole of the nominal move. The 30-year printed 5.27% on 21 August, off a 5.31% high on 17 August, which Colas describes as a 15-to-20-year peak in yield.
He gives four reasons real rates broke out. The Fed's bond buying artificially depressed them and that unwind is now complete. There has been no recession since 2020 despite two oil shocks, a trade-policy shock and 2022's rate-hiking campaign, which implies the neutral rate is simply higher than it used to be. Federal deficits keep growing, with the year-to-date figure already above all of 2025. And the fourth is new: about $1.75tn of corporate bond issuance this year, running 20% to 30% ahead of the same point in 2025, much of it from companies funding AI.
That fourth point is the one that connects to equities. Colas's argument is that every dollar buying a high-grade corporate bond is a dollar not buying a Treasury. Asked which he would rather hold, a ten-year or a US ten-year, he says a bond investor only wants coupons and principal back, and Alphabet is as likely to deliver those as the government is. and are named alongside it as issuers. So the demand that used to clear the Treasury's supply is being bid away by the same capital cycle that is lifting the equity index.
The consequence he draws is a duration call, not a stock call: keep portfolio duration under five years until the US economy weakens. His evidence is , the long-dated Treasury ETF, which compounded at roughly 7.8% a year through the 2010s with coupons reinvested and at negative 4.4% a year so far this decade. The asset that was held as the risk-off hedge has been the thing destroying returns.
Today's tape ran the other way, and the reason matters. rose 0.91% and the seven-to-ten-year part of the curve rose 0.45%, while crude fell 3.14% to $82.34. A one-day fall in oil pulls inflation expectations down and bonds up; it does nothing to the real-rate story, which is about deficits, the neutral rate and who else is issuing. A cheaper barrel is not a fix for this. Only a weaker economy is, which is precisely the trade-off Colas is describing.
2A September rate rise is priced at one-in-three, and a cut at almost nothing
Polymarket's "Fed Decision in September" contract, resolving 16 September, carries $52.0m of volume and $2.85m of liquidity — the deepest book checked today. No change trades at 66.5%, unchanged on the day. A 25 basis point increase trades at 33.5%, up one point. The two cut legs together are worth 1.65%, so the hike is priced at roughly twenty times the combined chance of a cut.
The drift is the story rather than the level. That hike leg was 30.5% a week ago and 32.5% yesterday, and has now risen on three consecutive readings. This is the same conclusion Colas reaches from the yield decomposition, arrived at independently: an economy that has absorbed every shock thrown at it since 2020 without a recession is an economy whose neutral rate is higher than the Fed's current 3.63% funds rate suggests.
The hard data underneath is thinner than the pricing implies. Unemployment printed 4.1% for July, down from 4.2% in June and 4.3% in May, while payrolls fell by about 23,000 over the same month — a labour market that is loosening on one measure and tightening on the other. No new monthly macro release landed inside today's window. On Kalshi, which would give a second venue for the same question, the fetch returned HTTP 429 for the third consecutive day.
The asymmetry is what a reader should take from it. With a cut priced at under two cents, there is far more room for a dovish surprise to move the market than a hawkish one, because the hawkish case is already most of the way in the price.
3The Hormuz headlines de-escalated faster than the Hormuz prices
President Trump posted at 14:30 UTC that the US Navy has removed or detonated all mines from international waters in the Strait of Hormuz, and warned Iran that any vessel laying new ones will be destroyed. Reuters reported the same day that the US has begun returning staff to Middle East embassies after the war evacuations, and that fresh sanctions landed on Iran and its trading partners, including the oil trader Wellbred. Crude read all of it as relief and fell 3.14% to $82.34, extending a slide Reuters put at more than 4% the day before, when the sanctions themselves were "shrugged off".
The prediction markets did not follow. The contract on Hormuz traffic returning to normal by 15 September trades at 1.85%, up 0.3 points, on $539k of volume — effectively ruled out. The 31 December version trades at 36.5%, up one point, on $9.4m of volume and $250k in the past day. Neither price moved materially on a demining announcement, which is the interesting part: clearing mines is not the constraint the market thinks is binding.
The ceasefire ladder splits by tenor. Continuation through 31 August firmed to 94.5%, up three points, and through 15 September to 83.5%, up two. But 30 September softened to 71.5%, down three, and 31 October sits at 65.0%. Near-term calm got more likely today and the medium term got less so.
Reuters also reported that six months into the war, almost half of global oil flows now transit war zones, and that Iran has threatened 45 tankers with fines and confiscation. So what these prices describe is not the end of the disruption but its normalisation — a longer, quieter conflict that shipping routes around rather than through. That distinction matters for the second scenario in the next section, which needs oil back near $60-65 to work.
4The whole of this year's S&P gain is earnings, and both routes to paying more get tested tomorrow
Colas's second point is the one with the most direct read on equities. The S&P 500 is up about 12.1% year to date, and earnings revisions are up 15% for this year and 13% for next. There has been no multiple expansion at all this year — slightly the reverse. He notes that after thirty years watching the data, consensus estimates start a year high and get trimmed through it; a year of upward revisions is the rare case, and this one came from technology and energy.
He lays out three upside paths for the next twelve months and the conditions each needs. The base case, up 6% to 16%, needs only continued earnings growth with the multiple flat around 20. The better case, up 13% to 28%, needs a resolution to the US-Iran conflict, crude back to $60-65 and diesel down from above $100 to $70-80, which would let the multiple reach 22. The best case, up 12% to 39%, needs all of that plus AI capital spending showing up in reported earnings, taking the multiple to 24. He is explicit that the middle case was his base until about six weeks ago and that the length of the conflict is what demoted it.
Two things have changed about what the index is. Sector and single-stock correlations sit at ten-year lows, which Colas reads as a market confident enough about the absence of a recession to pick and choose; the hyperscalers' willingness to lever up to fund AI is the same bet expressed in balance sheets. And the mega-caps have stopped being buyback machines. They now issue stock and debt, which is why is both the worst-performing mega-cap and the one whose debt rating has come closest to junk. Roughly 38% of the MSCI All Country World Index is technology; the equal-weighted S&P is 14%.
Both of the multiple-expansion routes get marked tomorrow. reports after the close on 26 August, with consensus at $2.13 of EPS on $93.6bn of revenue, and it rose 1.84% today into the print. The same morning brings the Personal Income and Outlays release, which carries the PCE inflation measure the Fed uses, alongside the GDP release. The AI-earnings question and the inflation question land within hours of each other.
5The long-run return came from dividends and from what happens after the pick
Two videos land on the same uncomfortable point from opposite ends. Don Hamson, of Plato Investment Management, told Equity Mates that over the twenty years to the end of 2025, two-thirds of the Australian equity market's total return came from dividends including franking credits — index capital growth was only 3.5% a year. Over more than a century, he puts the dividend share above half. Yet the index yield is now near an all-time low, at roughly 3.2% cash plus 1% franking, against a franking yield closer to 1.5% five or six years ago.
Ian Cassel, a micro-cap investor promoting his book Stock Picker on Excess Returns, supplies the other half. About 10% of active managers beat the S&P 500 over ten years, and about 2.5% to 2.7% over twenty. His framing is that stock picking is five skills — identifying, analysing, buying, selling and holding — and most practitioners have a process for the first two only. Buying is sizing conviction; selling is limiting losses and capturing gains; holding is maintenance work on a thesis you already own. His own screen is a business that grows through a recession, a balance sheet that lets it act while competitors cannot, and a valuation that can double in three years without any multiple expansion.
Hamson's warning is the sharpest single-name moment in the day. , G8 Education on the ASX, screens on a trailing yield near 30%, which he calls driving by the rearview mirror: the company cut its dividend to nothing in February and the share price has collapsed since. He flags consumer discretionary as the sector where trap probability is currently highest, and notes that is "no longer a high yield bank". He is also, by his own description, running income funds — the episode's thesis and his book are the same thing, which is context for how firmly it is put.
The undirected mentions in that conversation are worth recording. now earns more than half its profit from copper and has just raised its dividend by more than 40%, which he says has led some to reframe both as electrification plays, a label he distances himself from while agreeing on the copper shortage. he would have called Australia's greatest earner five years ago and now describes as having lost its shine after a takeover that has not worked.
What the sources recommended
Three videos produced two explicit calls, both on the sell side, and no buy recommendations at all. Neither is a fresh single-name idea in the usual sense, and both carry a qualification from the person making it.
Colas's is a duration instruction before it is a ticker call. Keep portfolio duration under five years until the US economy weakens, with the long-dated Treasury ETF cited as the instrument whose −4.4% annualised return this decade demonstrates the cost. The caveat travels with it: he says the time to extend duration is when the data weakens and the risk of a shock becoming an event rises, and that there is no sign of that now. It is a call with a stated expiry condition rather than a permanent view.
Hamson's is a screening warning rather than a position. G8 Education's trailing yield near 30% reflects a payout made about eleven months ago; the dividend was cut to nothing in February and the share price has fallen since. He generalises it to consumer discretionary as the sector where he currently sees the highest probability of the same trap.
Two names sat close to the buy line and are recorded outside the tables. Hamson says his fund has made money in over many years and always likes it, which describes a holding rather than an entry. Colas concedes a point to equal-weighted S&P exposure he has long argued against — 14% technology against roughly 38% in the global cap-weighted index — but names no instrument, so no symbol was inferred into a table.
Both the inflation answer and the AI answer arrive on the same day
The 26 August session is where three of the claims above get tested at once. Personal Income and Outlays lands in the morning with the PCE measure inside it, and GDP with it. If PCE runs hot, the 33.5% hike leg has its justification and the real-rate story gains a fifth reason. reports after the close, and it is the closest thing available to a read on whether the AI capital spending in claim one is producing the earnings that claim four's best case requires. and report the same afternoon; , , and follow on 27 August. The next employment report is not until 4 September, so the labour-market ambiguity behind the Fed pricing stays unresolved for another ten days.
