Core PCE Printed 3.0% on a Revision, Not Because Inflation Fell. October Hike Odds Went From 70% to a Coin Flip.

Wednesday, September 30, 2026, written at 9:30 a.m. Eastern, half an hour before the open, on the last session of the third quarter. The 8:30 data is in hand and is discussed below. Futures and pre-market levels are indicative and can move before the bell. Micron reports after the close. Mainland China shuts from tomorrow through October 7 for Golden Week and will be absent for the ISM and for payrolls.

Top Call: The Disinflation the Market Bought at 8:30 Was a Change of Method, Not a Change of Prices

Core PCE for August printed 3.0% year over year against a 3.3% consensus, with the monthly rate at 0.2% against 0.3% expected. Headline held at 3.4% against 3.7% expected. Within ninety minutes the October meeting went from better than 70% priced for a hike to roughly a coin flip, the ten year came off Tuesday’s 5.278% to 5.22%, gold reversed a seven week low, and all four equity contracts turned positive.

Now read the rest of the release. July core was revised down to 3.0% from 3.3%. Second quarter core was revised to 3.3% from 3.6%. July headline was revised to 3.4% from 3.7%. Core PCE did not fall from 3.3% to 3.0% in August. It was 3.0% in July as well, once the Bureau of Economic Analysis applied a methodology change retroactively to 2021 covering how it prices legal services, software and computer accessories, and portfolio management services. The year over year core rate was unchanged on the month. The entire surprise against consensus exists because the consensus was built on a July number that no longer exists.

That distinction is the whole note. A market that repriced the most contested Federal Reserve meeting of the year by more than twenty points of probability did so on a revision to how three service categories are measured, announced in advance, applied to five years of history, and containing no new information whatsoever about what anybody paid for anything in August.

Two things sharpen it further. The same 8:30 release revised second quarter GDP up to 2.2% final against 1.5% expected. Fifteen minutes earlier ADP reported 90,000 private jobs in September against a 68,000 consensus, snapping a run of declining monthly totals after an August of 38,000, with education and health adding 55,000, leisure 22,000, manufacturing 17,000 and construction 15,000. Both of the real activity numbers published this morning beat, and both argue for a hike. Only the inflation number argues against, and it argues against on a definitional basis.

Where we would temper this. A 3.0% core is the level the Federal Reserve now forecasts from, however it was arrived at, and levels genuinely matter for policy even when the change in them is statistical. A committee looking at 3.0% rather than 3.3% has more room, and it is not irrational to price that. Our objection is to the size of the move, not to its direction. Going from 70% to 50% on a redefinition, in the same hour that growth was revised up seven tenths and private payrolls more than doubled, is too much. Note also that John Williams said Tuesday there was no need for urgency in October, and our honest reading is that the New York president probably did more of the work in this repricing than the data did. That is a better reason to be at 50% than the one the tape is citing.

Rates: The Front End Is Doing the Repricing and the Long End Is Not Listening

The thirty year touched 5.61% on Tuesday, its highest since June 2002, in a sixth consecutive session of rising yields. The ten year reached 5.278%, a fresh 2007 high, and has since eased to 5.22% to 5.24%. The two year is near 4.9% and is where the policy repricing is showing up.

The shape of that is the actual trade in this morning’s data. A soft inflation print pulls the front end down because it moves the October meeting. It does nothing to the long end, because the long end is not trading policy. The thirty year has risen for six straight days on term premium, on heavy corporate supply, on a widening deficit and on oil above $100, and none of those four were touched by a revision to how portfolio management services are priced. If this morning holds into the close, today is a steepener, and a steepener driven by a statistical artifact at the short end with the long end pinned at 2002 levels is a poorer quality steepener than it looks.

The dollar is softer after the print, with the euro at 1.1358 and the yen at 156.94, against an index that had gained roughly 2% through September.

Gold: The Framework Gets Its Second Test and Passes in the Other Direction

Gold is up about 1.4%, quoted between roughly $4,185 and $4,241 depending on venue, rebounding from a seven week low. On Monday we retracted a read that framed gold as a bid for something that is nobody’s liability and replaced it with the plainer claim that the carry cost of a non-yielding asset is the dominant term. That framing predicted three consecutive declines into rising real yields, and it got them. This morning the real yield fell and gold rose immediately, with bitcoin up 1.42% to about $85,387 and the Russell contract the strongest of the four.

We want to be careful about how much credit to take. A framework that only works when the variable moves one way is a coincidence. This one has now worked in both directions inside three sessions, which is the first genuine evidence that it is describing the mechanism rather than the outcome. It also continues to explain why gold, bitcoin and small caps keep moving as one expression rather than three.

Energy: A Pipeline Did What Six Weeks of Diplomacy Could Not

Brent is near $103.50 and WTI near $90.25 to $90.41, both up about one percent. Be careful reading the WTI level against yesterday’s page, because the October contract has expired and the front month is now November, so part of the apparent decline from $92 is the roll rather than the market.

The substantive development is physical rather than diplomatic. Saudi Arabia has restored roughly half the capacity of its East-West pipeline, with shipments to the Red Sea running at a minimum of 3.5m barrels a day. That is a route around the strait rather than through it, and it is the best explanation available for why Brent sits at $103 instead of $110 in a week that produced no diplomatic progress at all. President Trump rejected the seven day roadmap and says he has offered nothing. Foreign Minister Araghchi expected a formal American response on Tuesday. Day two hundred and fourteen, and both sides are still talking to mediators rather than to each other.

The read we would add is that the market has begun pricing partial physical workarounds rather than waiting for a settlement, and that meaningfully lowers the ceiling on any future headline. A strait that can be routed around at three and a half million barrels a day is worth less as a threat than one that cannot.

Earnings: Micron Tonight Is the Only Print That Matters

Concentrix is the more interesting of the two and almost nobody will write about it. The company outsources contact centre and customer experience work, which is precisely the labour that agentic systems are being sold to replace. A beat on the quarter with a fourth quarter guide below consensus is the shape you would expect if the current book is fine and the forward book is being repriced by customers who have started running pilots. One quarter from one company is not a trend and we are not calling it one. It is worth watching alongside ADP, because if this cycle displaces anything first it will be this category, and the displacement will show up in guidance before it shows up in payrolls.

Micron reports after the close and it is the test of the capacity thesis this desk has run since Saturday. The company guided to $50.0bn of revenue plus or minus $1.0bn, an adjusted gross margin near 86%, and $31.00 of adjusted earnings per share plus or minus $1.00. The street carries roughly $50.8bn to $51.2bn and about $31.50, above the midpoint on both.

Two things to hold in mind before the tape overreacts tonight. HBM4 shipments passed $1bn ahead of schedule, the HBM customer base has grown to six, and 2026 supply is contracted, which means the quarter is already known and the upside surprise is capped by the same fixed price contracts that give the visibility. And the fiscal fourth quarter runs fourteen weeks against thirteen in the first, so a sequential guide that looks soft is arithmetic before it is demand. If the stock sells off on a guide in the low fifties tonight, that is the calendar being mispriced, and we would say so rather than read it as the memory cycle turning.

Corporate and Sector

TSMC is evaluating a second American manufacturing hub in Texas, with as many as six advanced wafer fabs and Dallas the reported location, on top of the $265bn already committed to Arizona. Because the new fabs would run more advanced nodes, total investment could exceed the Arizona figure. Nothing is finalised and the company has not confirmed it. The point worth extracting is one we have made through the memory work: packaging and memory, not logic lithography, have become the binding constraints on accelerator supply, and a second hub is a statement about where the foundry thinks the bottleneck will be in 2030 rather than about where it is now.

Jensen Huang, speaking alongside Elon Musk on Tuesday, put the AI buildout at roughly one million American jobs, with 10 to 20 gigawatts of new infrastructure annually, and described it as the first re-industrialisation of the United States in fifty years. The number deserves the caveat that it counts infrastructure employment and is not net of displacement anywhere else. Set it beside the Concentrix guide above and the interesting question is not whether the buildout creates jobs, it is whether the jobs it creates and the jobs it removes land on the same people. They do not.

Elsewhere: Hormel is buying Brakebush Brothers for about $1.055bn. Shell took final investment decision to double LNG Canada to 28 million tonnes a year. Apple Pay launched in India with Axis Bank. Robinhood is introducing weekend round-the-clock trading in selected US stocks and ETFs. Roche had fenebrutinib accepted for priority FDA review in multiple sclerosis. Citigroup cut Dow and LyondellBasell to Neutral on demand weakness despite higher oil, and cut Moderna to Sell while raising its target to $80. Germany has attached conditions to UniCredit over Commerzbank, including a domestic listing and job guarantees. Anthropic’s Nasdaq listing is still targeted for October.

Asia and Europe

Tokyo reversed hard, the Nikkei up 1.94% to 66,753 after two consecutive losses, which is the cleanest expression anywhere of a market trading lower American rate expectations through the yen. Seoul was the exception, the Kospi down about 0.4% at 6,843 in a third session of foreign selling, and the pattern there has been consistent for a week: foreign outflow absorbed almost entirely by a domestic bid, producing small index moves on very large flows.

Hong Kong added 89 to 24,613 and Shanghai 11 to 3,842 in the final session before Golden Week. The mainland is now gone until October 8, which removes the second largest equity market in the world for the ISM on Thursday and payrolls on Friday. That is worth positioning around rather than discovering.

Europe is where the more important divergence sits. German import prices rose 8.3% year over year against 8.0% expected and 1.0% on the month against 0.7%, and preliminary September inflation accelerated faster than forecast across Germany, France, Spain and Italy, driven by the energy shock. The bloc print on Friday is expected at 3.6% against 3.2% in August. So American inflation is being revised down by statisticians on the same morning European inflation is accelerating on physical energy, and the gap is passthrough and currency rather than policy. The euro area imports its barrels and has no shale; the United States does not and does. If this holds, the ECB is closer to a hike than the market has it and the rate differential trade that has driven the dollar through September has a problem.

Tuesday’s European closes: the Euro Stoxx 50 up 0.38% at 6,325, the DAX flat at 25,375, the FTSE 100 down 0.45% at 10,637, the CAC 40 down 0.53% at 8,036, the FTSE MIB up 0.09% at 51,805 and the IBEX down 0.42% at 19,518. This morning the DAX is 44 lower at 25,355 and the FTSE 13 higher at 10,650.

The Quarter Closes Today

The third quarter delivered almost nothing at the index level and a great deal underneath it. The S&P was roughly flat in July, up 1.2% in August and flat again through September, against a rise of about 45% in oil and a long end that went from under 5% in early July to 5.61% yesterday. An index that absorbs both of those and finishes the quarter close to where it started is not a calm index, it is a market in which the dispersion went entirely into the constituents. Meta is the proof: up about 30% in September on Muse, its best month since July 2013, inside a flat tape.

Expect month and quarter end rebalancing flows into the close today, which will make this afternoon a poor guide to anything.

Key Events Today and This Week

Friday is the session that settles this morning. If payrolls confirm ADP, the October repricing that happened at 8:30 today reverses, and the market will have spent two days trading a revision to how legal services are priced.

Accountability

We did not flag the methodology revision and we should have. Yesterday we wrote that Wednesday carries the macro constraint at 8:30 and named core PCE as the print. Correct on importance and incomplete on content, because the BEA had pre-announced a change of method applied back to 2021, and that change turned out to be the single most consequential feature of the release. It was public before the print. We had the day to find it and we did not. That is a preparation miss rather than an analytical one, which makes it more annoying and not less.

The gold correction works in both directions on its third session. The carry cost framing replaced a wrong read on Monday, predicted three declines into rising real yields, and this morning predicted the rebound when the real yield fell. We are upgrading this from a correction that is behaving to a framework we will use, with the caveat that three sessions is three sessions.

The capacity thesis gets its real test tonight. Saturday’s framing said capacity is the binding input and that the market is sorting companies by whether they own it, rent it or must fund it. Micron owns it, is sold out at fixed prices into 2026, and reports after the close. We have committed in advance to two readings so we cannot pick one afterwards: a soft-looking sequential guide caused by a thirteen week quarter against a fourteen week quarter is calendar and we will say so, and a guide below $55bn on comparable weeks is the thesis taking real damage.

The crude discipline rule pays for a fifth week. We have carried no directional view on crude since September 23 on the grounds that a strait, not a barrel, is the marginal input. Brent has traded from under $99 to above $108 and back to $103.50 inside two weeks on commentary alone, and the one thing that actually moved the physical picture, half of an East-West pipeline, came from neither side of the negotiation. Having no view remains the correct view.

Monday’s breadth question is unresolved and we are leaving it unresolved. We flagged a contradiction between a reported sector count and the advance decline line and said we would settle it. We do not have clean settled data and we are not going to manufacture a conclusion to close a loop we opened. It stays open.