Closing Bell Overtime Sep 18, 2026 18m 5m saved
With Warren Pies, Co-Founder of 3Fourteen Research · John Flood, American Equities Execution Head at Goldman Sachs · Pramol Dhawan, Head of Emerging Markets at PIMCO · Max Levchin, Founder and CEO of Affirm
S&P 500 second-quarter earnings excluding energy grew 14% from a year earlier, John Flood told the program, and Goldman Sachs clients are hedging anyway.
The market spent the week reading the Federal Reserve's rate rise as a threat. Three of the four guests on Friday's Overtime read it as the removal of one, and the fourth spent his segment arguing that the technology everyone is afraid of has so far cost his company no jobs at all.
"Just extreme hedging happening in the market to combat all the various macro headlines that we have to deal with in our client conversations"
Warren Pies downgraded equities on 10 August because he thought the market had underpriced a hike, and now says the hike cleared the way. Flood sees institutional positioning first from Goldman's execution desk. Pramol Dhawan runs emerging-markets investing at PIMCO. Max Levchin has spent 15 years building credit models at Affirm and launched one built on transformers this week.
The full segment is covered here so you can skip it. 18 minutes of audio, 13 minutes of reading.
Here are the 11 calls that matter.
Key Takeaways
Pies cut equities to benchmark weight on 10 August and put them back to overweight after the hike
He reads the Fed's projections as one more hike and a high bar to a third
Goldman's tripwire is the speed of the move: the 10-year at 5.5% before Halloween
Second-quarter S&P 500 earnings excluding energy still grew 14% year over year
Flood's call on software is that it is where semiconductors were 18 months ago, with hedge funds still short
Emerging-market inflation excluding China is now lower than in the United States
Dhawan prefers Latin America and sub-Saharan Africa to the Taiwan and Korea AI trade
A globally diversified bond portfolio gets to 7–7.5%, which he calls insurance against an AI winter
Affirm's new model is built to say yes to more borrowers, not to cut underwriting staff
Transformers read the footnotes on a credit file that a decision tree would discard
1. Back Into Equities
Pies was asked whether the moment had come to get more aggressive on stocks, having spent the past month more cautious. He reversed course on air.
He downgraded equities on 10 August because a hike was underpriced
It was a little over a month ago, August 10th, where we downgraded equities from an overweight to benchmark weight
Warren Pies
His thesis was that as the odds of a September rise climbed, the rates market would reprice and equities would struggle. It repriced and the index absorbed it: the S&P 500 fell a couple of percent, with the weakness in financials and industrials.
The hike was a clearing event, not a threat
I think the Fed cleared the decks. It was a clearing event. Yields have repriced. And it's time to push back into the equity trade.
Warren Pies
2. One Hike, Then A High Bar
Pies reads the Fed as having one more rise in it and a much harder case for anything beyond that.
His base case is one more, then a high bar
I think it's one more hike with a pretty high bar to an additional hike.
Warren Pies
He compared it to 1997, when the Fed raised once and stopped, and said the market will come to read this cycle the same way. Two numbers in the committee's own projections are what give him the high bar. Core PCE is projected at 3.4%, before the revisions that are about to land, which he expects inflation data to come in under.
An inflation forecast the Fed can beat
That's going to be easy for inflation data to come under. The Fed can declare a little victory there.
Warren Pies
The second is a 4.1% unemployment rate in both 2027 and 2028.
The projections assume the job market never softens
It also assumes the labor market won't weaken even a little bit.
Warren Pies
A host pushed back on the certainty, and on the 1997 comparison.
The host's objection: the Fed has not decided this
But to say that they think it's one more I think maybe overstates how much they have decided.
A host
And 1997 only looks like one hike in hindsight
And the 97 example, it was one, but who knows if it was intended to be one at the start because you got the Asian financial crisis hitting not that long after that.
A host
Pies conceded that the data can always change, and said what he liked about the meeting is that for one moment the data is frozen and the committee's reaction function visible.
3. Hyperscalers, Then Semis
Asked where the leadership comes from, Pies named an order.
Hyperscalers first, semiconductors next
I think it's hyperscalers first and then semis right after that.
Warren Pies
He has not upgraded bonds alongside equities, and the curve looks flat to him.
Still underweight bonds, with the 10-year not finished
We're still underweight bonds. I think the two year is at fair value. The ten year might have another 10 or 15 basis points to reprice.
Warren Pies
Which argues for narrow leadership, not a broad rally
So to me it says that narrow mega-cap leadership is going to be the place to be while that process is ongoing.
Warren Pies
Once yields settle, he said, the rally can broaden out. The news he thinks the market missed on the day of the hike was a neocloud provider raising its on-demand rental rates for graphics processors, a price rise rather than a discount.
Everything he sees points at compute demand
And so everything we see is pointing to this crazy demand for compute.
Warren Pies
4. Hedging, Not Deterioration
Flood's observation was that institutional positioning has turned defensive while the underlying numbers have not. He was asked whether the two are connected, and said they are not.
The defensiveness is hedging, not a view on fundamentals
Just extreme hedging happening in the market to combat all the various macro headlines that we have to deal with in our client conversations, oil rates, geopolitics come up constantly.
John Flood
Earnings, by contrast, have been strong
But when we look at earnings, they've been stellar. Q2 earnings season was one of the best we've seen.
John Flood
The objection he hears from clients is that the growth is all artificial intelligence. His answer is the ex-energy number: with crude spiking, second-quarter S&P 500 earnings outside energy still grew 14% from a year earlier.
Which is why he is still bullish
We think this trend is intact. And we're bullish because earnings are very very solid.
John Flood
5. Software Is The New Semis
Flood's one bullish pocket inside the artificial-intelligence complex is software, which he compared to where semiconductors sat 18 months ago.
Positioning in software is light and shorts are covering
And the positioning there is very light hedge funds have started to cover shorts.
John Flood
The two sub-themes are security and data
And specifically within software, we like cybersecurity and data infrastructure.
John Flood
He expects institutions to lean into both
Protect the home front. Organize data are two themes that our institutional investors are going to start leaning into.
John Flood
A host raised a broader feature of this market: paired-up trades that decide what works on a given day, which has helped keep index-level volatility down. Flood said that stays as long as multi-strategy funds dominate trading and factors keep mattering more.
Pair trades and factor moves are here to stay
I think that the pair trades factor movements underneath the surface are here to stay.
John Flood
6. The 5.5% Tripwire
Asked what would break the bull case, Flood put it in yields. The market, he said, is pricing three more rises before the end of next year; Goldman's economists expect one more quarter-point this year and two cuts in 2027.
The market is priced for three more hikes
So the market right now is pricing in three more hikes before year end of next year.
John Flood
His own house expects one more, then cuts
We think we get one more 25 basis point hike this year and then two cuts next year.
John Flood
That gap is why he thinks a hawkish surprise is hard from here. What matters for rotation out of stocks is not the level of yields but the speed of the move.
The number that would worry him is a fast one
So if we see the ten year yield hit 5.5% before Halloween, it's essentially a two standard deviation move
John Flood
At that speed money leaves equities for bonds
That is when you start to see assets move away from equities into bonds. That would be concerning to me.
John Flood
On seasonality, he pointed at a mechanical cause rather than a pattern. September is the biggest fiscal-year-end month for mutual funds, which produces selling of losing positions for tax purposes.
September is the mutual-fund year end
Seasonality for me, where we see it the most is because September is the number one fiscal year end month for the mutual fund communities in terms of year end.
John Flood
So he expects choppy trading to continue
So I would not be surprised to see continued chop through September
John Flood
7. EM Policy Credibility
Dhawan was introduced with the fact that emerging markets are outperforming the S&P 500 this year despite rate and energy volatility, and with his claim that they are winning the policy-credibility race against developed markets. He gave three supports.
He expects the outperformance to keep going
We're really excited about emerging market assets.
Pramol Dhawan
The first is valuation: a discount to developed-market peers in both equities and fixed income, with South Korean and Taiwanese companies generating what he called really good free cash flow. The second is credit.
Credit growth is at a post-crisis high
We're at the highest level of credit impulse across emerging markets that we've seen since the global financial crisis, and that's helping to raise nominal growth.
Pramol Dhawan
The third is monetary policy. Emerging-market central banks raised rates quickly and have been slow to cut, which he said has brought inflation down far enough to invert the usual comparison.
Inflation outside China is now below the US
So much so that inflation in emerging markets, excluding China, is now lower than that in the United States.
Pramol Dhawan
8. Latin America Over Taiwan
Asked which countries look best, and how he handles high commodity prices, Dhawan named a region.
Latin America threads the needle
We really like Latin America.
Pramol Dhawan
His reasoning was high commodity prices, US policy support and governments moving toward more market-friendly positions, set against cheap valuations in Brazil and Colombia and a strong year in Argentina. Sub-Saharan Africa runs a similar playbook, he said, except that its policy support comes from China.
He calls it durable rather than tactical
I think Latin America is just a very durable story.
Pramol Dhawan
A host put the implication to him: preferring those regions means preferring them to Taiwan and South Korea, which have effectively been an artificial-intelligence trade. Dhawan said both will keep benefiting as capital spending widens beyond the United States.
Data centers are moving offshore for political reasons
We're seeing AI data centers move beyond the US borders because of the political issues that we have here.
Pramol Dhawan
And they are landing in three regions
We're seeing data centers being built out across East Asia, across Latin America, in Kazakhstan as well.
Pramol Dhawan
His problem with Taiwan and Korea is not the cash flow, which he called very strong, or the current-account surpluses, which he called great. It is that owning them is owning one position.
One trade in many countries is not diversification
But it is an AI story. It's hard to sort of get away from the fact that there's very little diversification across the equity markets.
Pramol Dhawan
So the diversification has to come from bonds
We almost think to ourselves, if you want diversification, you need to do that in fixed income.
Pramol Dhawan
9. Insurance For An AI Winter
On US yields, Dhawan would not call the current level on 10-year Treasuries correct, and said it could rise further.
He will not defend the current level
Could it go a little bit higher? Sure it can.
Pramol Dhawan
What interests him is what a globally diversified bond portfolio yields once sensible credit spread is added on top.
A diversified bond book gets to 7–7.5%
You can sort of bump into that seven, 7.5% type yield. And that is a really positive yield for us.
Pramol Dhawan
At that yield the margin for error is wide
A lot of things have to go wrong for you to sort of lose money in that state of the world.
Pramol Dhawan
The scenario he wants it for is the one nobody is positioned for. The whole market is arguing about inflation, and he said the paths to the opposite outcome are numerous.
The bubble popping is a deflation path
Everyone's thinking about inflation right now, but there are lots of paths to deflation, lots of paths to an AI bubble popping
Pramol Dhawan
High-quality fixed income, on his account, is the insurance that gets needed in that state of the world.
10. Affirm's AI Underwriter
Levchin came on to discuss a new underwriting model that reads a borrower's credit history and decides what that borrower can responsibly repay. The question put to him was whether it means fewer human underwriters.
The point of the model is approving more people
It's fundamentally about saying yes to more people, giving more folks access to credit
Max Levchin
It finds lending it would previously have declined
So we are finding efficient ways of lending money to people where we otherwise might have said no.
Max Levchin
And it is the output of 15 years of model building
It is a product of 15 years of doing research and building AI and machine learning models.
Max Levchin
Levchin called himself a huge techno optimist on the wider argument about the technology, while agreeing that safety work, monitoring and evaluation all matter.
He expects scientific breakthroughs within quarters
But by and large, we are probably quarters away from foundational breakthroughs in all kinds of sciences.
Max Levchin
Then the line the segment was titled for. Affirm's position since the transformer paper, he said, has been to make the people it already has more productive.
The headcount has not paid for the compute
We've never had to let go of a person to buy more tokens.
Max Levchin
And the compute has been cheap enough to use heavily
We've been able to, quite economically, put a lot of tokens to work to build some amazing things.
Max Levchin
11. What Makes It AI
A host asked the simple version: if Affirm has run decision-making algorithms on creditworthiness for 15 years, what is new? Classical machine learning, Levchin said, takes buckets of information with a limited number of dimensions, a human decides which are worth looking at, and the result is a large decision tree ending in a yes or a no.
The underlying technology is not new
They've been around since the 70s, but the current rebirth of that technology has fundamentally allowed us to broaden our ability to look at data far, far, far beyond what a human can look at.
Max Levchin
The practical gain, he said, comes from combining large multidimensional embeddings with the simpler decision-tree models. His example was the free text on a credit report: a borrower writing in to say a payment was late because a change of address delayed the bill.
An old model throws the explanation away
Traditional underwriting model would never look at it. It would just discard it.
Max Levchin
A transformer-based model, he said, reads the note, treats it as a mitigating circumstance and changes the answer. On whether the job goes away:
Human underwriters stay, but they cannot do this
Not really. But human underwriters could never do what we're doing with transformers.
Max Levchin
Bonus Insights
The hosts also walked through what they were watching into the close: Friday's large options expiration rolling off, and the pattern of September and October weakness. One noted that emerging markets closed lower for the week after the rate rise, which frames Dhawan's case rather than contradicts it, since the outperformance he claims is year to date.
All four guests read the rate rise as clearing the way rather than closing it, and Levchin's addition was that the models everyone is arguing about have so far cost his company no jobs.
Products, Companies & Tools Mentioned
3Fourteen Research (Pies's firm, which cut equities to benchmark weight on 10 August and restored the overweight after the hike)
Goldman Sachs (Flood's desk sees institutional positioning; its economists expect one more quarter-point rise this year and two cuts next)
PIMCO (Dhawan's firm, which wants emerging-market equity risk in Latin America and its diversification in fixed income)
Affirm (Launched an underwriting model built on transformers, which Levchin says approves borrowers the old model turned down)
Books & Resources Mentioned
Attention Is All You Need (The transformer paper Levchin dates Affirm's current position on artificial intelligence to)
If this was worth your time, send it to someone who has to have a view on this.
Get the latest market chatter as it happens: