Intro
Claudia Sahm of New Century Advisors previews Kevin Warsh's first Jackson Hole speech as chairman and explains why she has just moved from the hold camp to thinking the Fed should be raising rates. Former Philadelphia Fed President Patrick Harker follows with the case for a signaling hike and why the AI build-out will not feel it, and Gene Munster of Deepwater Asset Management takes apart Nvidia's growth curve, its Hugging Face acquisition and its customer concentration.
Guest: Claudia Sahm, New Century Advisors
Host: Jonathan Ferro
Also on: Patrick Harker, former Philadelphia Fed President; Gene Munster, Deepwater Asset Management
Published: 28 August 2026 on Bloomberg Surveillance
Listen on Omny | 20 min
Key Takeaways
Sahm has moved from holding to wanting hikes, and it was the headlines that did it
"I think we're shifting to a place where it would be most appropriate for them to start raising rates." — Claudia Sahm
"I have been in the hold camp up until about this week, and it's not just on the data." — Claudia Sahm
The risk she now sees is inflation getting stuck nearer 4% than 2%
"They might not get a lot worse, but if we get stuck, we're a lot closer to 4% on PCE inflation than we are 2%" — Claudia Sahm
The Jackson Hole speech is a test of whether Warsh can lead a committee, not a policy signal
She expects no forward guidance and no lead on the September meeting
Harker would raise the funds rate purely as a signal, knowing it will not move 30-year yields
"It's not enough just to say we're going to be tough on inflation. You actually have to do something." — Patrick Harker
Moving the inflation target mid-fight has to come off the table
"We're going on to six years here soon of inflation above target." — Patrick Harker
The supply shocks the textbook says to look through are policy choices now
"They're not so much random acts of God. These are deliberate actions by nation states or within our own administration." — Patrick Harker
The AI build-out has not been rate sensitive, and the productivity payoff arrives long after the spending
Higher rates would not slow the hyperscalers; skittish investors would
"What they're going to be affected by is people who start to question, where's the beef?" — Patrick Harker
Munster expects Nvidia to grow about 90% again next year, well ahead of both guidance and the Street
"the plus 70% revenue growth rate versus the Street at plus 45%" — Gene Munster
The constraint is inference capacity, and Munster still calls it the third inning
Buying Hugging Face is Nvidia doing to its customers what they are doing to it
Meta and Google build their own silicon; Nvidia builds its own models
Non-hyperscalers are what flattens the deceleration investors are afraid of
Hyperscaler revenue growth just over 50% next year against around 85% for everyone else
What Warsh's First Jackson Hole Speech Is a Test Of
The program opened with Sahm's own written note, read on air: "We have a clear sense of what Warsh doesn't like at the Fed, but very little sense of what he's proposing as a replacement. His Jackson Hole speech is an opportunity to bring depth to his vision."
She is treating the speech as a read on the man rather than on policy — she says she is confident she will learn more about Kevin Warsh as a leader and as a Fed chair today, and adds that she may not like what she learns
"We don't know leaders until they have been tested." — Claudia Sahm, who calls the July press conference a test, though not a really tough one
Asked what mistakes Warsh made in that press conference, she declined to grade a new chair out of the gate and reframed the question as how he moves on from the moment
What she is not expecting: "I don't expect us to get forward guidance. We're not going to get a big lead on September 11th." — Claudia Sahm
What she does expect is evidence he is building a consensus and responding to the committee's feedback rather than going it alone
"It's a big transition to go from the outside critic to the leader of the institution." — Claudia Sahm
The Dissenters, and the Majority Still on Hold
The show put to her that this is the most dissents since September 2016, alongside the hawkish speeches since the July conference, and asked whether the chair has to address them
"The chair needs to understand all of the arguments that are on the table." — Claudia Sahm
"The hawks have been very articulate in what they are seeing in the data, what they are concerned about, why they have dissented." — Claudia Sahm, who says that is great
The dissents are speaking for a quieter majority: she says the hawks are giving voice to the hold camp, the majority at the Fed who have been on hold and are clearly looking for more improvement
She is not sure Warsh will be the one to voice that debate today, but says inflation and the inflation outlook will be in the air at Jackson Hole
Asked whether he will lay out the data he watches, she said she would hope for meat on some bone, and pointed at the scale of what he has already proposed: "if you look at the five task forces, they touch almost everything the Fed does"
"consistently, Kevin Warsh has stayed at 30,000 feet" — Sahm thinks the big picture is where he is comfortable, and the likeliest thing to get today
Why Sahm Left the Hold Camp This Week
"I think we're shifting to a place where it would be most appropriate for them to start raising rates." — Claudia Sahm
"I have been in the hold camp up until about this week, and it's not just on the data." — Claudia Sahm
She calls the inflation data itself mildly encouraging, with some improvement showing
The headlines are what changed her mind: "Trade war with Canada, Middle East progress completely stalled out, and a lot of demand for the AI build-out that's pushing up chip prices."
"If I look at the outlook a year out, I am not confident we are moving to 2% on a steady clip" — Claudia Sahm
The Fed's case for holding rests on the forces abating — the tariff effect waning, the gas price effect waning, the AI effect being narrow — and she says the forward-looking information now arriving undercuts it: "am I really confident that those inflationary effects are waning?"
"They might not get a lot worse, but if we get stuck, we're a lot closer to 4% on PCE inflation than we are 2%" — Claudia Sahm
There is a lot of data still to come before September, she says, but the last week or two has not produced the good news headlines the hold case needs
What a Rate Hike Can and Cannot Fix
The show asked what a hike does about geopolitics and high memory prices. "So a Fed rate hike pulls some demand out of the economy" — but she grants it would not go at the root causes of a Middle East supply shock or of tariffs
AI is the exception she draws: some of the build-out runs through capital markets with bond issuance, so interest rates are relevant, and she calls it a demand story rather than a purely supply-side shock
"those higher interest rates, they're going to hit people on the margins, small businesses on the margins" — Sahm on where the demand actually comes out
"They can't stop those tariffs. They can't get the peace in the Middle East. But their job is to get inflation down." — Claudia Sahm
"I'm not talking about anything aggressive." — she frames it as moving a little more restrictive, and says it is likely not a done deal
Harker's Three Elephants, and the One the Fed Controls
Harker's own written note, read on air: "Warsh has three elephants in that room, inflation, the Treasury market, and the fiscal situation underneath both. Watch what he addresses rather than what he promises."
"he clearly doesn't have influence over the long end of the yield curve. That's never been the case" — what the chair does have influence over, Harker says, is inflation and the credibility of the Fed as an inflation fighter
"It's not enough just to say we're going to be tough on inflation. You actually have to do something." — Patrick Harker
On the program's earlier discussion of a quarter or half point move: "raising the short end by 25, 50 basis points. Is it going to make a huge difference? No. But it's signaling what the Fed means when it says it's tough on inflation."
Six Years Above Target, and Why the Goalposts Stay Put
Asked whether this chair faces questions the last one did not, Harker said no — the questions come from the circumstances, not from the switch
"We're going on to six years here soon of inflation above target." — Patrick Harker, who says the Fed has to show it is serious about getting back to 2%
"the idea that we change the target, change the goalposts right now in the middle of this fight, that's just, you have to take that off the table. It makes no sense." — Patrick Harker
The show pushed back with the argument that Warsh has no cleaning up to do, noting that 30-year yields have moved only 10 basis points and that his aim is not to hold the market's hand
Harker's answer was action rather than language: "as people say, words are cheap", and asked what action looks like, "I think they just have to raise the Fed funds rate."
He points at Susan Collins, whom he calls his old colleague, as the tell: "the fact that she is now saying we're probably not restrictive is a real tell"
Supply Shocks That Are Not Acts of God
Asked what has changed in the economy, Harker started with AI: "It's sucking. It's a great sucking sound. It's sucking all kinds of capital in, all kinds of goods, transformers, you name it."
The scale and the speed are what make it different — an incredible build-out compressed into a very rapid time period
The second change is the war, and what it does to the textbook: policymakers are supposed to look through supply shocks, but "these supply shocks keep coming at us. They're off, on, off, on, tariffs on, tariffs off. It's hard to keep looking through those."
"They're not so much random acts of God. These are deliberate actions by nation states or within our own administration." — Patrick Harker
That makes them different in kind from what the Fed has dealt with before, he says, not just different in size
Why Rates Don't Bite the AI Build-Out, and When the Payoff Shows Up
The show asked how rate sensitive the build-out is, a debate it says it has had on the program before. Harker's answer: it has not been to date, and right now it really doesn't matter, because companies are rushing to build the infrastructure
"what we know through history is, you know, that it's going to take time to actually see the productivity on the other side" — Patrick Harker
His analogy for the lag was the re-engineering wave: "we're going back to the 1980s, baby, bell bottoms, disco, and the word reengineering"
"we have to redesign our production processes, our service delivery processes, to take advantage of new technologies. That doesn't happen overnight." — Patrick Harker
"we're not going to see the payoff for quite a while" — it will take time to move into the economy
The show pressed on the transmission mechanism, putting to him that "the new central bank for them in town seems to be Nvidia that's happy to finance everything from the entire ecosystem"
"I don't think that the hyperscalers will be affected by a rate increase. What they're going to be affected by is people who start to question, where's the beef?" — Patrick Harker
"Another phrase from the 1980s. Where's the money? And so investors may get a little skittish at some point." — Patrick Harker
Nvidia's Growth Curve Next to Apple's
The program set the segment up with its own number: Nvidia added $442 billion in market value the previous day, which it called the second-largest one-day gain by any stock in history
Munster took the Apple comparison and ran the arithmetic: "calendar 22, Nvidia did $27 billion in revenue. Calendar 27, so five years later, they'll likely do more than $700 billion."
"what's happened with Nvidia has been breathtaking, even relative to the iconic tech moves that we've seen" — Gene Munster
The eye-opener from the results was the growth rate, not the level: "the plus 70% revenue growth rate versus the Street at plus 45%", for calendar 27
"I expect that next year calendar 27 will probably grow at 90%." — which would leave calendar 27 growing at about the same rate as calendar 26
He says the interesting part is no longer what happened from 22 to 26 but what happens from 27 to 30
"the customers simply just can't get enough capacity to do inference, and that is what is driving these numbers up" — Munster, setting aside the big picture on AI and looking only at the business
"I still believe we're probably in the third inning of this build-out" — he admits it is almost impossible to wrap his head around, and says we're very early. The show's response was that it struggles with that, but that this is the way things are heading
Where the Pricing Power Comes From
Asked about the way Nvidia fosters its own demand by helping build out the AI ecosystem elsewhere, Munster went to the product: it is the speed of the chips that drives all of it
"kind of next year, probably 20, 30% potential pricing increase on some of their products" — the pricing leverage that speed buys
The piece he says gets missed in the conversation: "consumers or their customers really need these chips more than any other custom silicon"
"they basically have the best stuff in town" — Gene Munster
Hugging Face, Poolside and Hedging Against Its Own Customers
The show asked what Nvidia is doing buying Hugging Face, a big platform for open source modeling and discussion, and said the deal was rather lost in the news flow around the results
"it's effectively the GitHub of AI development" — Munster's framing for the technical audience; for everyone else, a library of open source AI models that developers pull together and plug into what they are building
"this is the theme around open source, and this is something that Jensen's been very supportive of" — Gene Munster
The story under the headline is a hedge: "What's most important here is that Nvidia is basically building out, is doing the inverse of what their customers are doing to them."
Customers like Meta and Google are building their own custom silicon because they want alternatives to Nvidia. "What Nvidia is doing here with Hugging Face, what they did last week with Poolside, a small acquisition, is essentially they're building their own models"
The payoff, in his telling, is that if key customers move further into their own silicon, Nvidia has the option to focus on the models and make money from the models themselves
He calls it a fascinating dynamic — as tightly as the mega caps and Nvidia hold their hands, Nvidia is quietly getting some diversification
SpaceX, Sovereigns and the Slope of the Deceleration
Asked about the other customers beyond the hyperscalers, sovereign AI and enterprise, Munster answered with concentration first
"the customer concentration, actually the top two declined in terms of total percentage of revenue. It went from 38% in the April quarter to around 30%." — Gene Munster
"SpaceX basically came out of nowhere. It was a couple percentage of total revenue in the April quarter. It was probably about 5% in the July quarter." — Gene Munster
"the script is flipping in terms of where they're getting their growth from" — for next year he has hyperscalers growing revenue just over 50% and non-hyperscalers, including the sovereign buyers, around 85%
What investors are actually worried about is not size: "it's less about the law of large numbers. Google is a similar-sized business. The issue is the slope of the deceleration of revenue."
"that decline from 90% next year to maybe 50% in calendar 28 won't be as sharp" — Munster's case for why the customer mix matters, and what he says "allows investors to sleep just a little bit better at night"
Sahm's bottom line is that something has shifted in the US inflation picture and the Fed should be moving toward tighter policy rather than waiting, and both Harker and Munster describe an AI build-out that a higher policy rate would barely touch.
Products, Companies & Tools Mentioned
Nvidia (The stock the show says added $442 billion of market value in a day; Munster has revenue going from calendar 22 to more than $700 billion in calendar 27, driven by customers who "simply just can't get enough capacity to do inference")
Hugging Face (Bought by Nvidia — "effectively the GitHub of AI development," and Munster's evidence that Nvidia is building its own models as a hedge)
Poolside (A small acquisition the week before, part of the same push into models)
Meta and Google (The customers building their own custom silicon to have alternatives to Nvidia; Google is also Munster's example of a similar-sized business, to show the issue is deceleration rather than the law of large numbers)
SpaceX ("basically came out of nowhere" as an Nvidia customer, going from a couple percent of revenue in the April quarter to about 5% in the July quarter)
Apple (The comparison Ferro reached for — years of talk about deceleration as it got bigger — and the benchmark Munster says Nvidia's run has beaten)
Get the latest market chatter and takes as they happen:
X | Threads | Instagram | YouTube | TikTok | Facebook