Abandon Ship!

(DESCRIPTION)

Title card: JP Morgan, Eye on the market. JP Morgan. Presentation. Title card: May 2026, Abandon ship! Image: A digital illustration of a large ocean liner resembling the Titanic, with the side of the hull labeled 2026 midterms. An American flag flies at the bow of the ship, and three large smokestacks emit dark smoke against a cloudy sky. Multiple anthropomorphic elephants in business suits leap from the deck into the choppy ocean below, with one elephant prominently splashing into the water in the foreground. A small lifeboat floats in the water to the right. A video box on the right. Michael Cembalest has short hair, glasses, and wears a dark zip-up jacket. He sits in front of a virtual modern shelf that holds vases, gold decorative items, books, and geometric objects.

(SPEECH)

Good morning. This is Michael Cembalest with the May 2026 Eye in the Market podcast. This one's called "Abandon Ship" which I will explain shortly. As you can see here, there's an image of a bunch of well-dressed elephants jumping off the SS 2026 Midterm ship into the ocean, which is the theme part of this discussion.

I've

(DESCRIPTION)

Slide: Transportation modes. A man with light brown hair and a beard wears a t-shirt, dark jeans, and a backpack as he rides a motorized white suitcase scooter through an airport corridor. The background has large windows and a blurred motion effect. Bold text reads IT'S SO FAST!

(SPEECH)

been traveling a lot recently. I was in Scottsdale and someplace called the Salamander Hotel in DC. And then I was in Bozeman and then Miami and then Laguna Beach-- I still don't where exactly that is-- and then Los Angeles.

And so I was thinking of getting this motorized suitcase. But then I decided that I would look completely ridiculous on it. So I decided against it. And I'm continuing to walk through airport journeys.

(DESCRIPTION)

Slide: Abandon ship. A bullet point list.

(SPEECH)

In any case, this Eye on the Market is kind of a compilation of a lot of the different topics that I've been talking about to clients recently-- the midterm elections, relevant barometers, and then all the issues around Congressional redistricting

The spring thaw in US economic conditions, US earnings and economic resilience, and the issue of energy intensity, a rationing timeline for the Strait of Hormuz-- I'll explain what I mean by that.

An update on hyperscaler earnings, tech valuations, debt financing, some issues around data center build-out issues, and then an idea called the Gulf Super Express, which will probably never happen, but is worth reviewing just in case it does.

So before we get started, a couple of weeks ago, we published a piece on Mythos, given its exceptional ability to wreak havoc by identifying software vulnerabilities, and then the race between patching and hacking.

Last week, JP Morgan published a really interesting, well-researched note on this cybersecurity vulnerability of operational technology, such as aircraft autopilot systems, factory robots, power grid controls, railway track shifting systems, things like that. It's a very sobering read, but I think it's worth the effort.

OK, so what is abandon ship all about?

(DESCRIPTION)

Slide: Midterms, Abandon ship. A scatter plot titled GOP House retirements vs party gains/losses, 1930 - 2024 with a subtitle reading Change in GOP House seats in subsequent election. The vertical axis ranges from -120 to 100 and the horizontal axis is labeled Number of GOP retirements, ranging from 0 to 36. Red data points scatter across the chart, with the plot divided into three vertical zones labeled Coast is Clear on the left, Mixed Bag in the middle, and Get Out of Dodge on the right. Most points cluster between 4 and 20 retirements, with seat changes ranging mostly between -40 and 60. A red dashed vertical line at 36 retirements is annotated with Note: 36 GOP retirements before 2026 midterms. Source: Brookings, US House of Reps, Ballotpedia, JPMAM, 2026.

(SPEECH)

So you probably have heard that so far, 36 house Republicans have decided not to run for reelection this fall. It's the highest number on record since the data begins in 1930s, so almost 100 years.

And so can we infer anything from this? Well, when the number of GOP retirements is low-- let's say less than 10-- usually there's not much going on in the next election. When the retirements are between 10 and 20, it's a mixed bag. Sometimes Republicans lose seats, sometimes they gain seats.

But when more than 20 Republicans retire from the House, it has historically been a signal that they are about to get routed in the midterm elections. And so we have a chart here that shows that trifecta of different conditions in terms of GOP retirements and what tends to happen in the subsequent elections. So you can take a look at that.

But some of these dots imply pretty large losses for the GOP in the midterms. We'll see.

(DESCRIPTION)

Slide: decline in blue-collar employment. A stacked bar chart titled Year on year change in US blue collar employment with a vertical axis labeled Thousands of jobs ranging from -400 to 1,000 and a horizontal axis spanning 2023 to 2026. The bars represent Utilities, Transportation & warehousing, Natural resources & mining, Construction, and Manufacturing, with a red Total line overlaid. The Total line starts near 900 in early 2023, declines steadily through 2024 to around 100, then drops below zero in 2025 and reaches roughly -250 by 2026. Source: BLS, JPMAM, March 2026.

(SPEECH)

Some of the challenges the GOP will be facing, of course, is an outright decline in blue collar employment, which I think would be very disappointing to the administration since that was a priority of theirs.

So blue collar employment in terms of utilities, transportation, natural resources, mining, extraction, construction, manufacturing, these things are absolute declines in employment since Trump was inaugurated.

Everybody

(DESCRIPTION)

Slide: Soaring commodity prices. A bar chart titled Absolute YTD US commodity price changes with a vertical axis labeled % price change ranging from -30% to 120%. Bars descend from highest to lowest across commodities including Naphtha, Wholesale gasoline, Shipping fuel, Jet fuel, Sulfur, Methanol, Propylene, WTI (crude), Urea, Ethylene, Benzene, Xylenes, Retail diesel, Toluene, Retail gasoline, Fertilizer, Sulfuric acid, Aluminum, and Natural gas. Naphtha tops the chart above 120% while Natural gas sits at the bottom near -20%. Source: Bloomberg, JPMAM, April 30, 2026.

(SPEECH)

knows there's been huge increases year-to-date in terms of commodity prices. Natural gas has gone down. But the rest of the entire suite of oil-related commodities and refined products, fertilizers, propylene, methanol, sulfur, jet fuel, shipping, fuel, fertilizer, you name it have gone up quite substantially, anywhere from 40% to 120%.

And

(DESCRIPTION)

Slide: Rising inflation expectations. A line graph titled US short term inflation expectation surveys of households and consumers, Percent. The vertical axis ranges from 0% to 8% and the horizontal axis spans 2019 to 2026. Three lines track Conference Board 1 yr households, University of Michigan 1 yr households, and New York Fed 1 yr consumers. All three lines remain near 2% to 3% from 2019 to early 2021, then rise sharply to peaks between 6% and 7% in 2022. The lines decline through 2023 and 2024 to around 3% to 4%, then spike upward again in 2025 before fluctuating between 3% and 5% through 2026. Source: Conference Board, NY Fed Survey, UMich Survey, Bloomberg, JPMAM, April 2026.

(SPEECH)

inflation expectations are unsurprisingly rising. They were declining coming into the end of last year. The war has reversed some of that. So now you're starting to see inflation expectations pick up.

And there are some other things going on that voters may also focus on. There

(DESCRIPTION)

Slide: Soaring ACA premiums. A bar chart titled ACA Marketplace average premium changes with a vertical axis ranging from -10% to 120% and a horizontal axis spanning 2019 to 2026. Teal bars represent yearly premium changes labeled 1.5% in 2019, -3% in 2020, 1.1% in 2021, 3.5% in 2022, 3.5% in 2023, 6% in 2024, and 7% in 2025. A red bar for 2026 reaches 26%, and a taller red bar also labeled 2026 reaches 114%. An annotation with a red arrow reads Assuming expiration of Enhanced Premium Tax Credit received by 90%+ of ACA enrollees. Source: CMS, KFF, RWJF, JPMAM, 2026.

(SPEECH)

are soaring ACA premiums. So from 2019 to 2025, the ACA marketplace for health care insurance had annual premium increases that were at most 7%, usually closer to 4% to 5%.

In 2026, they're going to go up 26%, the premiums, on an absolute basis. And when you couple that with the loss of the enhanced premium tax credit, it's going to result in an effective 114% annual increase in premiums for the ACA enrollees. So those are some pretty eye-popping numbers.

As

(DESCRIPTION)

Slide: A measles epidemic. A bar chart titled Measles cases reported in the US, 2024-2026 with a vertical axis labeled Count ranging from 0 to 300 and a horizontal axis spanning Jan-24 to early 2026. Bars remain low near 0 through most of 2024, then begin to rise sharply in early 2025. An arrow points to the start of this rise with the annotation RFK Jr. sworn in as Secretary of HHS. Cases climb through 2025 with peaks around 200, then surge to nearly 300 by early 2026 before declining to around 50. Source: CDC, JPMAM, April 19, 2026.

(SPEECH)

you all know, as soon as RFK jr. Gets sworn in as Secretary of Health and Human Services, we have the outbreak of a measles epidemic. I'm sure that's just a coincidence. And

(DESCRIPTION)

Slide: Strange doings at the DOJ. An area chart titled Cases terminated by the Department of Justice with a subtitle reading Number of criminal cases terminated, monthly. The vertical axis ranges from 0 to 12,000 and the horizontal axis spans 2004 to 2024. The shaded blue area fluctuates between roughly 1,000 and 5,000 cases for most of the period, with frequent spikes. A sharp peak near the right side of the chart reaches close to 10,000, labeled Trump 2.0, Bondi DoJ. Source: Ken Morales (ProPublica), March 31, 2026.

(SPEECH)

then there are some very strange doings at the Department of Justice that are starting to get surfaced as well, that I thought would be interesting to talk about.

We have a chart in here that looks at, going back to 2004, the number of cases terminated each year by the Department of Justice. And sometimes the Department of Justice terminates cases because they go stale. Sometimes they terminate them because their priorities are shifting compared to prior administrations.

But there was a by far-- by more than a factor of two-- the largest spike in case terminations by the Department of Justice under the last year or so. And the weird thing is that when you look at it by category, the administration reduced the terminated immigration cases, because it was picking up prosecutions of immigration.

But they increased the pace of case terminations-- labor racketeering, national security, terrorism, organized crime, white collar crime, drugs, corruption, civil rights, you name it. Every single other category listed showed an increase in declined cases.

300

(DESCRIPTION)

A table titled Referred cases that the DOJ declines to prosecute with a subtitle reading First six months of each administration. The table has four columns labeled Case type, Avg of prior 3 administrations, Trump (2025), and Change in declined cases. Rows list Labor at 28, 64, and 129%, National security at 720, 1,391, and 93%, Organized crime at 98, 182, and 86%, White-collar crime at 3,787, 6,009, and 59%, Drugs at 3,447, 4,999, and 45%, Corruption at 471, 623, and 32%, Violent crime at 4,858, 6,108, and 26%, Civil rights at 457, 535, and 17%, Other at 2,786, 2,950, and 6%, and Immigration at 864, 674, and -22%. Source: Ken Morales (ProPublica), March 31, 2026.

(SPEECH)

cases involving material support to foreign terrorist organizations, 60 union corruption and labor racketeering cases, 5,000 cases of money laundering and federal drug law violations. So some very strange doings at the Department of Justice that are also kind of swirling around the midterm elections.

Now, the Republican Party has tried to change some of the math through redistricting. And this is very much of a process. And we have a chart in here that keeps tabs on what's happened so far.

So

(DESCRIPTION)

Slide: Redistricting, GOP advantage could range from -1 to +9 seats. A stacked bar chart titled Number of new House seats that may result from state redistricting efforts and Voting Rights Act decision. The vertical axis ranges from 0 to 16. The Republicans bar totals 14 seats and includes Texas with 5, N. Carolina with 1, Ohio with 2, Missouri with 1, Florida with 4, Louisiana with 1, and Tennessee with 1. The Democrats bar totals 10 seats and includes California with 5, Utah with 1, and Virginia with 4. A legend identifies three categories: Possible consequence of Louisiana vs Callais VRA Supreme Court decision, Redistricting: Subject to State Supreme Court challenges, and Redistricting: State legislation enacted with no further challenges expected. Source: JPMAM, 2026.

(SPEECH)

the GOP-- and remember, all of these redrawn districts don't automatically become either Republican or Democrat. The midterm results-- the people vote in the midterms. But the issue is the way that those new districts are drawn presumably will result in the assumed gain for the parties that redraw them.

So Texas picks up five. North Carolina picks up one. Ohio picks up two. Missouri may pick up one, subject to some state Supreme Court challenges. Then on the Democratic side, California picks up five. Utah picks up one.

The reasons we explain in the piece. I don't want to go into it now. And then you had the Virginia referendum, which could pick up four seats.

DeSantis has recently had a new map proposed that the state legislature approved last week. But Florida has an explicit prohibition on partisan gerrymandering. So that's going to go straight to the Florida State Supreme Court.

And then in Tennessee and Louisiana, there might be one seat each that flips to the GOP based on the decision of the Supreme Court last week related to the Voting Rights Act, where they decided that using race as a purpose for creating these so-called majority/minority districts is unconstitutional.

And so Tennessee and Louisiana each may rush to redo their primaries in redrawn districts. We'll see what happens. The bottom line from all of this is that in the worst case, the GOP probably loses a seat on that, and in the best case, could pick up as much as nine. And then the swing is somewhere in there.

I did want to talk for a couple minutes about Virginia, because there may be a really embarrassing unforced error that Democrats made in Virginia as it relates to the referendum. Let me just spend a minute on that because I think it's interesting.

So

(DESCRIPTION)

Slide: depending on what happens with the Virginia referendum.

(SPEECH)

Virginia State Constitution is similar to California's, and it includes protections against partisan gerrymandering. And so in Virginia, the normal rule is that you need a bipartisan commission to propose new maps. And they can only meet once a decade.

So just like California, Virginia held a referendum so voters could say, yeah, we want to keep these provisions, but we want to temporarily suspend them until the end of the decade so that the legislature, instead of a bipartisan commission, can redraw the maps, and can do it in a very partisan way.

So the new Congressional map in Virginia would replace the six Democrat, five Republican balance with something like 10 to 1. But on the referendum ballot, unlike in California, Virginia added the phrase that they were doing this to restore fairness.

And they put that language actually on the ballot. And so that's raising issues. Fairness to whom? Like other redistricting efforts all across the country by both parties, these measures don't advance fairness in that state, because they create legislative balances that are much more extreme than the voter population balances by party.

And there are legal questions regarding the use of that fairness language on the referendum itself that the Virginia Supreme Court may now evaluate. And that's one of the reasons that a state trial court enjoined state officials from certifying or preventing state officials from certifying the results.

We'll see what happens. Maybe they just let it go through. They can require them to hold it again. But this does seem like something of an unforced error by the state of Virginia to include that kind of language on the ballot itself.

OK, so let's go to the next topic. What's

(DESCRIPTION)

Slide: U.S. economy more resilient than expected. A line graph titled Dallas Fed US Weekly Economic Index with a subtitle reading Index (composed of 10 series including unemployment claims, retail sales, fuel sales and electricity utility output). The vertical axis ranges from 0 to 8 and the horizontal axis spans 2018 to 2026. The line hovers around 2 to 3 from 2018 to early 2020, then drops sharply to near 0 in 2020. It surges to a peak above 7 in 2021, then declines steadily through 2022 and 2023 to about 1.5. From 2023 onward, the line gradually rises and stabilizes between 2 and 3 through 2026. Source: Dallas Fed, Bloomberg, JPMAM, April 25, 2026.

(SPEECH)

a little surprising about so many House Republicans bailing is that the economic data-- not necessarily the labor market data-- but the economic data is a lot more resilient than expected.

And we have a series of charts at Eye on the Market that we have in our Trump Tracker that shows this. So there's a weekly index that tracks the economy, comes out of the Dallas Fed. Looks pretty good.

(DESCRIPTION)

A line graph titled US economic surveys & business cycle indicators surprise index, Actual data - median analyst forecast, %. The vertical axis ranges from -2.0% to 2.0% and the horizontal axis spans 2018 to 2026. The line fluctuates between about -1.5% and 1.5% throughout the period, dipping below -1% during 2020 and again in 2023, then rising sharply toward 1% by 2026. An annotation reads Includes Phil Fed Bus Outlook, Empire Mfg, UofM Sentiment, ISM Services & Mfg, MNI Chicago PMI, Conf Board Cons Confidence & Leading Index, Dal Fed Mfg, Rich Fed Mfg. Source: Bloomberg, JPMAM, April 30, 2026.

(SPEECH)

Surveys of business cycle indicators and things, like the Empire Manufacturing Survey, regional surveys, Conference Board consumer confidence, Dallas Fed, things like that. That looks pretty good.

(DESCRIPTION)

A line graph titled ISM new orders less inventories with a subtitle reading ISM manufacturing index. The vertical axis ranges from -10 to 20 and the horizontal axis spans 2018 to 2026. The line starts near 18 in 2018 and declines to around 0 by 2019. It drops sharply to near -10 in 2020, then surges to a peak above 15 in 2021. From 2022 onward, the line fluctuates between -5 and 5, before rising to nearly 10 by 2026. Source: ISM, Bloomberg, JPMAM, March 2026.

(SPEECH)

My favorite indicator is this thing called manufacturing orders less inventories. It measures the pace at which manufacturing orders for new equipment are outstripping inventory growth. And that looks pretty good.

(DESCRIPTION)

A line graph titled US trucking market demand with a subtitle reading Indexed ratio of load postings divided by available truck postings. The vertical axis ranges from 0 to 450 and the horizontal axis spans 2018 to 2026. A blue line representing Flatbed trucks fluctuates between 50 and 200 from 2018 to 2020, then surges to a peak above 400 in 2021. The line drops sharply through 2022 and 2023 to around 50, then rises again to nearly 250 by 2026. A yellow line representing Refrigerated trucks remains lower overall, peaking near 250 in 2021 and 2022, then declining to around 100 by 2024 before rising to about 150 by 2026. Source: Bloomberg, JPMAM, April 24, 2026.

(SPEECH)

A measure of flatbed trucking demand is picking up, for whatever that's worth. Some people like that. Loan

(DESCRIPTION)

A line graph titled Loan demand surveys with a subtitle reading % of respondents reporting an increase net of % reporting a decrease. The vertical axis ranges from -90% to 90% and the horizontal axis spans 2017 to 2026. A blue line labeled Dallas Fed US Banking Outlook Total Loan Demand Index fluctuates between 0% and 60% for most of the period, with a sharp drop near 2020 and a peak above 60% by 2026. A yellow line labeled Senior Loan Officer Opinion Survey US Demand for C&I Loans hovers near 0% from 2017 to 2019, drops below -60% in 2020, and stays negative through 2024 before rising to around 20% by 2026. Source: Dallas Fed, Federal Reserve, Bloomberg, JPMAM, March 2026.

(SPEECH)

demand at the banks is rising. And so there's a lot more people-- banks reporting an increase in loan demand.

(DESCRIPTION)

A line graph titled US durable capital goods ex aircraft with a subtitle reading US$, billions, seasonally adjusted. The vertical axis ranges from $55 to $85 and the horizontal axis spans 2014 to 2026. A yellow line representing New orders and a blue line representing Shipments track closely together, starting near $67 in 2014, dipping to about $60 by 2016, then climbing to roughly $75 by 2019. Both lines drop sharply to about $55 in 2020, then rise steadily through 2022 and 2023 to around $77. From 2024 onward, both lines climb further, with New orders reaching above $82 and Shipments near $80 by 2026. Source: Census Bureau, JPMAM, March 2026.

(SPEECH)

And then just basic durable goods orders and shipments are also rising. So the US economy is actually proving to be pretty resilient in the face of this war. Now, along with that resilience has come some inflation measures which are no longer falling.

And

(DESCRIPTION)

Slide: June eye on the market. A two-panel layout with a bulleted list at the top reading New Fed chair set to take over, Higher oil prices set to boost producer prices, Rising inflation surprise index, Prices paid, Core PCE rising again. The left panel is a line graph titled ISM surveys: prices paid with a subtitle reading Index (50+ = increasing), 3 month moving avg. The vertical axis ranges from 40 to 90 and the horizontal axis spans 2017 to 2026. A yellow line for Manufacturing and a blue line for Services both peak above 80 in 2021, drop to around 50 by 2023, then rise back near 70 by 2026. A dashed horizontal line marks the 50 level. Source: ISM, Bloomberg, JPMAM, March 2026. The right panel is a line graph titled PCE inflation measures with a subtitle reading y/y % change. The vertical axis ranges from 0% to 8% and the horizontal axis spans 2018 to 2026. Lines representing Median PCE, PCE, Core PCE, and Trimmed PCE all hover around 2% from 2018 to 2020, then surge to peaks between 5% and 7% in 2022. The lines decline through 2023 and 2024 to around 2% to 3%, then trend upward again toward 4% by 2026. Source: Federal Reserve Bank of Dallas, JPMAM, March 2026.

(SPEECH)

so Eye on the Market in June to celebrate the new Fed Chair-- good luck to him. What's that phrase? That some don't wish for something. You may get it.

The new Fed Chair is going to be stepping in right around the time that higher oil prices are set to boost producer prices. There's an inflation surprise index in the US that's going up.

And the new Fed Chair is also going to have to deal with rising prices paid data and then core PCE is rising again. Now, I saw some recent commentary where the new Fed Chair says, you know, we should probably use trimmed PCE to be tracking inflation.

That's the one inflation measure that happens to be going down right now. The problem is, that was an abysmal measure if you were trying to use it to track the inflation surge that took place under Biden, which presumably all these Trump people are still criticizing.

And that would have been an abysmal way to try and track the Biden inflation era. So interesting that they have cherry picked the one area that is currently showing lower inflation. We'll see how that works out. Anyway, that's for the June Eye on the Market.

And

(DESCRIPTION)

Slide: S&P 500 earnings more resilient than expected. A line graph titled S&P 500 change in forward earnings vs price with a subtitle reading Percent change from January 1, 2025. The vertical axis ranges from -15% to 25% and the horizontal axis spans Jan-2025 to Apr-2026. A yellow line representing 12 month forward EPS rises gradually from 0% to about 25% by April 2026. A blue line representing Price drops sharply to around -15% by April 2025, then climbs steadily through the rest of 2025 and 2026, reaching above 20% by April 2026. Source: Bloomberg, JPMAM, April 30, 2026.

(SPEECH)

it's not just the US economy that's been proving resilient. It's also S&P earnings. And that's really, I think, the best explanation for why you've had this V-shaped recovery in the equity markets recently.

Now, while the commodity shock from the war is still playing through, there's other forces upsetting it-- lower tariffs, particularly after the Supreme Court decision, individual tax cuts, corporate tax cuts, all of those things are providing stimulus of roughly 2% of GDP this year.

As for Q1 earnings, so far, around half of the companies have reported. Sales growth of 10%, earnings growth around 20% to 25%, positive news on surprises, and only literally a handful companies have revised their 2026 earnings guidance down so far in spite of the war.

So this looks pretty resilient in terms of an earnings picture. Now, of course, one of the reasons for that is the primary driver are technology-related issues, which we'll talk about in a few minutes.

But

(DESCRIPTION)

Slide: U.S. energy intensity. A line graph titled Oil intensity of the US economy with a subtitle reading Index (100 = 1990). The vertical axis ranges from 0 to 140 and the horizontal axis spans 1990 to 2025. A blue line labeled Oil intensity of S&P 500 profits starts near 100, rises to about 125 in the early 1990s, then declines with notable spikes around 2000 and 2010, ending near 25 by 2025. A red line labeled Oil intensity of GDP starts at 100 and declines steadily to about 50 by 2025. A yellow line labeled Oil intensity of NIPA profits also begins at 100 and trends downward to about 15 by 2025, with peaks around 2000 and 2008. Source: Energy Institute, S&P, BEA, JPMAM, 2025.

(SPEECH)

I did want to show this one chart, because I think it's really important. And we've been talking about this ever since the war started in our energy paper in March.

Gasoline prices still matter. The US is a commuter society. We don't have very good public transit. A lot of people drive. They drive SUVs and other cars primarily that have pretty crappy gas mileage. We don't have a national large gasoline tax, et cetera, et cetera.

That all said, the oil intensity of the United States is either half or 20% of what it was when I started working at JP Morgan almost 40 years ago. And we have a chart in here that looks at the oil intensity of GDP, which is down by half since then.

And better, the oil intensity of S&P and economy-wide profits is down anywhere from 75% to 80%. And so I think that's an important way of understanding why the economy and the stock market might be more resilient to oil shocks than it would have been 10 years ago, 15 years ago or 20 years ago.

And as semiconductor consumption continues to rise and as energy efficiency and fuel switching continue, I would expect these oil intensity numbers to continue to fall in the years ahead. So this is one way of understanding why profits and growth may be more resilient to these oil shocks in the United States than you might have thought.

Now,

(DESCRIPTION)

Slide: Strait of Hormuz timeline. A line graph titled Global oil and refined product inventories with a subtitle reading Million barrels. The vertical axis ranges from 7,500 to 9,100 and the horizontal axis spans Jan to Dec. Multiple lines track inventory levels for years 2017 through 2026. The 2020 line stands out with the highest values, peaking near 9,000 in mid-year. Most other years cluster between 7,900 and 8,500. The 2026 line, marked with red squares, starts near 8,400 in January and declines sharply to about 7,650 by June. Source: JP Morgan Global Commodities Research, April 2026.

(SPEECH)

that said, the coast is not entirely clear, because what's happening is the world is facing the prospect of declining global oil inventories. And you can't draw them down past a certain level.

So we have a chart in here that looks at global oil and refined product inventories that have ranged somewhere in the neighborhood of 8 to 8.5 billion barrels over the last few years.

And global oil and refined product inventories refers to oil that's sitting around in storage tanks, in terminals, in pipelines, in floating tankers, or in strategic reserves. Now they're being drawn down by about 8 million barrels a day.

And by June or July sometime, you may hit an operational stress point. And for those of you that work in the oil and gas industry or the pipeline industry, you'll understand, you can't let the pressure in oil and gas pipelines fall below a certain level, or else you kind of lose the ability to have it function properly.

And so by June or July, we're going to start hitting globally some operational stress points, below which I think it would be difficult for inventories to continue to be drawn down. And if that's the case, if the Strait of Hormuz hasn't reopened by then, you're going to start seeing more fuel rationing, primarily in Asia and Europe.

But that's going to have some economic aftershock effects for us as well. Now, if the strait's not open by September, then you're definitely going to hit some kind of floor more a little bit below 7 billion barrels, where you really start to see a lot of demand destruction.

So putting a rough timetable on it, where we're sitting right now, the administration has about a month and a half to try to get the strait reopen before you start seeing more severe economic consequences, from the best we can tell.

Now,

(DESCRIPTION)

Slide: AI trade is alive and well. A table titled Contribution of 42 AI stocks to S&P 500 returns, earnings and capex/R&D growth. The columns are labeled Since Nov 2022, Since Nov 2023, Since Nov 2024, and Since Nov 2025. Rows list Price return at 76%, 68%, 81%, and 42%, Earnings growth at 87%, 105%, 99%, and 193%, and Capex/R&D growth at 75%, 83%, 88%, and 84%. Source: Bloomberg, JPMAM, April 30, 2026.

(SPEECH)

you remember a few minutes ago, I talked about how the oil intensity of GDP has gone down-- and of profits. That's because the AI intensity of GDP and profits has gone up.

And we have this table in here that we had in the beginning of the year. Since ChatGPT was launched at the end of 2022, there's about 42 AI-related stocks, including some of the turbine manufacturers and other companies that feed off of the AI ecosystem.

These 42 companies in the S&P 500 have accounted for 75% to 85% of the price returns, earnings growth, CapEx, and R&D growth in the entire S&P 500. So this is an unbelievably concentrated bet that's taking place in the US.

The latest earnings from the hyperscalers and things like that suggest that corporate adoption of AI and agentic AIs is generating some revenue growth. The productivity numbers are generally moving in the right direction.

So the AI trade is very much alive and well.

(DESCRIPTION)

Slide: While broad tech sector has repriced. A line graph titled S&P 500 tech forward PE drawdown with a subtitle reading Percent drawdown from previous high. The vertical axis ranges from -45% to 0% and the horizontal axis spans 2016 to 2026. The line fluctuates near 0% to -10% from 2016 to 2018, then drops to about -25% labeled Rate hikes. It dips to around -30% labeled COVID in 2020, then plunges to nearly -40% labeled 2022 bear market. Another sharp drop in 2025 is labeled Liberation Day, followed by the lowest point near -45% in 2026 labeled Software/Iran. Source: Bloomberg, JPMAM, April 30, 2026.

(SPEECH)

What's interesting is, as we've been talking about, the damage that the agentic AI programs like Claude and OpenAI have crushed some of the software stocks to the point where there was a huge drawdown in broader technology PEs of almost 40% at its worst levels. The PEs are still down 30%.

And this is the amazing thing. You're not going to believe this chart until you see it. But if we look at all these different sectors and regional markets, US technology stocks have the lowest ratio of price-to-earnings divided by earnings growth.

So

(DESCRIPTION)

Slide: Leaving tech as one of the cheapest industries/markets on a P/E vs earnings growth basis. A bar chart titled The price investors pay for earnings growth with a subtitle reading P/E ratio divided by projected 2-year consensus earnings growth. The vertical axis ranges from 0.0 to 4.0. Yellow bars descend from highest to lowest across categories including Staples, MSCI Japan, Industrials, Financials, Healthcare, Discretionary, Utilities, Stoxx 600, Comm services, S&P ex mag 7, S&P 500, Mag 7, MSCI ACWI, Nasdaq 100, Materials, and Energy. A teal bar at the far right represents Tech. Staples tops the chart near 3.8 while Tech sits at the bottom near 0.5. Source: Bloomberg, JPMAM, April 30, 2026.

(SPEECH)

this is what's called a PEG ratio. And it has to do with what's the PE divided by earnings growth-- what's the price you pay for earnings growth. And tech at this point looks to be the cheapest on this entire chart.

Now, some of these stocks are now cheap for a reason because some of the SaaS vendor stocks are going to find it very difficult to recover. But as we've written about before, we think there's been too much selling and that there's going to be more resilience in that sector that's been priced in so far.

But I thought it was interesting to look at just how cheap the tech sector has become on using this particular lens. The latest projections from the hyperscalers is 2026.

The next 12 months or so is the last huge surge in capital spending in R&D before it starts to level out in 2027. But

(DESCRIPTION)

Slide: Hyperscaler apex and R&D. A line graph titled Hyperscaler capex and R&D as a share of revenues with a vertical axis labeled Percent ranging from 0% to 90% and a horizontal axis spanning 2017 to 2027. Lines track Meta, Alphabet, Amazon, Microsoft, S&P 500 Info Tech median, and S&P 500 median. Meta peaks above 80% around 2026, while Alphabet and Microsoft rise to about 50% to 60% by 2027. Amazon climbs more gradually to around 30% by 2027. The S&P 500 Info Tech median stays near 20% throughout, and the S&P 500 median remains around 10%. A vertical dashed line near 2026 marks the start of Consensus estimates. Source: Bloomberg, JPMAM, April 30, 2026.

(SPEECH)

just to be clear, these companies are spending 40%, 50%, and 60% of their revenues on capital spending.

I think that's gargantuan. That compares to the average tech stock in the S&P 500, which spends 18% of its revenues on capital spending in R&D. So that's just kind of amazing.

And because of those soaring expenditures in terms of capital spending, the hyperscaler free cash flow margins are starting to come down.

(DESCRIPTION)

Slide: Hyperscaler cash flows. A line graph titled Hyperscaler free cash flow margins (net of capex) with a subtitle reading Percent, trailing 12 months. The vertical axis ranges from -10% to 40% and the horizontal axis spans 2021 to 2027. Lines track Microsoft, Meta, Alphabet, Amazon, and S&P 500 median. Microsoft and Meta hover between 25% and 35% from 2021 through 2025, then decline sharply toward 5% by 2027. Alphabet stays near 20% to 30% before dropping to about 5% by 2027. Amazon fluctuates between -5% and 10% throughout the period. The S&P 500 median remains relatively flat near 10% to 13%. A vertical dashed line near 2026 marks the start of Consensus estimates. Source: Bloomberg, JPMAM, April 30, 2026.

(SPEECH)

I think the markets will have some patience here. But at some point, these projected free cash flow margins are going to have to pick back up again.

(DESCRIPTION)

Slide: Hyperscaler debt financing. A bar chart titled Annual change in hyperscaler long term debt (bonds, loans and leases) with a subtitle reading US$, billions. The vertical axis ranges from $0 to $150 and the horizontal axis spans 2015 to 2026. Bars from 2015 to 2024 fluctuate between roughly $10 and $50. The 2025 bar reaches about $150 and is segmented into Q1-Q3 plus Amazon Q4, Meta Q4, Microsoft Q4, Oracle Q4, and Google Q4. The 2026 bar reaches about $115 and is segmented into Amazon Q1, Oracle Q1, and Google Q1. Source: Bloomberg, company sources, JPMAM, Q1 2026.

(SPEECH)

I wanted to spend a minute on something because we're getting a ton of questions on this. And it's an example of something where I think that this issue is less of a problem than it's often made out to be.

So we're among the first people to show this chart, which is, yes, there's an enormous amount of hyperscaler debt financing going on. Until the fourth quarter of 2025, you didn't really see Google and Microsoft and Meta and Amazon and Salesforce issuing lots of debt.

And then all of a sudden in the fourth quarter of last year and in the first quarter of this year, we're starting to see literally hundreds of billions of dollars of hyperscaler debt issuance to finance data centers.

But

(DESCRIPTION)

Slide: AI stock debt to cash flow ratio. A bar chart titled Net debt to trailing 12m EBITDA of direct AI stocks with a subtitle reading Multiple, includes bonds, loans and SPV triple net leases. The vertical axis ranges from 0x to 8x. Blue bars represent As of Q3 2025 and yellow dots represent As of Q1 2026 across companies including Palantir, Palo Alto, Arista, Tesla, ServiceNow, Fortinet, AMD, NVIDIA, Google, Salesforce, Cadence, Apple, Adobe, Meta, Microsoft, Micron, Amazon, Qualcomm, Uber, Super Micro, Intel, Broadcom, Dell, NXP, IBM, Oracle, HP, ARM, and TSMC. Most companies sit near or below 1x, with Oracle reaching about 4x in Q3 2025 and surging to nearly 7x by Q1 2026. A note marks the S&P 500 median around 3x. Source: Bloomberg, JPMAM, April 30, 2026.

(SPEECH)

if we use a lens of looking at the amount of debt you have relative to your cash flow-- and in debt, we're including bonds and loans and triple net leases and all that kind of off-balance-sheet stuff too-- Oracle is still the outlier.

Oracle is still the company that has a much higher ratio of debt to cash flow than even the median, the S&P. Whereas the rest of them, even Meta and Microsoft and Apple and Google, these numbers are pretty low.

And so even with all the-- in other words, even with all the borrowing that has taken place, their ratios of debt to cash flow are still pretty low because they're so profitable and they have a lot of cash and marketable securities on their balance sheet to offset that.

(DESCRIPTION)

Slide: The big picture. A line graph titled Capex financing vs capex cycle with a subtitle reading Russell 3000 universe: US tech and communications companies. The left vertical axis ranges from 0% to 45% and the right vertical axis ranges from 3% to 15%. The horizontal axis spans 1996 to 2026. A yellow line representing Share of capex + dividends that are financed with debt rather than by internally generated cash flow from ops peaks above 40% around 2001, declines through the 2000s, and rises sharply near 2026 to about 28%. A blue line representing Capex to sales fluctuates between 5% and 15%, peaking around 2001, declining through 2010, and trending upward again to about 5% by 2026. Source: Bloomberg, JPMAM, December 2025.

(SPEECH)

And we've shown this chart before, but the biggest difference between the Dot-com boom and today is in both of those periods, you had a spike in capital spending. The difference was last time, capital spending got financed with debt. And this time, it's still mostly being financed with internally-generated cash flow.

And

(DESCRIPTION)

Slide: Spreads well-behaved other than Oracle. A line graph titled Credit default swap comparison, select AI stocks with a subtitle reading Spread, five year CDS, bps. The vertical axis ranges from 0 to 200 and the horizontal axis spans 2022 to 2026. Lines track Oracle, Meta, Amazon, Alphabet, and Microsoft. Most companies hover between 20 and 60 bps for much of the period. Oracle stands out with a sharp surge starting in late 2025, rising above 180 bps by 2026. Meta also climbs to about 70 bps near the end of the period, while Amazon, Alphabet, and Microsoft remain relatively flat between 20 and 50 bps. Source: Bloomberg, JPMAM, April 29, 2026.

(SPEECH)

that's the reason why credit spreads-- other than Oracle, which is a special case-- that's the reason why credit spreads for the other hyperscalers are still pretty tight in trading and investment grade levels.

(DESCRIPTION)

Slide: Will the data center boom continue at the same pace? A line graph titled US construction spending with a subtitle reading US$, billions, seasonally adjusted annual rate. The vertical axis ranges from $0 to $125 and the horizontal axis spans 2014 to 2026. A red line representing Electric power starts near $75 in 2014, fluctuates through 2020, then rises steadily to about $120 by 2026. A blue line representing Office buildings starts near $35 and climbs to about $75 by 2020, then stays mostly flat through 2026. A yellow line representing Data centers starts near $5, rises gradually through 2020, then surges sharply from 2022 onward to about $50 by 2026. Source: Census Bureau, JPMAM, January 2026.

(SPEECH)

OK. Just a couple more things on AI and data centers. The latest data from the Census Bureau as of January shows that electric power generation in construction spending and data center construction spending are continuing to go up and up and up.

The question is, can it really continue at that pace? And are there any things that suggest that we may see a partial slowdown? And I think we are starting to see signs of a possible slowdown.

And

(DESCRIPTION)

Slide: Signs of a possible slowdown in the pace. A horizontal stacked bar chart titled US data center capacity by status & completion year with a vertical axis labeled GW. The horizontal axis ranges from 0 to 50. Bars represent the years 2025, 2026, and 2027. Each bar is segmented into Accelerated, On time, Delayed, No construction observed, and Not monitored. The 2025 bar reaches about 12 GW, the 2026 bar reaches about 22 GW, and the 2027 bar reaches about 47 GW with a large dark blue segment representing No construction observed. Source: SynMax Vulcan Platform, IRR Energy, April 9, 2026.

(SPEECH)

there are some people that did some satellite and other measurements of all the data center capacity that's supposed to be worked on in 2027. And way more than 50% of it, there's no construction observed. Now, sometimes it's permitting issues that are temporary.

But other times, it's because they can't get the necessary combustion turbines, they can't get the enough skilled labor, they can't get the transformers to connect it to the grid. And so some of these equipment, labor, and permitting issues are beginning to get in the way of the pace of the data center construction buildout.

And

(DESCRIPTION)

Slide: equipment. A scatter plot titled Producer price inflation: core goods with a subtitle reading % increase vs 2020 for each of the 47 core goods categories. The vertical axis ranges from 0% to 90%. Blue dots descend from about 60% on the left to near 0% on the right. A red dot at the top, near 85%, is labeled Transformers & Power Regulators. Source: Bloomberg, BLS, JPMAM, March 31, 2026.

(SPEECH)

as a reminder, of all the 47 categories in the Producer Price Report, the second highest level of inflation that we've seen is in transformers and power regulators. So this data center build out has really created some supply chain bottlenecks that may now start slowing the pace of data center expansion.

(DESCRIPTION)

Slide: Transformers. A line graph titled Transformer delivery times in the US with a vertical axis labeled Weeks ranging from 70 to 170 and a horizontal axis spanning Q1 2022 to Q2 2025. A blue line representing Generation step-up transformers starts near 50 weeks in early 2022, rises steadily to about 130 weeks by 2023, dips slightly, then climbs sharply to about 165 weeks by Q4 2024, with a final dot near 150 in Q2 2025. A yellow line representing Power transformers starts near 80 weeks, rises to about 130 weeks in 2023, declines slightly, then peaks near 140 weeks before dropping to about 110 weeks by Q4 2024, with a final dot near 130 in Q2 2025. Source: Wood Mackenzie, Bloomberg, Q2 2025.

(SPEECH)

And just since 2022-- and so in 3 and 1/2 years-- the delivery time to get one of the generation step-up transformers that's needed to connect these things to the grid has risen from about a year to three years.

So that's not even for a turbine. That's just for the transformers. Now, these kinds of problems can be solved with additional investment in productive capacity. Question is, who's going to do that? And right now, we're not seeing the supply chains increase that rapidly.

OK, last topic.

(DESCRIPTION)

Slide: Gulf Super Express: Just a pipe dream? A geographic map of the Middle East and Arabian Peninsula region. Labeled countries include Syria, Iraq, Iran, Lebanon, Israel, Cyprus, Jordan, Saudi Arabia, Kuwait, Qatar, United Arab Emirates, Oman, Yemen, Eritrea, Ethiopia, and Djibouti. Bodies of water labeled include the Arabian Sea, Gulf of Aden, Gulf of Oman, and Indian Ocean. Red dotted lines mark routes connecting points labeled Basra Position intake, Strait of Hormuz, Sohar Port, and Bab el-Mandeb.

(SPEECH)

I read this interesting paper from Rice University. By the way, when I grew up, people used to call Rice the Harvard of the South. I don't know if they continue to do that. I have no idea if it's accurate. But I like to always assume that it is.

So Rice University has this thing called the Baker Institute for Public Policy. And they issued a report. And they went into some detail on something called the Gulf Super Express.

Now, I don't that this is ever going to happen, because it would require a lot of coordination between the Gulf countries. And you just saw the Emirates pull out of OPEC. So maybe now's not the best time for me to be writing about this.

But the region, if it decides to cooperate, can reduce the Iranian threat by essentially building a pipeline from Basra and Southern Iraq all the way to the Indian Ocean along the Coast of Oman.

And it would bypass the Strait of Hormuz. It would bypass the Bab Al-Mandab Strait where the Houthis are active. It would mean they wouldn't have to go through the Red Sea and the Suez Canal, and they would just be able to go straight to Asia and India.

(DESCRIPTION)

A bullet point list.

(SPEECH)

And some of the numbers are interesting. So they specced out two 56-inch pipelines that could carry 10 million barrels a day each, a spur to a local port, a lot of storage capacity.

Their capital cost estimates would include both passive and active defenses, a strategic long-lead equipment reserve in case one was damaged. And the cost for this entire multi-country project would be about $55 billion, which is around what Saudi Arabia has spent on the NEOM Project so far. I'm not sure what that would show for it.

And it would take let's call it five years to build. Now, if it did cost that much, at its full capacity of 10 million barrels a day, you'd end up with about $55 per barrel per day.

And that would rank at the lower end of real pipeline costs within a universe of all the pipelines that have been built around the world since the year 2000. So this could be done at a cost competitive basis.

Now obviously, shipping oil first to a pipeline and then through a VLCC tanker would cost more than just shipping it through the tanker. But if you can't ship it through the tanker because Iran's going to close the Strait, then it's a moot point, isn't it?

So anyway, I thought this was interesting. And there are options for the region to try to reduce their exposure to Iran and its influence over the Gulf. And I think chances are better than 50/50 that either this project or something like it at some point starts to get built in the region.

So

(DESCRIPTION)

Title card: May 2026, Abandon ship! Image: A digital illustration of a large ocean liner resembling the Titanic, with the side of the hull labeled 2026 midterms. An American flag flies at the bow of the ship, and three large smokestacks emit dark smoke against a cloudy sky. Multiple anthropomorphic elephants in business suits leap from the deck into the choppy ocean below, with one elephant prominently splashing into the water in the foreground. A small lifeboat floats in the water to the right.

(SPEECH)

anyway, that's enough for today. And thank you for listening. And again, in June, we're going to take a look at the challenges facing the new Fed Chair.

And then in July we'll have a special-- let me get this word right-- semiquincentennial Eye on the Market, looking at the dollar, US equities, and US fixed income at a time of the 250th anniversary of the United States. So thank you for listening, and I'll see you next time. Bye.

(DESCRIPTION)

Logo: J.P. Morgan.