Future Shock: What comes after the US invasion of Iran and after the rise of Agentic AI

(DESCRIPTION)

Logo: JP Morgan. Disclaimer. Text: PLEASE NOTE: This session is closed to the press. Investing in alternative assets involves higher risks than traditional investments and is suitable only for sophisticated investors. Alternative investments involve greater risks than traditional investments and should not be deemed a complete investment program. They are not tax efficient and an investor should consult with his/her tax advisor prior to investing. Alternative investments have higher fees than traditional investments and they may also be highly leveraged and engage in speculative investment techniques, which can magnify the potential for investment loss or gain. The value of the investment may fall as well as rise and investors may get back less than they invested. The views and strategies described herein may not be suitable for all clients and are subject to investment risks. Certain opinions, estimates, investment strategies and views expressed in this document constitute our judgment based on current market conditions and are subject to change without notice. This material should not be regarded as research or as a J.P. Morgan research report. The information contained herein should not be relied upon in isolation for the purpose of making an investment decision. More complete information is available, including product profiles, which discuss risks, benefits, liquidity and other matters of interest. For more information on any of the investment ideas and products illustrated herein, please contact your J.P. Morgan representative. Investors may get back less than they invested, and past performance is not a reliable indicator of future results. This information is provided for informational purposes only. We believe the information contained in this video to be reliable; however we do not represent or warrant its accuracy, reliability or completeness, or accept any liability for any loss or damage arising out of the use of any information in this video. The views expressed herein are those of the speakers and may differ from those of other J.P. Morgan employees and are subject to change without notice. Nothing in this video is intended to constitute a representation that any product or strategy is suitable for you. Nothing in this document shall be regarded as an offer, solicitation, recommendation or advice, whether financial, accounting, legal, tax or other, given by J.P. Morgan and/or its officers or employees to you. You should consult your independent professional advisors concerning accounting, legal or tax matters. Contact your J.P. Morgan representative for additional information and guidance concerning your personal investment goals. INVESTMENT AND INSURANCE PRODUCTS: NOT A DEPOSIT, NOT FDIC INSURED, NOT INSURED BY ANY FEDERAL GOVERNMENT AGENCY, NO BANK GUARANTEE, MAY LOSE VALUE.

(SPEECH)

This session is close to the press. Welcome to the JP Morgan webcast. This is intended for informational purposes only. Opinions expressed herein are those of the speakers and may differ from those of other JP Morgan employees and affiliates. Historical information and outlooks are not guarantees of future results.

Any views and strategies described may not be appropriate for all participants and should not be intended as personal investment, financial, or other advice. As a reminder, investment products are not FDIC insured, do not have bank guarantee, and they may lose value. The webcast may now begin.

[MUSIC PLAYING]

(DESCRIPTION)

A shimmering strip of gold-plated handwriting swirls elegantly across a dark surface. It spells JP Morgan. Presentation. Title card: Future Shock: What comes after the US invasion of Iran and the rise of Agentic AI. Michael Cembalest, Chairman of Market and Investment Strategy, J.P. Morgan Asset and Wealth Management March 2026. A middle aged man with short brown hair and glasses wears a dark zip up sweater over a white shirt. He sits at a table and speaks, and a glass of water rests in front of him. Text: Michael Cembalest, Chairman of Market and Investment Strategy. In the background, there is a city skyline through a window.

(SPEECH)

OK. Good morning, everybody. A lot to talk about this morning. Thanks for joining. I originally wanted to do this call on what's going on with agentic AI in the SaaS markets, but obviously, given what's going on with the US invasion of Iran and commodity markets, we need to talk about that first. And I'm going to try to get through both topics in about 50 minutes. We're

(DESCRIPTION)

Slide: Worst military events in US history as per the National Interest. Black and white photograph: A crashed helicopter lies in pieces on the ground with debris scattered around. The broken rotor and torn fuselage sit near a flat open landscape, and parts of the aircraft are charred and twisted.

(SPEECH)

going to go through a lot of information in this deck. If anybody wants a copy of these slides, you can contact your sales coverage, and they can get it to you.

I've been asked why did the US invade Iran, and I don't have any opinions on that other than what you've read. Except I looked at this article in The National Interest, which was written by someone from the services, and they wrote an article on the worst five military events in US history. And the number one worst military event, going all the way back to the 1700s, was the Desert One disaster that took place in 1980, in the deserts of Iran, in a failed effort to rescue the hostages.

And I'm only mentioning this, because I think for anybody that's under the age of 50 or 55, it's hard to understand the hold that event has on older people in this country, whether they're politicians or policymakers or people in the administration. And that's the only context I wanted to share, which is don't underestimate the psychological impact of that event on people of a certain age. And you can see the other events here date all the way back to 1776, the Battle of Long Island.

(DESCRIPTION)

Slide: Strait of Hormuz is shut, unlike prior other disruptions. A line chart titled Strait of Hormuz crude exports before and after disruptions tracks million barrels per day across days before and after events. Multiple colored lines represent different disruptions such as Hormuz drills, US pull out of JCPOA, Israel strike on Iranian Embassy in Damascus, 12 Day War, and Epic Fury, and most lines fluctuate between roughly 10 and 20 million barrels per day with varying patterns around the disruption point. One line labeled Epic Fury drops sharply to near zero after the event and remains very low, while others recover more quickly after initial volatility. Source: JPMorgan Global Commodities Research, March 12, 2026.

(SPEECH)

So I'm going to try to go through this quickly because you know all of this. The Strait of Hormuz, unlike other disruptions, the Strait of Hormuz is practically completely shut at this point to container ship traffic and commodities. Efforts

(DESCRIPTION)

Slide: What about US escorts to keep the Strait of Hormuz open?

(SPEECH)

to keep the Strait open are very complicated because of drone swarms, suicide speedboats, missile attacks, and mines.

There are wooden mine, clearing vessels and underwater drone mine, hunting equipment. But obviously, these things take time and are another example of asymmetric warfare, because you're using things that cost millions to destroy things that cost thousands. And obviously, even if the traffic partially resumes, Iran looks like it will retain some capabilities to continue to torment certain entities using the Strait, and they appear prepared to do so.

(DESCRIPTION)

Slide: Will the US seize Kharg Island which accounts for 90% of Iranian oil exports? A map shows the Persian Gulf region with Iran, Iraq, Kuwait, Saudi Arabia, Qatar, and the UAE labeled, and the Strait of Hormuz highlighted near the entrance to the gulf. An inset map focuses on Kharg Island, labeling a crude oil terminal, an airfield, and an industrial port. A small locator inset shows the island’s position within the broader region, and a scale reads 100 miles.

(SPEECH)

Kharg Island, at this point, I think we all know, easy to seize, hard to protect. So the US could try to seize Kharg Island, through which 90% of Iranian exports flow. But defending it and keeping it there and protecting the troops against missile and drone attacks, I think, would be much more complicated.

(DESCRIPTION)

Slide: Declining global energy intensity. A line chart titled Energy intensity of GDP, global tracks exajoules of consumption per $2021 global real GDP from 1935 to 2025. A red line for Oil rises sharply through the mid 20th century, peaks around the 1970s, and then declines steadily, while a brown line for Coal trends downward over time and a gray line for Natural gas rises mid century before gradually declining. Vertical dashed lines mark the Iranian Revolution and the Gulf War, after which all energy intensity measures generally trend lower. Source: Energy Institute, IEA, OWID, Shift Energy Portal, JPMAM, 2025.

(SPEECH)

Here's where I wanted to get into some things that you might not have seen before. On a glass-half-full basis, which I'm trying as hard as I can to use here, the energy intensity of global output has fallen a lot and particularly with respect to oil. This goes back all the way to the 1920s.

The oil intensity of GDP actually peaked in the 1970s, around the time of the Iranian revolution. It's fallen by 2/3 since then and has fallen in half since the Gulf War. And this has happened mostly because of more efficient cars and planes and furnaces and lighting and refrigeration and air conditioners and also from fuel switching in certain countries from coal to gas. You have much more efficiency in gas turbines than you do in most coal turbines.

So

(DESCRIPTION)

Slide: Declining energy share of household spending. A line chart titled US spending on gasoline and other energy goods tracks the share of total consumer spending from 1970 to 2025. The line rises to peaks above 5.5 percent around the late 1970s and early 1980s, then trends downward over time with smaller spikes around 2008 and the early 2010s before falling to near 2 percent by the 2020s. The overall trend shows declining energy spending as a share of consumer expenditures despite periodic volatility. Source: FRED, BEA, JPMAM, December 2025.

(SPEECH)

the good news is that the energy intensity of GDP has come way down, and US household spending on gasoline and other energy goods is now 1 and 1/2% to 2% of total consumer spending, also way down from prior periods. So if we're looking for silver linings here, every year, the world gets less sensitive to what's taking place in the Gulf, or obviously, there are still some big impacts.

(DESCRIPTION)

Slide: Why oil is a global market. A line chart titled Brent vs WTI crude oil prices tracks US dollars per barrel from 2019 to 2026. Two lines represent Brent and WTI, and both move closely together, rising sharply to peaks above $120 around 2022 before declining and fluctuating between roughly $60 and $90 in subsequent years. Near 2026, both lines rise again toward around $100, maintaining a narrow spread. Source: Bloomberg, JPMAM, March 13, 2026.

(SPEECH)

The thing to remember here-- and I think this is really important-- oil is a global market. Yes, the US is at this point a marginal oil exporter, or certainly not the oil importer that it used to be. But look at this chart-- Brent global non-US oil prices and West Texas Intermediate prices have gone up almost the same amount.

So US consumers and businesses aren't necessarily shielded here in any material way from what's taking place, because oil is a global market. Very

(DESCRIPTION)

Slide: …natural gas is not. A line chart titled Natural gas prices by region tracks US dollars per MMBtu from 2019 to 2026. Lines for Dutch TTF Natural Gas Futures and Japan Korea Marker LNG Futures spike dramatically above $30 to $40 around 2022 before falling and stabilizing in the $10 to $20 range, while the US Natural Gas Futures line remains much lower throughout, mostly below $10 with smaller fluctuations. Source: Bloomberg, JPMAM, March 13, 2026.

(SPEECH)

different than natural gas markets, where Henry Hub prices have barely budged in contrast to pipeline and LNG prices in Europe and in Asia. So that distinction, I think, is really important to understand when we get to the issue of who's exposed to this and by how much.

(DESCRIPTION)

Slide: Doubling of oil prices tends to coincide with material S&P selloffs. A two panel chart titled S&P 500 vs WTI crude oil price shows percent year over year changes from 1970 to 2025. The top panel displays S&P 500 returns fluctuating widely with periodic sharp declines, while the bottom panel shows WTI crude oil price spikes with large upward surges around major geopolitical events such as the Yom Kippur War, Iranian Revolution, Gulf War, and postpandemic surge. Vertical dashed lines align oil spikes with equity drawdowns. A table at right labeled Length of WTI crude oil price spikes lists historical episodes and notes whether prices reversed quickly or remained elevated. Source: Bloomberg, Macrotrends, JPMAM, March 13, 2026.

(SPEECH)

Now, doubling of oil prices-- I know this chart is hard to see, you can look at it later. But typically, when oil prices double, you get material S&P sell-offs. They either coincide with them or precede them. Oil prices are up 50% year on year, so we're not near that 100% threshold. But obviously, it's something that we need to watch.

(DESCRIPTION)

Slide: Highest Strait of Hormuz oil and gas exposure as a share of energy consumption. A vertical bar chart titled Crude oil, condensate and LNG imports via the Strait of Hormuz as a share of primary energy consumption, 2024 shows country level percentages up to about 35 percent. Greece and South Korea appear highest above 30 percent, followed by Thailand, Taiwan, Singapore, Pakistan, and India in the 20 to high 20 percent range, while countries such as China, Belgium, France, the US, Spain, and the United Kingdom show much lower exposure near or below 10 percent. Source: EIA, IEEFA, OEC, Vortexa, JPMAM, 2025.

(SPEECH)

And here is-- I think there's a lot of information there about who imports what. There's not enough information that's being normalized. Like, yes, China imports more oil than Luxembourg from the Strait of Hormuz. Let's look at it relative to something meaningful. So here we look at crude oil, condensate, and LNG imports, from the Strait of Hormuz as a share of each country's primary energy consumption.

So you can see that the US and China are at the low end of the range here relative to some countries in Asia and in Northern Europe.

(DESCRIPTION)

Slide: Three broader factors determine a country's sensitivity to these events. A bullet point list.

(SPEECH)

But because oil is a global market, there's actually three things that determine how countries and markets get impacted here-- how much do you import in terms of oil and gas as a percentage of your energy? How much of oil and imported gas is as a percentage of your total energy consumption? And what is your oil and gas consumption relative to your GDP energy intensity?

So we combine these three measures into one, and we create an index of oil and gas sensitivity to the events that are unfolding. And this is what it looks like. And

(DESCRIPTION)

Slide: An index of oil & gas sensitivity (imports, consumption, GDP). A vertical bar chart titled Oil & gas sensitivity index shows index values up to about 70 across countries. Singapore and South Korea rank highest near the top, followed by the Netherlands, Italy, Thailand, Mexico, Spain, and Japan in descending order, while countries such as the United Kingdom, the United States, and Norway appear lower on the scale. Red arrows highlight selected countries including Malaysia and the US. Source: EIA, IEA, JPMAM, 2025.

(SPEECH)

the US is a little more than halfway down. China's all the way down. At the end of the day, China between its gas-- I'm sorry, between its domestically produced gas, its heavy reliance on coal, its growing share of renewables, and its nuclear power, is actually much less exposed to what's going on in the Gulf than a lot of people think. And also, they have a lot of domestic energy stocks to draw down on.

So this is primarily an issue for other countries in Southeast Asia and countries in Northwest Europe. Those are the ones that are going to have the biggest issues. And when we plot this oil and gas sensitivity index against changes in equity markets year to date, there's a rough correlation that works pretty well.

(DESCRIPTION)

Slide: Biden drawdown of SPR for political/inflation reasons substantially altered flexibility. A line chart titled US Strategic Petroleum Reserve total inventory tracks barrels in millions from 2010 to 2026. The line remains relatively stable around 700 million barrels through the early and mid 2010s, then trends downward and drops sharply around 2022 to near 350 million barrels before partially recovering toward about 400 million by 2025, followed by a labeled March 2026 drawdown point near 250 million. Source: DOE, Bloomberg, JPMAM, March 6, 2026.

(SPEECH)

So where are we here? At first, on the Strategic Petroleum Reserve, the Biden drawdown of the Reserve, which took place for political and inflation reasons, substantially is impairing US flexibility here. I've written about that a lot. There is going to be some drawdown. And as you can see, it's going to bring the SPR to levels that it hasn't been at in a long time.

Other things the administration is thinking about are, on the margin, increasing the ethanol blend from 10% to 15% of vapor pressure waiver, which means you can sell winter-grade gasoline in summer. A possible ban on refined products exports, which are about 30% of US production.

Maybe a Jones Act suspension, which prohibits foreign ships from moving oil from one place to the US to another. And then maybe waiving federal taxes on oil and gas. But again, these things are all on the margin. I don't blame the administration for talking about them, but they're no substitute for an end of hostilities and some reversion to where we were before.

(DESCRIPTION)

Slide: Global SPR release: not the reaction they were hoping for. A line chart titled Brent crude oil intraday prices tracks US dollars per barrel over several days in March. The line spikes sharply above $115 around March 8, then drops quickly below $90 before rebounding toward $100 by March 12, with a dashed vertical line marking the IEA oil stock release announcement. Source: Bloomberg, JPMAM, March 13, 2026.

(SPEECH)

This is probably one of the most depressing charts in the deck, which is the global SPR release, which was announced a few days ago. Brent oil prices went up. That's not the reaction they were looking for. They released 400 million barrels. It's equivalent to about 23 days of imports coverage for the IEA members.

It's about 20% of the existing global reserves, counting both public and private emergency reserves. Very different from the reaction during the Gulf War when the SPR release caused oil prices to drop pretty substantially. So this is the market saying they don't think the SPR release is going to help, because it's not enough, and it's going to get exhausted too soon.

(DESCRIPTION)

Slide: Gulf War III could be the most severe oil market imbalance since World War II. A table titled Approximate oil disruptions and global spare capacity compares historical crises with supply disrupted and available spare capacity as percentages of world supply. Earlier events such as the Suez Crisis, Six Day War, Arab Oil Embargo, and Iranian Revolution show disruptions generally between 5 and 10 percent with varying spare capacity, while more recent events such as Gulf War II and Abqaiq show low spare capacity near 1 percent. The row labeled Gulf War III 2026 present shows a much larger disruption of about 20 percent alongside zero spare capacity. Source: Rapidan Energy Group, EIA, BP, St. Louis Fed, US Senate, March 9, 2026.

(SPEECH)

And as a reminder, this Gulf War III, which is what some people are referring to it as, is the worst oil market imbalance since World War II. Not the same in terms of impacts on GDP, we've talked about that, or percentages of consumption, because energy consumption is down. But if you're looking just specifically at the issue of how much supply is disrupted as a percentage of world supply and the amount of available capacity after the disruption, this one's the worst, because 20% of supply is offline, and there's no more spare capacity. So this is a pretty tight one.

And I was looking at some charts on jet fuel this morning-- in the US, Asia, and Europe, jet fuel prices are up more than 100%, even though Brent oil prices, crude oil prices, are only up 50%, which is something that you can see happening in the refined products market.

(DESCRIPTION)

Slide: If the situation is not resolved soon, price increases needed to balance supply and demand (either through demand destruction or increased supply) are large. A bar and scatter chart titled Realized price increases vs price needed to balance lost volumes from Middle East supply disruption compares percent changes across commodities. Blue bars show front month price appreciation for oil, LNG, TTF, Henry Hub, aluminum, urea fertilizer, ethane, and propane, while gold dots indicate the higher price increases needed to fully balance lost supply. In most cases the dots sit well above the bars. Source: Bridgewater, March 10, 2026.

(SPEECH)

So just to wrap up on this question of Iran, if the situation is not resolved soon, meaning three to six weeks, three to four weeks, we're going to need price increases on a lot of commodity-related items to balance supply and demand. And there's two ways of balancing supply and demand in a situation like this. Prices can go up and destroy demand, or prices can go up and elicit more supply.

At some oil price, the Canadian tar sands are profitable, maybe even at levels close to today's prices. We'll see. But when you look at oil, natural gas, and aluminum fertilizer, ammonia-related products made from natural gas, ethane, propane, the break-even prices, which are the dots on this chart, which are required to balance the market, those are substantially higher than the front-month contracts are showing today. So this is a really severe disruption and one that will need to be resolved within the next few weeks in order to avoid a bigger hit to markets and to GDP.

As

(DESCRIPTION)

Slide: US economy heading into GWIII.

(SPEECH)

a reminder, the US economy heading into this whole thing was looking OK. The labor markets were kind of weak, but we had improving manufacturing services surveys, Fed easing some fiscal stimulus, low corporate and household delinquency rates, some nice improvements in productivity. So if we can get past this, the prior trend was actually pretty favorable.

So

(DESCRIPTION)

Slide: What is citrine? Most is not naturally occurring, it’s simply anthropogenically melted quartz. A visual collage labeled Citrine shows yellow to orange quartz crystals in raw cluster form alongside two faceted gemstones with bright amber and golden tones. The raw crystals are translucent with pointed terminations, while the cut stones have reflective surfaces and geometric facets that enhance color and brilliance.

(SPEECH)

now to the other topic. First question, what is citrine? So just to be clear, citrine is a gemstone. But 95% of what you see isn't naturally occurring citrine. It's man-made melted quartz, and I have an example of it here. And it's nice and shiny, but this is melted quartz.

The reason I'm bringing this up is it's in reference to the Citrini memo from about a month ago that roiled global markets and created a lot of reactions to it. And the fakeness of this citrine and its unreliability is a metaphor for how I feel about parts of the Citrini memo, but not all of it. So let's go through it

(DESCRIPTION)

Slide: Collapsing share prices with little to no erosion in earnings I have seen this movie before. A line chart titled Managed care earnings vs performance tracks an index from 2020 to 2025. A blue line labeled Performance rises steadily through 2021 to 2024 with volatility, then drops sharply in 2025, while gold and red lines for 12 month trailing earnings and 12 month forward earnings trend upward more smoothly and remain relatively stable. Source: Bloomberg, JPMAM, March 4, 2026.

(SPEECH)

Now, what's happened in the SaaS market, so software as a service, is that the markets have collapsed even though earnings haven't budged. Now, we've seen this movie before, just a year ago. Here's a chart on the managed care industry within health. The blue line shows a collapse in the performance of the stocks. They got obliterated by 40% to 50%. If you looked at forward earnings projections and trailing earnings projections, there didn't seem to be that much going on. So what's going on here?

Well, in this case, the markets turned out to be right.

(DESCRIPTION)

Slide: The markets were right. A line chart titled Managed care earnings vs performance tracks an index from 2020 to 2026. A blue line labeled Performance rises through 2024 with volatility, then drops sharply in 2025 before partially recovering, while a gold line for 12 month trailing earnings climbs steadily and then declines into 2026. A labeled point marks 12 month forward earnings near 2026. Source: Bloomberg, JPMAM, March 13, 2026.

(SPEECH)

There was a surge. There was a problem with Medicare Advantage-- the federal funding cuts, lower reimbursement rates, DOJ probes. And then somewhere between 10 to 15 million people are going to lose their insurance by 2034 based on some of the administration decisions. And so in this case, the markets were right. And within a few months, the earnings projections caught up to what the markets were pricing in.

So I'm using this as an example because this is the template that, I think, some people are assuming is going to happen in software, but I'm not quite so sure.

(DESCRIPTION)

Slide: Is the market right about SaaS as well? A bar chart titled Disruption is coming compares market cap in US dollars billions for the Agentic AI market and the SaaS market. The SaaS market bar towers above at over $300 billion, while the Agentic AI market appears much smaller near the bottom of the scale. Source: Fortune, MarketsandMarkets, 2026.

(SPEECH)

Now, is the market right about the SaaS sell-off? Right now, the size of the SaaS market is over $300 billion in terms of market cap. And the agentic AI market is only a few billion dollars, whether it's 5 or 10 billion.

But disruption is coming. And yes, there will be some convergence here where the agentic AI number goes up and the SaaS market goes down. The question is when and to whom. And to put some definitions in place here-- traditional software as a service companies, companies adapt to the SaaS interface, and you have to adapt to everything that they do. Whereas with agentic AI, it does what you want, and it's an agent.

And just to put some context on it-- the average company-- I thought these numbers were amazing. The average large enterprise has over 290 SaaS relationships, up from 110 just in 2020. So there's a lot of SaaS companies out there that have their claws into different parts of these large companies. And the question is, can companies do it themselves now and destroy the value proposition of these companies?

(DESCRIPTION)

Slide: I hate surveys. A horizontal bar chart titled How AI agents are delivering value to businesses shows share of survey respondents across categories. Increased productivity ranks highest at around 70 percent, followed by cost savings, faster decision-making, and improved customer experience in the 50 to 65 percent range, while categories such as improved resilience and continuity and enhanced stakeholder trust are lower near 20 to 30 percent. Source: PwC AI Agent Survey, May 16, 2025.

(SPEECH)

I generally hate surveys. The population samples are low, and they're over extrapolated. But there was a PwC agentic AI survey a few months ago and a little more than 50% of the respondents said that they were getting cost and productivity benefits and faster decision making. If agentic AI was really as good as advertised, I would expect those numbers to be higher. But anyway, here's that survey.

(DESCRIPTION)

Slide: Software as a Service (SaaS): the epicenter of the AI shock. A bullet point list.

(SPEECH)

Just to get into more of the details. There's three primary risks that agentic AI programs that you've heard about, like Claude Cowork and things like that, pose. First, companies could build their own in-house tools and replace their SaaS relationships. The second risk is that companies could use AI to reduce the number of people that have to use those SaaS programs. And since there's a lot of them charge on a per-head basis, that's a problem.

And the third risk is that, if the SaaS companies have to start adopting AI within their own product offerings, now, all of a sudden, the marginal acquisition costs of a new customer are not zero. So you can see some margin pressures on those companies. So those are the big risks that the market's been responding to.

And

(DESCRIPTION)

Slide: Most vulnerable SaaS functions. A bullet point list.

(SPEECH)

if you look through what the most vulnerable SaaS functions are, it makes sense-- data entry, pipeline management, customer support, simple data analysis and reporting, project management timelines, things like that. Those are the SaaS functions that are, I think, most at risk here.

(DESCRIPTION)

Slide: An open-source proxy for enterprise Agentic AI DIY. A line chart titled A proxy for enterprise use of coding agents tracks coding agent edits to Github codebases in thousands from Sep 2025 to Mar 2026. A blue line labeled Claude Code rises sharply from low levels to over 200 thousand edits by early 2026, while a gold line labeled OpenAI Codex remains relatively flat and lower, around 40 to 80 thousand. Vertical dashed lines mark releases such as Anthropic Opus 4.5 and OpenAI GPT-5.3 Codex and Anthropic Opus 4.6, coinciding with acceleration in Claude Code usage. A note states 4% of current GitHub code edits and new additions are coauthored by Claude Code at a minimum. Source: Paul Kedrosky, March 2, 2026.

(SPEECH)

The hardest thing to do is to get a sense for how much this is actually happening behind the scenes. The closest I've been able to get to a proxy for enterprise use of agentic AI, do-it-yourselfism, is this chart that looks at what's going on within publicly available GitHub code bases.

So there's a lot of jargon here, but the bottom line is when Anthropic's Opus 4.5 was released a few months ago, and then last month you got an update to 4.6 and OpenAI's version, you started to see a material uptick in the number of edits and additions within these GitHub code databases, which suggests that people are taking advantage of these tools. It's a pretty crude proxy, but it tells us that something is actually going on.

And

(DESCRIPTION)

Slide: A promotional survey from a company that creates DIY SaaS tools. A vertical bar chart titled SaaS tools at risk for replacement shows the percent of respondents who plan to replace SaaS tools. Workflow automations rank highest at about 35 percent, followed by internal admin tools and business intelligence tools near 30 percent, while CRMs or sales tools and form builders sit around 25 percent. Categories such as project management, customer support, approval or routing, and data integrations appear lower in the 15 to 20 percent range, with Other at the lowest level. Source: Retool, February 17, 2026.

(SPEECH)

obviously, this is a promotional survey from a company that creates the do-it-yourself SaaS tools, but they talk about all the different workflows here in this chart that respondents are saying that they plan on building now for themselves to replace their SaaS relationships. But that's very promotional.

(DESCRIPTION)

Slide: Not just a SaaS issue: What is the value proposition for pure merchant intermediaries when merchants can in principle use agentic AI to deal with banks directly? A table titled Top US merchant acquirers lists companies ranked by 2023 US processing volume, transactions, and number of merchants. JPMorganChase ranks first with about $2.4 trillion in processing volume and 893,000 merchants, followed by Fiserv and Worldpay with large volumes and transaction counts, while firms such as Stripe and Square show smaller processing volumes but large merchant bases. Source: Payment Industry Intelligence, April 2024.

(SPEECH)

And this is not just a SaaS issue. This is the one that really-- if I was working at certain companies, would keep me up at night.

So you have banks, and then you have merchants. And then there's a number of entities in between them that I've heard described as pilot fish, which are just fish that graft onto the backs of other fish. There's a bunch of pilot fish companies that don't really do much other than facilitate the relationships between merchants and banks. And I just wonder, what is the value proposition for those merchant intermediaries, Stripe being one of them?

When the merchants can, in principle, start to use agentic AI to deal with the banks directly and Stripe and Square-- entities like that charge anywhere from 2 and 1/2% to 3%, so there's real money here. And this table looks at the processing volume, number of transactions, and the number of merchants for the top 10 merchant acquirers. For the nonbank entities here, this is where you have to think about what these agentic AI programs will do to them.

(DESCRIPTION)

Slide: SaaS companies are not standing still and have incumbent benefits. A bullet point list.

(SPEECH)

Now, the SaaS companies as well as the companies on the prior page and the merchant acquisition business are not standing still. There are incumbent benefits. Some of them are migrating away from fees per seat to fees based on utilization, whether it's measured in tokens or API calls. These companies can build agentic AI into their own product offerings, maybe faster than some of their clients can, and become what's referred to as orchestration layers.

And then, when I meet with people within JP Morgan who are responsible for a lot of the systems that operate in the firm this size, they talk about the gravity of existing SaaS systems. And so what do they mean by that? Well, a lot of the SaaS relationships we have are related to things on regulatory audit trails, payrolls, financials. These things took several years to implement. Thousands of people were trained in order to use them. There's deep operational embedding. There are lots of regulatory entanglement issues.

And then the last point I bolded because reliability is paramount. And I think sometimes there's been a little bit of a rush to judgment on the market about what these AI tools can reliably do in different industries. So let's talk about that for a minute.

So

(DESCRIPTION)

Slide: While it looks like the functionality of LLMs is exponentially improving, this is at 50% accuracy (i.e., the worst employee you’ve ever had). A scatter plot titled AI software engineering task complexity shows time in hours for tasks AI models can complete at 50% accuracy versus LLM release date from 2022 to 2026. Data points rise sharply over time, with earlier models near zero hours and later models such as GPT-5, Gemini 3 Pro, GPT-5.2, and Claude Opus 4.5 reaching several hours, while Claude Opus 4.6 stands out near 12 hours. A dotted curve illustrates accelerating progress. Source: METR, March 2026.

(SPEECH)

guilty as charged, I have shown this chart before. So this is a chart that shows what AI programs are capable of doing by measuring the time that humans take to do the same thing. And you can see over time that it's gone from a few minutes to several hours. And the latest Claude Opus release, it can now do what takes humans 12 hours.

But there is a little thing in the subheader of this chart. I always knew it was there. But this is only looking at doing it at 50% accuracy. So now let's look at what these dots would look like if you needed these programs to be 80% accurate, which is still worse than the worst analyst I've ever had. But let's raise the threshold to 80% instead of 50% percent. And the dots collapse.

So

(DESCRIPTION)

Slide: At 80% accuracy, which is still unacceptable, the picture is quite different… A scatter plot titled AI software engineering task complexity shows time in hours for tasks AI can complete at 50% vs 80% accuracy across LLM release dates from 2022 to 2026. Blue points for 50% accuracy rise sharply over time, reaching several hours and up to around 12 hours for Claude Opus 4.6, while gold points for 80% accuracy remain much lower, generally near zero to about 1 to 2 hours even for newer models. Source: METR, March 2026.

(SPEECH)

this is like an hour-long webcast, almost. This, to me, is one of the most important charts in the deck, because what it's showing you is the extent to which the ability of these things to reliably accomplish certain tasks collapses once your accuracy threshold goes up.

(DESCRIPTION)

Slide: Autonomous AI agents need a lot of safety work before companies can adopt them. A bullet point list. At right, a man wearing a spacesuit helmet looks forward with colored light reflections across his face against a dark background.

(SPEECH)

And so yeah. A lot of these bugs can be worked out, but these autonomous AI agents need a lot of safety work before companies can adopt them.

And I'm going to give you a few examples that I consider some of these to be hilarious, although they probably weren't to the people that were affected by them. So a bunch of researchers looked at the new agentic AI programs like Claude, Opus, and Moonshot, and they saw some pretty remarkable things. And I have a picture here from 2001, Space Odyssey.

Remember the guy, he goes out to fix something and HAL, the computer, won't let him back in. That's what we're talking about here. Look at this. Unauthorized compliance with nonowners, disclosure of sensitive information, denial-of-service conditions. They attacked other companies that they thought were competitors. Uncontrolled resource consumption, identity spoofing, system takeovers, cross agent propagation of unsafe practices.

It's one thing for these programs to be wrong. It's another thing for them to be engaging in these destructive behaviors. And

(DESCRIPTION)

Slide: Autonomous AI agents need a lot of safety work before companies can adopt them. A bullet point list.

(SPEECH)

there's plenty of examples, if you want to look for them. Alibaba reported that some of its AI agents repurposed GPU to do crypto mining on its own accord without any prompting. Amazon is now requiring senior approval after some AI-assisted code decided that it was just going to wipe out a bunch of protocols in a production environment.

And then probably the most important, METR, which does a lot of research on this thing, determined that around half of the proposed coding changes that were proposed by AI agents were actually too flawed to be implemented, even though they passed one of the code benchmark tests called SWE.

And then lastly, researchers have identified-- and this is hardly surprising-- over 40,000 vulnerabilities in OpenClaw. So there's a lot of work that has to get done to increase the reliability of these things, and that's what I think the investors who have rushed for the exit in SaaS stocks may be missing.

So

(DESCRIPTION)

Slide: SaaS investors rush for the exit. A line chart titled Indicative software ETF daily turnover shows percent of market cap from Jan 2024 to early 2026. The line remains relatively low and stable below 10 percent through most of 2024 and 2025, then spikes sharply in early 2026 to peaks above 60 percent before dropping but staying elevated. Source: Bloomberg, JPMAM, March 13, 2026.

(SPEECH)

this is a chart that looks at the ETF volume. Its come down a bit. But a month ago, there was just a scramble to dump these SaaS stocks.

(DESCRIPTION)

Slide: Investors selling despite stability in forward earnings projections… A line chart titled Software earnings vs performance tracks an index from Jan 2025 to early 2026. A blue line labeled Performance rises modestly through mid 2025, then declines sharply into 2026 toward the low 80s before a slight rebound, while a gold line labeled Consensus 12 month forward earnings trends upward more steadily and remains elevated despite a late dip. The divergence shows falling stock performance even as forward earnings expectations stay relatively stable. Source: Bloomberg, JPMAM, March 13, 2026.

(SPEECH)

And here's the chart I was referring to earlier, which is-- 12-month forward earnings look reasonably stable, and yet the performance of these stocks has collapsed. So this rhymes with the managed care chart that we saw earlier.

Now, a lot of SaaS companies have multi-year contracts and higher renewal rates, and so maybe it's not surprising that what we're talking about here isn't impacting their 12-month projected earnings. It would take maybe three to five years, or two to four years maybe, for those things to show up. Let's see what else.

Now,

(DESCRIPTION)

Michael skips a slide. Slide: …with no apparent deterioration in earnings revisions. A line chart titled Large cap software & services 3 month earnings revisions tracks the 3 month percent change in forward earnings as a 6 month moving average from 1990 to 2025. The line fluctuates between roughly minus 10 percent and plus 15 percent over time, with cyclical ups and downs but no sustained downward trend in recent years, and it remains slightly positive into 2025. Source: Bloomberg, JPMAM, February 27, 2026.

(SPEECH)

again, there haven't been any deterioration. There hasn't been any deterioration in earnings revisions in the large-cap software and services space. So again, not just the earnings projections, but the earnings revisions are pretty stable. And this is what the valuation picture looks like. This is a chart that looks at the price-to-sales ratio relative to the market from large-cap software and services. We're back all the way to 1991.

So to me, this is pricing in a lot of bad news. And

(DESCRIPTION)

Slide: SaaS valuation aftermath: pricing in a lot of bad news. A line chart titled Large cap software & services relative forward price to earnings ratios shows the forward P/E ratio relative to the S&P 500 from 1991 to 2026. The line peaks above 2.0 in the late 1990s, declines sharply in the early 2000s, and fluctuates between roughly 1.0 and 1.6 over the following decades, before falling back near 1.0 by 2026. A dashed line marks parity with the S&P 500, and the latest point sits close to that level. Source: Bloomberg, JPMAM, March 13, 2026.

(SPEECH)

similarly, if you want to look at PE relative, PE ratios, you get the same story. So there's a lot of really bad news priced in here. And I would just counsel our clients not to be too reactive to the next things that you read, because the markets have already reacted. And I actually think there's an opportunity for active managers to sift through this to pick what the ultimate winners are going to be, because software is suddenly very unloved.

The

(DESCRIPTION)

Slide: Software is suddenly unloved: positioning and short interest. A two panel chart shows trends in positioning and short interest. The left chart titled North America hedge fund TMT positioning shows a blue line for software and services declining sharply into negative territory by 2026, while a gold line for semiconductors rises strongly into positive territory, indicating divergence in investor positioning. The right chart titled Software median short interest as a percent of shares outstanding shows a steady rise from around 3 percent in 2023 to nearly 6 percent by 2025. Source: JP Morgan Positioning Intelligence, February 2026, and JPM Equity Strategy & Global Thematic Research, March 6, 2026.

(SPEECH)

first chart here looks at TMT positioning. Semiconductors-- everybody is off the charts. But software and services is highly negative. And then the chart on the right shows that short interest, as a percentage of shares, has really skyrocketed. So from an investment perspective, I would not be surprised at all to see the software stocks catch a bid at some point within the next few months because of these dynamics.

(DESCRIPTION)

Slide: The premise of the labor and economic AI-pocalypse. A bullet point list.

(SPEECH)

So let's go back to this Citrini thing for a minute, because what they wrote about was the premise that there was going to be some labor and economic AI apocalypse here, where AI models continue to take over high labor, high-skilled labor tasks. Adoption increases productivity. More people get fired, consumer spending weakens, corporate profits collapse, earnings go down, and the whole thing spirals on itself. Household net worth goes down. Federal and local tax receipts suffer. People start defaulting on their mortgages.

And so I'm getting asked lots of questions about this Citrini memo and its copycats, which have been appearing. And

(DESCRIPTION)

Slide: The premise of the labor and economic AI-pocalypse. A bullet point list.

(SPEECH)

the sectors mentioned in this Citrini AI apocalypse are-- it's hard to find sectors that didn't get mentioned in this thing. I just want to walk through a few examples of how I consider a lot of the underlying analysis there to be uncertain or flawed, for better or for worse.

And

(DESCRIPTION)

Slide: Piling on: Anthropic’s now viral labor AI-pocalypse chart. A radar chart titled Observed vs theoretical share of job tasks AI can perform compares percentages across categories such as management, business and finance, computer and math, architecture and engineering, legal, education, healthcare, and others. A red shaded area labeled Theoretical spans much larger portions across most categories, while a smaller blue shaded area labeled Observed remains closer to the center, showing a large gap between potential and actual performance. Source: Massenkoff and McCrory (Anthropic), March 5, 2026.

(SPEECH)

by the way, it's not just Citrini. Anthropic piled on last week with now what is probably one of the most viral charts I've seen and which most people have seen, which is this polar chart. And the center of the polar chart has a little blue region, which looks at the percentage of job tasks that AI are performing now, and then the red splotch is what they could do at some point in the future.

Again, everybody started freaking out about this. Anthropic was pretty honest about it that this is a long way away, and a lot of this isn't feasible. And this was just a little bit of navel-gazing on their part. I would also add that Anthropic and other AI labs rarely, if ever, address the issue of four-nines reliability, which is critical for enterprise adoption. But anyway, let's keep going.

(DESCRIPTION)

Slide: Also: Anthropic sees no meaningful difference between AI and non-AI exposed unemployment rates… A line chart titled Unemployment differences for AI exposed vs non-AI exposed workers tracks the difference in differences coefficient from 2016 to 2026. The line fluctuates around zero, with a sharp negative dip around 2020 before recovering and stabilizing near zero in later years, and a dashed vertical line marks the ChatGPT release. Source: Massenkoff and McCrory (Anthropic), March 5, 2026.

(SPEECH)

Anthropic was at least honest that when they looked at their measure of AI and nonAI-exposed unemployment rates, they're not seeing a difference. So if we were in the midst already of a big AI adjustment, you would start to see the line on this chart go up a lot, because the unemployment rate for AI-exposed sectors would be much higher than for nonAI-exposed sector, not happening. And

(DESCRIPTION)

Slide: …even for the youngest cohort of workers. A line chart titled New job start differences for AI exposed vs non-AI exposed workers age 22–25 tracks the DiD coefficient as a percent of the 2022 baseline from 2016 to 2026. The line shows high volatility with large negative dips around 2020 and smaller fluctuations thereafter, including a brief rise near 2022, before trending modestly negative again into 2025 and 2026, and a dashed vertical line marks the ChatGPT release. Source: Massenkoff and McCrory (Anthropic), March 5, 2026.

(SPEECH)

you don't even really see any meaningful difference when you look at the age cohorts of recent college graduates.

And so there's something-- I think the term is AI washing. There are times when companies-- and I think this may have happened recently with Block-- there are times that companies are going to want to make rational firing decisions because they overhired, and then they're going to blame it on AI. And so I don't know. Just be careful about that kind of thing.

Now,

(DESCRIPTION)

Slide: Why is the entry-level tech job unemployment rate declining again? A line chart titled Unemployment rate in tech occupations, age 20–30 only tracks percent from 2020 to 2025. The line fluctuates between roughly 1 percent and 8 percent, with spikes around 2020 and 2024, followed by a decline into 2025 toward the lower end of the range. A bullet point at right reads Block and AI-washing. Source: IPUMS CPS, JPMAM, January 2026.

(SPEECH)

if we're in the midst of this job-eating AI apocalypse, why are entry-level tech unemployment rates declining again? So a few months ago, it looked like this number was going up and up and up. Then there was a gap because of the government shutdown. These numbers have been declining again.

So again, there's a lot of data that raises questions about how much this is even happening right now. The unemployment share of tech jobs is still stable.

(DESCRIPTION)

Slide: Why is the employment share of tech jobs still stable? A line chart titled Employment share in tech occupations tracks percent from 2020 to 2025. The line rises sharply to around 4.6 percent in early 2020, then drops to near 3.7 percent, and stabilizes with fluctuations between roughly 3.9 percent and 4.2 percent through 2025. The overall pattern shows relatively steady levels after the initial volatility. Source: IPUMS CPS, JPMAM, January 2026.

(SPEECH)

This number is not going up. Next chart, the employment share of software developers is rising. Why

(DESCRIPTION)

Slide: Why is the employment share of software developers rising? A line chart titled Employment share in software developer occupation tracks percent from 2020 to 2025. The line fluctuates between roughly 1.2 percent and 1.5 percent, with dips around 2021 and 2023, and a steady upward trend into 2025 where it reaches near the top of the range. Source: IPUMS CPS, JPMAM, January 2026.

(SPEECH)

is that happening if we're in the midst of agentic AI and Copilot era that's destroying all those jobs? I don't know.

And

(DESCRIPTION)

Slide: Why are software job postings rising again?? A line chart titled Indeed US job postings for software engineers tracks an index from Jan 2024 to Jan 2026. The line declines from the low 70s in early 2024 to the low 60s by early 2025, then increases steadily through 2025 into 2026, reaching around 70. Source: FRED, Indeed, JPMAM, March 6, 2026.

(SPEECH)

look at this one-- job postings for software engineers. It was easy last summer to fall into a pit of despair on this thing, because the job listings were plummeting. Now, they're surging again. And so this is raising to me a lot of questions about the Citrini and Anthropic and a lot of the thesis here about job destruction.

Then

(DESCRIPTION)

Slide: Real estate? A line chart titled US real estate brokerage and leasing revenues tracking seasonally adjusted US billions from 2012 to 2024. A blue line labeled Seasonally adjusted revenue rises steadily from around $90 billion in 2012 to near $190 billion by 2024, while a gold line labeled 2 year trailing growth rate fluctuates widely, dipping below zero around 2020, spiking above 20 percent around 2021, and settling near mid single digits by 2024. Source: US Census Bureau, JPMAM, February 20, 2026.

(SPEECH)

a few other things, just that I wanted to get into. Citrini makes the argument that this is going to kill real estate brokers. Real estate brokers are like Rasputin. You can shoot them, stab them, and drown them, and poison them, but they're still there. The internet, even before the latest agentic AI solutions, had completely made private information obsolete. There's information abundance now rather than asymmetry.

And the fact that real estate agents still exist is a sign that consumers do see value in what they're providing. Brokerage commissions continue to grow at about 6% a year, which is not that different from pre-COVID levels. So for all the talk about disintermediation in the real estate brokerage industry, maybe it's going to happen. We're not seeing it.

(DESCRIPTION)

Slide: Merchant payments and stablecoins? A bullet point list.

(SPEECH)

And then one of the other Citrini thesis was about merchants would broadly adopt and embrace stablecoins. The notion that they would do that in the absence of a standardized fraud liability framework, to me, is dubious. As we know, everything crypto touches eventually becomes a magnet for illicit and fraudulent activity.

And so the average merchant that already has pretty thin margins, I think, is going to be very hesitant to embrace something broadly where there's no fraud liability framework. Stablecoins are used in 70% of all crypto scam transactions and on more than 80% of crypto payments to sanction countries and individuals. So I'm not buying it.

(DESCRIPTION)

Slide: Merchant payments and stablecoins? A bullet point list.

(SPEECH)

These are the three issues about stablecoins-- irreversibility-- and we all this. But once a transaction is confirmed on the blockchain, it can't be reversed. And if a merchant is defrauded, there is no central repository or authority to trigger a chargeback. Lack of uniform protections, unlike you have in the traditional banking system. And then the GENIUS Act was very focused on reserves and compliance issues, but not on the issue of consumer and merchant protection against theft.

And

(DESCRIPTION)

Slide: Stablecoins, merchants and payments fraud. A vertical bar chart titled Payment methods currently accepted & added in past year shows percent of merchants surveyed across payment types. Blue bars for currently accepted methods are highest for cards near 80 percent and digital wallets near 70 percent, followed by bank transfers or direct debit and ecommerce mobile payment around 40 to 60 percent, while categories such as real time payments, buy now pay later, gift cards or vouchers, and cash on delivery appear lower. Gold bars for added in past 12 months are smaller across categories, with modest additions in digital wallets, account to account payments, and buy now pay later, and a highlighted category labeled Cryptocurrency appears at a relatively low level. Source: 2025 Global eCommerce Payments & Fraud Report, Visa.

(SPEECH)

when we look at some data that just came out from a Visa report that looks at the payment methods that are currently accepted and have been added in the last year, crypto is all the way at the bottom of this chart. And so we're not seeing evidence that there's a major movement in the direction of merchant or merchant acquirers to somehow incorporate crypto

(DESCRIPTION)

Slide: Radiology? A bullet point list.

(SPEECH)

Radiology. This goes all the way back. This is one of my favorite ones. There's a guy named Geoffrey Hinton, who's the godfather of AI, and Vinod Khosla, both of whom predicted around a decade ago that radiologists would no longer be needed. Why are we graduating them? We don't need them anymore.

Actually, what's happening is imaging is increasing 3% to 4% a year due to an aging population and chronic disease and expanded access to preventative care. But the number of practicing radiologists is only growing at about 1%. We have a radiology shortage. We don't have a radiology obsolescence issue. And we have some facts and figures here in terms of the degree of understaffing. But again, this is another one that I think is completely off the mark.

And

(DESCRIPTION)

Slide: Customer service and call center reps show some decline… A line chart titled Employment share in customer service representative occupation tracks percent from 2020 to 2025. The line fluctuates around roughly 1.5 percent to 1.8 percent through 2020 to 2023, then trends slightly lower through 2024 and 2025, ending near about 1.4 percent. Source: IPUMS CPS, JPMAM, January 2026.

(SPEECH)

then let's go to the epicenter of the shock. What's the thing that you think that agentic AI would be most able to displace? Would be customer service agents and call center reps. Very low value added and things that are supposedly the sweet spot for some of these agentic AI programs. When you look at their share of total employment, they've ticked down a tiny bit. You have to squint to see it.

So this job-pocalypse-- there's definitely going to be some changes in the future as we adapt to these new technologies. But in my opinion, the markets and some of the doomsayers are overreacting to some very minor blips in the data as things stand right now. Look at this. Job

(DESCRIPTION)

Slide: But why are job listings rising again? A line chart titled Indeed US job postings for customer service representatives tracks an index from 2020 to 2026. The line drops sharply to around 55 in early 2020, then rises to a peak near 140 by 2021 to 2022, followed by a steady decline through 2023 and 2024 into the mid 80s, and then increases modestly into 2026 toward around 90. Source: FRED, Indeed, JPMAM, March 6, 2026.

(SPEECH)

postings for customer service reps are going up again after falling for five or six years. So some of the data just does not conform to the thesis that are being published on the average Substack.

And

(DESCRIPTION)

Slide: And why does India, the epicenter of global calls centers, keep making more money on call center/tech support functions? A line chart titled India information and communication technology service exports tracks current US dollars in billions from 2000 to 2024. The line rises steadily from under $10 billion in 2000 to around $50 billion by 2008, continues upward to about $75 billion by the mid 2010s, then accelerates after 2020, reaching roughly $180 billion by 2024. Source: World Bank Group, JPMAM, 2024.

(SPEECH)

by the way, why does India, which is the epicenter of global call centers, keep making more and more money on tech support functions? This is a chart on India information and communication technology service exports, basically the dollars and billions that they make from doing this. This thing has been rising, not falling.

(DESCRIPTION)

Slide: Some weakness on job postings for legal and accounting but no smoking gun, yet. A line chart titled Indeed US job postings by industry tracks an index from Jan 2024 to Jan 2026. A gold line for Legal remains near 100 in early 2024, then trends gradually downward into the mid 90s by 2026, while a blue line for Accounting declines more sharply from around 100 to the low 80s in 2024, drops further to the high 60s in 2025, and then recovers modestly into the low 80s by 2026. Source: FRED, Indeed, JPMAM, March 6, 2026.

(SPEECH)

Now, where do we see some job weaknesses? We're seeing weaknesses in legal. We're seeing some weaknesses in accounting. No smoking gun yet, but you can see it happening, and you hear about it happening in terms of anecdotes.

(DESCRIPTION)

Slide: The big picture. A bullet point list.

(SPEECH)

But there are some big picture things that the Citrini note and other similar Substack articles get very wrong about economics, in my opinion.

The average American household spends, let's say, 8 to $12,000 a year on services whose primary value proposition is making their lives easier like tax preparation, insurance brokerage, travel booking, financial advice, real estate, and other kinds of subscription fees. In principle, the latest revolution should completely change that, because it's going to give me the ability to do all of that and cut out all sorts of fees.

We'll see. People still shop a lot on Amazon, even though it's not the cheapest option in many cases. But OK, let's accept the premise. If these costs go down or go away, the value is then rerouted to other things in the economy. It doesn't vanish, and it doesn't get disintegrated. And that's the premise of the Citrini death spiral is that it inherently assumes that the value's destroyed, whereas it's dead. It's just repurposed. And the same thing takes place with the disintermediation of corporate expenses. They get rerouted to other expenditures, including business expansion and hiring.

(DESCRIPTION)

Slide: The big picture. A bullet point list.

(SPEECH)

And so the 20th century is full of examples of productivity changes, which tended to result in faster job growth, not slower job growth, within three to five years. So let's look at a chart that looks at what happens after there's a productivity shock. There

(DESCRIPTION)

Slide: Unemployment impacts of productivity shocks tend to be modest and brief. A line chart titled Impact of a 1% technology-driven labor productivity shock on unemployment tracks change in unemployment rate across quarters since a 1% productivity increase. The line rises from about 0.2 percent in quarter 1 to a peak near 0.35 percent around quarter 2 to 3, then declines steadily through subsequent quarters, reaching near zero by quarter 8, with vertical error bars around each point indicating variation. Source: Goldman Sachs Global Investment Research, February 27, 2026.

(SPEECH)

does tend to be-- for a 1% technology driven labor productivity shock, you will get, let's say, 40 basis points, 30 basis points, increase in the unemployment rate within a few quarters. But by two years, it's gone.

And

(DESCRIPTION)

Slide: Over the long run, productivity growth tends to coincide with employment growth. A scatter plot titled Productivity & employment growth in expansionary cycles shows nonfarm payroll growth versus nonfarm productivity growth across historical periods. Points range from about 1 percent to 4 percent productivity growth and 1 percent to 5 percent employment growth, with labeled periods such as 1949–1953, 1961–1969, 1975–1980, and 2020–2025. A dotted upward sloping trend line runs through the points, indicating a positive relationship between productivity and employment growth across cycles. Source: Bloomberg, BLS, NBER, JPMAM, January 2026.

(SPEECH)

then the important issue is over the longer run, when you look at the business cycles, the higher the productivity growth rate, the higher actual employment rate is. For the most part, productivity shocks eventually result in more employment growth than in job destruction, and that's been the case over the last 100 or so years.

And

(DESCRIPTION)

Slide: The positive employment consequences of the AI buildout. A line chart titled Data center-exposed construction growth tracks employees added since Jan 2022 in thousands. A gold line labeled data center-exposed construction rises steadily from near zero in 2022 to around 450 thousand by 2026, while a blue line labeled other construction increases to around 300 thousand by 2024 and then declines slightly to the mid 200 thousand range by 2026. Source: BLS, JPMAM, December 2025.

(SPEECH)

by the way, let's not forget that there are some pretty notable benefits to this whole AI build-out. Look at the job growth in things related to data centers. This chart is looking at utilities and HVAC contractors, electricians, transmission equipment, transformers, et cetera. Those jobs are growing pretty quickly and are a partial offset against some of the jobs that are being lost elsewhere.

And then I'll just conclude with this chart. David

(DESCRIPTION)

Slide: The Autor chart. A stacked bar chart titled 60% of today’s employment is in occupations that did not exist in 1940 shows employment in millions across categories such as professionals, managers, clerical and admin, production, construction, personal services, transportation, technicians, sales, cleaning services, health, and farming. Each bar splits into two colors representing occupations that existed in 1940 and those that did not, with the latter comprising a large share across most categories. A final Total bar highlights that occupations not present in 1940 make up the majority of current employment. Source: Autor et al, NBER, 2022.

(SPEECH)

Autor is a professor at MIT who I've spoken with many times. He does some of the most interesting work on issues related to labor markets and productivity. And a lot of you may have seen this chart before, but I thought it was a great way to conclude this discussion, which is 60% of the jobs that exist today, those occupations didn't even exist in 1940.

And so it's hard to anticipate right now exactly what new jobs are going to arise. But through most productivity shocks, new jobs get created, and it avoids the job death spiral that Citrini and others have been referring to.

So anyway, back to the first topic on Iran and the war. The events of the weekend were not super encouraging. We saw attacks on energy infrastructure, specifically oil and gas loading stations in Fujairah and other attacks on other energy infrastructure in the Gulf.

So far, developed and developing countries are not responding robustly to the president's request for additional naval support to open the Strait. And so we're now in the part of this conflict where the uncertainty is rising.

But again, I think it'll be another three to four weeks before we get clarity on what's going to happen to energy markets. And I do think that some of the larger economies, like the US and China, are more insulated from the damage here than people might think. Refer back to those charts in the beginning of the deck in terms of which countries are more affected. So that's today's webcast. Thank you very much for participating. I'll be accepting bids for this piece of fake citrine, and I hope to talk to you all again soon. So long.

Thank you for joining us. Prior to making financial or investment decisions, you should speak with a qualified professional in your JP Morgan team. This concludes today's webcast. You may now disconnect.

[MUSIC PLAYING]

(DESCRIPTION)

IMPORTANT INFORMATION, This material is for information purposes only. The views, opinions, estimates and strategies expressed herein constitutes Michael Cembalest’s judgment based on current market conditions and are subject to change without notice, and may differ from those expressed by other areas of JPMorgan Chase & Co. (“JPM”). This information in no way constitutes J.P. Morgan Research and should not be treated as such. Any companies referenced are shown for illustrative purposes only, and are not intended as a recommendation or endorsement by J.P. Morgan in this context. GENERAL RISKS & CONSIDERATIONS, Any views, strategies or products discussed in this material may not be appropriate for all individuals and are subject to risks. Investors may get back less than they invested, and past performance is not a reliable indicator of future results. Asset allocation/diversification does not guarantee a profit or protect against loss. Nothing in this material should be relied upon in isolation for the purpose of making an investment decision.

NON-RELIANCE, Certain information contained in this material is believed to be reliable, however, JPM does not represent or warrant its accuracy, reliability or completeness, or accept any liability for any loss or damage (whether direct or indirect) arising out of the use of all or any part of this material. No representation or warranty should be made with regard to any computations, graphs, tables, diagrams or commentary in this material, which are provided for illustration/reference purposes only. Any projected results and risks are based solely on hypothetical examples, and actual results and risks will vary depending on specific circumstances. Forward-looking statements should not be considered as guarantees or predictions of future events. Nothing in this document shall be construed as giving rise to any duty of care owed to, or advisory relationship with, you or any third party. Nothing in this document shall be regarded as an offer, solicitation, recommendation or advice (whether financial, accounting, legal, tax or other) given by J.P. Morgan and/or its officers or employees. J.P. Morgan and its affiliates and employees do not provide tax, legal or accounting advice. You should consult your own tax, legal and accounting advisors before engaging in any financial transactions.

For J.P. Morgan Asset Management Clients, J.P. Morgan Asset Management is the brand for the asset management business of JPMorgan Chase & Co. and its affiliates worldwide. To the extent permitted by applicable law, we may record telephone calls and monitor electronic communications to comply with our legal and regulatory obligations and internal policies. Personal data will be collected, stored and processed by J.P. Morgan Asset Management in accordance with our privacy policies at https://am.jpmorgan.com/global/privacy. ACCESSIBILITY, For U.S. only: If you are a person with a disability and need additional support in viewing the material, please call us at 1-800-343-1113 for assistance. This communication is issued by the following entities, In the United States, by J.P. Morgan Investment Management Inc. or J.P. Morgan Alternative Asset Management, Inc., both regulated by the Securities and Exchange Commission, in Latin America, for intended recipients’ use only, by local J.P. Morgan entities, as the case may be, in Canada, for institutional clients use only, by JPMorgan Asset Management (Canada) Inc., which is a registered Portfolio Manager and Exempt Market Dealer in all Canadian provinces and territories except the Yukon and is also registered as an Investment Fund Manager in British Columbia, Ontario, Quebec and Newfoundland and Labrador, in the United Kingdom, by JPMorgan Asset Management (UK) Limited, which is authorized and regulated by the Financial Conduct Authority, in other European jurisdictions, by JPMorgan Asset Management (Europe) S.à r.l., in Asia Pacific (“APAC”), by the following issuing entities in the respective jurisdictions in which they are primarily regulated, JPMorgan Asset Management (Asia Pacific) Limited, or JPMorgan Funds (Asia) Limited, or JPMorgan Asset Management Real Assets (Asia) Limited, each of which is regulated by the Securities and Futures Commission of Hong Kong, JPMorgan Asset Management (Singapore) Limited (Co. Reg. No. 197601568K), whose advertisement or publication has not been reviewed by the Monetary Authority of Singapore, JPMorgan Asset Management (Taiwan) Limited, JPMorgan Asset Management (Japan) Limited, which is a member of the Investment Trusts Association, Japan, the Japan Investment Advisers Association, Type II Financial Instruments Firms Association and the Japan Securities Dealers Association and is regulated by the Financial Services Agency (registration number “Kanto Local Finance Bureau (Financial Instruments Firm) No. 330”), in Australia, to wholesale clients only as defined in section 761A and 761G of the Corporations Act 2001 (Commonwealth), by JPMorgan Asset Management (Australia) Limited (ABN 55143832080) (AFSL 376919). For all other markets in APAC, to intended recipients only.

LEGAL ENTITY, BRAND & REGULATORY INFORMATION, In the United States, bank deposit accounts and related services, such as checking, savings and bank lending, are offered by JPMorgan Chase Bank, N.A. Member FDIC. JPMorgan Chase Bank, N.A., and its affiliates (collectively “JPMCB”), offer investment products, which may include bank managed investment accounts and custody, as part of its trust and fiduciary services. Other investment products and services, such as brokerage and advisory accounts, are offered through J.P. Morgan Securities LLC (“JPMS”), a member of FINRA and SIPC. Insurance products are made available through Chase Insurance Agency, Inc. (CIA), a licensed insurance agency, doing business as Chase Insurance Agency Services, Inc. in Florida. JPMCB, JPMS and CIA are affiliated companies under the common control of JPM. Products not available in all states.

In Germany, this material is issued by J.P. Morgan SE, with its registered office at Taunustor 1 (TaunusTurm), 60310 Frankfurt am Main, Germany, authorized by the Bundesanstalt für Finanzdienstleistungsaufsicht (BaFin) and jointly supervised by the BaFin, the German Central Bank (Deutsche Bundesbank) and the European Central Bank (ECB). In Luxembourg, this material is issued by J.P. Morgan SE – Luxembourg Branch, with registered office at European Bank and Business Centre, 6 route de Trèves, L-2633, Senningerberg, Luxembourg, authorized by the Bundesanstalt für Finanzdienstleistungsaufsicht (BaFin) and jointly supervised by the BaFin, the German Central Bank (Deutsche Bundesbank) and the European Central Bank (ECB). J.P. Morgan SE – Luxembourg Branch is also supervised by the Commission de Surveillance du Secteur Financier (CSSF), registered under R.C.S Luxembourg B255938. In the United Kingdom, this material is issued by J.P. Morgan SE – London Branch, registered office at 25 Bank Street, Canary Wharf, London E14 5JP, authorized by the Bundesanstalt für Finanzdienstleistungsaufsicht (BaFin) and jointly supervised by the BaFin, the German Central Bank (Deutsche Bundesbank) and the European Central Bank (ECB). J.P. Morgan SE – London Branch is also supervised by the Financial Conduct Authority and Prudential Regulation Authority. In Spain, this material is distributed by J.P. Morgan SE, Sucursal en España, with registered office at Paseo de la Castellana, 31, 28046 Madrid, Spain, authorized by the Bundesanstalt für Finanzdienstleistungsaufsicht (BaFin) and jointly supervised by the BaFin, the German Central Bank (Deutsche Bundesbank) and the European Central Bank (ECB). J.P. Morgan SE, Sucursal en España is also supervised by the Comisión Nacional del Mercado de Valores (CNMV) and jointly supervised by the Banco de España as a branch under code 1567. In Italy, this material is distributed by J.P. Morgan SE – Milan Branch, with its registered office at Via Cordusio, n.3, Milan 20123, Italy, authorized by Bundesanstalt für Finanzdienstleistungsaufsicht (BaFin) and jointly supervised by the BaFin, the German Central Bank (Deutsche Bundesbank) and the European Central Bank (ECB). J.P. Morgan SE – Milan Branch is also supervised by the Commissione Nazionale per le Società e la Borsa (CONSOB), registered with Bank of Italy as a branch of J.P. Morgan SE under code 8076, Milan Chamber of Commerce Registered Number: REA MI 2536325.

In the Netherlands, this material is distributed by J.P. Morgan SE – Amsterdam Branch, with registered office at World Trade Centre, Tower B, Strawinskylaan 1135, 1077 XX, Amsterdam, The Netherlands, authorized by the Bundesanstalt für Finanzdienstleistungsaufsicht (BaFin) and jointly supervised by the BaFin, the German Central Bank (Deutsche Bundesbank) and the European Central Bank (ECB). J.P. Morgan SE – Amsterdam Branch is also supervised by the Netherlandsche Bank (DNB) and the Autoriteit Financiële Markten (AFM) in The Netherlands, registered with the Kamer van Koophandel as a branch of J.P. Morgan SE under registration number 72610220. In Denmark, this material is distributed by J.P. Morgan SE – Copenhagen Branch, filial af J.P. Morgan SE, Tyskland, with registered office at Kalvebod Brygge 39-41, 1560 København V, Denmark, authorized by the Bundesanstalt für Finanzdienstleistungsaufsicht (BaFin) and jointly supervised by the BaFin, the German Central Bank (Deutsche Bundesbank) and the European Central Bank (ECB). J.P. Morgan SE – Copenhagen Branch, filial af J.P. Morgan SE, Tyskland is also supervised by Finanstilsynet (Danish FSA) and is registered with Finanstilsynet as a branch of J.P. Morgan SE under code 29010. In Sweden, this material is distributed by J.P. Morgan SE – Stockholm Bankfilial, with registered office at Hamngatan 15, Stockholm, 11147, Sweden, authorized by the Bundesanstalt für Finanzdienstleistungsaufsicht (BaFin) and jointly supervised by the BaFin, the German Central Bank (Deutsche Bundesbank) and the European Central Bank (ECB). J.P. Morgan SE – Stockholm Bankfilial is also supervised by Finansinspektionen (Swedish FSA), registered with Finansinspektionen as a branch of J.P. Morgan SE.

In Belgium, this material is distributed by J.P. Morgan SE – Brussels Branch with registered office at 35 Boulevard du Régent, 1000, Brussels, Belgium, authorized by the Bundesanstalt für Finanzdienstleistungsaufsicht (BaFin) and jointly supervised by the BaFin, the German Central Bank (Deutsche Bundesbank) and the European Central Bank (ECB). J.P. Morgan SE – Brussels Branch is also supervised by the National Bank of Belgium (NBB) and the Financial Services and Markets Authority (FSMA) in Belgium, registered with the NBB under registration number 0715.622.844. In Greece, this material is distributed by J.P. Morgan SE – Athens Branch, with its registered office at 3A Irodou Attikou, Athens, Greece, authorized by the Bundesanstalt für Finanzdienstleistungsaufsicht (BaFin) and jointly supervised by the BaFin, the German Central Bank (Deutsche Bundesbank) and the European Central Bank (ECB). J.P. Morgan SE – Athens Branch is also supervised by Bank of Greece, registered with Bank of Greece as a branch of J.P. Morgan SE under code 124. Athens Chamber of Commerce Registered Number 158683760001. In France, this material is distributed by J.P. Morgan SE – Paris Branch, with its registered office at 14, Place Vendôme 75001 Paris, France, authorized by the Bundesanstalt für Finanzdienstleistungsaufsicht (BaFin) and jointly supervised by the BaFin, the German Central Bank (Deutsche Bundesbank) and the European Central Bank (ECB) under code 842 422 972. J.P. Morgan SE – Paris Branch is also supervised by the French banking authorities the Autorité de Contrôle Prudentiel et de Résolution (ACPR) and the Autorité des Marchés Financiers (AMF). In Switzerland, this material is issued by J.P. Morgan (Suisse) SA, with registered address at rue du Rhône, 35, 1204, Geneva, Switzerland, which is authorised and supervised by the Swiss Financial Market Supervisory Authority (FINMA) as a bank and a securities dealer in Switzerland.

In Hong Kong, this material is distributed by JPMCB, Hong Kong branch. JPMCB, Hong Kong branch is regulated by the Hong Kong Monetary Authority and the Securities and Futures Commission of Hong Kong. In Hong Kong, we will cease to use your personal data for our marketing purposes without charge if you so request. In Singapore, this material is distributed by JPMCB, Singapore branch. JPMCB, Singapore branch is regulated by the Monetary Authority of Singapore. Dealing and advisory services and discretionary investment management services are provided to you by JPMCB, Hong Kong/Singapore branch (as notified to you). Banking and custody services are provided to you by JPMCB Hong Kong/ Singapore Branch (as notified to you). For materials which constitute product advertisement under the Securities and Futures Act and the Financial Advisers Act, this advertisement has not been reviewed by the Monetary Authority of Singapore. JPMorgan Chase Bank, N.A. is a national banking association chartered under the laws of the United States, and as a body corporate, its shareholder's liability is limited. It is registered as a foreign company in Australia with the Australian Registered Body Number 074 112 011. With respect to countries in Latin America, the distribution of this material may be restricted in certain jurisdictions. Issued in Australia by JPMorgan Chase Bank, N.A. (ABN 43 074 112 011/AFS Licence No: 238367) and J.P. Morgan Securities LLC (ARBN 109293610). References to "J.P. Morgan* are to JPM, its subsidiaries and affiliates worldwide. "J.P. Morgan Private Bank" is the brand name for the private banking business conducted by JPM. This material is intended for your personal use and should not be circulated to or used by any other person, or duplicated for non-personal use, without our permission. If you have any questions or no longer wish to receive these communications, please contact your J.P. Morgan team, © 2026 JPMorgan Chase & Co. All rights reserved.

For J.P. Morgan Wealth Management Clients, Purpose of this material, This material is for informational purposes only. The views, opinions, estimates and strategies expressed herein constitutes Michael Cembalest’s judgment based on current market conditions and are subject to change without notice, and may differ from those expressed by other areas of J.P. Morgan. This information in no way constitutes J.P. Morgan Research and should not be treated as such. J.P. Morgan is committed to making our products and services accessible to meet the financial services needs of all our clients. If you are a person with a disability and need additional support, please contact your J.P. Morgan representative or email us at accessibility.support@jpmorgan.com for assistance.

J.P. Morgan Wealth Management is a business of JPMorgan Chase & Co., which offers investment products and services through J.P. Morgan Securities LLC (JPMS), a registered broker-dealer and investment advisor, member FINRA and SIPC. Annuities are made available through Chase Insurance Agency, Inc. (CIA), a licensed insurance agency, doing business as Chase Insurance Agency Services, Inc. in Florida. Certain custody and other services are provided by JPMorgan Chase Bank, N.A. (JPMCB). JPMS, CIA and JPMCB are affiliated companies under the common control of JPMorgan Chase & Co. Products not available in all states. This material is intended for your personal use and should not be circulated to or used by any other person, or duplicated for non-personal use, without our permission. If you have any questions or no longer wish to receive these communications, please contact your J.P. Morgan representative.

LEGAL ENTITY, BRAND & REGULATORY INFORMATION, The views, opinions and estimates expressed herein constitute Michael Cembalest’s judgment based on current market conditions and are subject to change without notice. Information herein may differ from those expressed by other areas of J.P. Morgan. This information in no way constitutes J.P. Morgan Research and should not be treated as such. The views contained herein are not to be taken as an advice or a recommendation to buy or sell any investment in any jurisdiction and there is no guarantee that any of the views expressed will materialize. Any forecasts, figures, opinions or investment techniques and strategies set out are for information purposes only, based on certain assumptions, current market conditions and are subject to change without prior notice. All information presented herein is considered to be accurate at the time of writing, but no warranty of accuracy is given and no liability in respect of any error or omission is accepted. This material does not contain sufficient information to support an investment decision and it should not be relied upon by you in evaluating the merits of investing in any securities or products. In addition, investors should make an independent assessment of the legal, regulatory, tax, credit and accounting implications and determine, together with their own professional advisers, if any investment mentioned herein is believed to be suitable to their personal goals. Investors should ensure that they obtain all available relevant information before making any investment. It should be noted that investment involves risks, the value of investments and the income from them may fluctuate in accordance with market conditions and investors may not get back the full amount invested. Both past performance and yield may not be a reliable guide to future performance.

Non-affiliated entities mentioned are for informational purposes only and should not be construed as an endorsement or sponsorship of J.P. Morgan Chase & Co. or its affiliates. J.P. Morgan Wealth Management is the brand for the wealth management business of JPMorgan Chase & Co. and its affiliates worldwide, J.P. Morgan Institutional Investments, Inc. • For J.P. Morgan Private Bank Clients, Please read the Legal Disclaimer. • For J.P. Morgan Asset Management Clients, Please read the Legal Disclaimer. • For J.P. Morgan Wealth Management Clients, Please read the Legal Disclaimer. • For Chase Private Clients, Please read the Legal Disclaimer. © 2026 JPMorgan Chase & Co., All rights reserved.

Logo: J.P. Morgan.