Outlook 2025: The Alchemists

(DESCRIPTION)

Logo: J.P.Morgan Private Bank. 2025 Outlook: The Alchemists. Michael Cembalest, Chairman of Market and Investment Strategy. Michael wears a suit and tie in a headshot photo. Next, Michael stands on stage, wearing a dark blazer over a dark shirt.

(SPEECH)

Thanks for coming. What-- is this February? Yeah.

Yeah.

Usually by this point in the year, we're looking at fourth quarter earnings, and maybe it's the first year of administration, so we're trying to get the outline of what a reconciliation bill might look like, and we're looking at the capital spending surveys, looking at fourth quarter, housing starts, and things like that. We're not paying any attention to any of that right now. [LAUGHS] We're trying to digest this peculiar policy mix that draws from very distinct strains in American history.

So Trump 2.0 takes a little bit from each one of these seven presidents. You've got Coolidge-- and a lot of our clients are-- and Jamie, I would say-- are very excited about the Coolidge element of this Trump administration in terms of pro-business policies, small government, DOGE, et cetera.

(DESCRIPTION)

Slide text: Trump 2.0 policies draw from some faces you may recognize. Portraits of seven presidents sit above their names, terms, and notable policies. Text: Calvin Coolidge (1923-1929) small-government and pro-business policies. James Polk (1845-1849) manifest destiny in the western hemisphere. Andrew Jackson (1829-1837) new political order and deep state battles. Woodrow Wilson (1913-1921) prewar isolationism. Richard Nixon (1969-1974) enemies lists and dreams of energy independence. Dwight Eisenhower (1953-1961) deportation policies. William McKinley (1897-1901) tariffs and protectionism.

(SPEECH)

There's also some James Polk in there. James Polk was all about manifest destiny in the Western hemisphere, whether it's-- Greenland was not on my 2025 bingo card, but now it is, and Panama, Colombia, Mexico, you name it.

Andrew Jackson is hard to avoid Andrew Jackson in this discussion. He created a new political order and told the Supreme Court, OK, you've made your decision, now try and enforce it. That may be coming soon. Wilson was an isolationist before events changed that. But you have to go back to Wilson to find the kind of isolationism that Trump is advocating.

And how could you avoid Nixon and his various vendettas and enemies lists? I mean, the parallels are screaming off the screen. Eisenhower was the deportation president. Eisenhower deported a million people in 1954. But as a sign of how quickly things can change, people didn't like the way it was handled, they didn't like the consequences. And by 1956, they deported 80,000 people. It fell by 95% in two years.

And then McKinley and tariffs and stuff like that. I mean, McKinley, the McKinley tariffs went south almost immediately in terms of how quickly they passed through inflation. Republicans lost 100 seats in the House in the 1890 midterms, and that's when the House was smaller.

So there's a couple of lessons here when you think about how quickly the deportation story changed under Eisenhower and how quickly the political winds shifted under McKinley with policies that can affect the economy.

So this is a weird mix of stuff, and we spend a lot of time trying to figure out how all of the various crosswinds are going to affect the economy at a time when, coming into this year, we're in the 95th percentile of valuations generally in US equities. With the Mag Seven, without the Mag Seven, no matter how you look at it, the Trump administration doesn't have a lot of room for error if it cares about the stock market.

(DESCRIPTION)

Text: Rasputin Eye on the Market, August 2023. In an image, men in lab coats and one in a suit sit around a dark room. One holds a photo of Rasputin under his arm. They all turn their attention to another man, gesturing to another photo of Rasputin.

(SPEECH)

And so, just as a bit of background to how we got to where we are here, there is a general consensus, very high recession forecast in 2023. And we began to notice that, like Rasputin, who shot, poisoned, drowned, and stabbed before they killed him, the US economy was remarkably resilient in 2023 for a place where there was just a 500 basis point increase in policy rates.

What was going on? This economy is a lot more resilient. And so, at that time, in spite of the inverted yield curve and all that stuff, we said, look, we're heading for a soft landing. And that was the cover of the 2024 outlook. The falling bears into the pillows was a metaphor for, we're going to have a soft landing here, and that's what we've had.

(DESCRIPTION)

In an AI-generated image, brown and polar bears fall from the sky onto a vast bed of pillows on the ground.

(SPEECH)

But now that we've had that soft landing, labor market conditions are easing a little bit, but not cratering. Consumer spending, business capital spending are all growing at a pace that's consistent with 2% to 2.5% growth. Single-family, multifamily housing starts that are underway should continue to roll into and be finalized, which reduces pressure on inflation, et cetera, et cetera.

(DESCRIPTION)

Text: US economy/markets. Two other bullet points say, Credit spreads are tight and both high yield and leveraged loan delinquency rates are low. M and A activity should benefit from $750 billion of private equity dry powder and changes at the FTC. On the next slide, a man in an AI-generated image wears a MAGA hat in a middle-ages style room. A large glass jar filled with red bubbles in front of him as he pours a vial into it. Other vials and bottles about the table are labeled, Agency purges, Tariffs, Oil and Gas, Deportation, Medical Freedom, Tax cuts, Cost cutting, and Crypto.

(SPEECH)

But how is this mix of stuff going to turn out? And this was the cover of our 2025 outlook, deregulation, deportations, tariffs, tax cuts, cost-cutting, oil and gas, crypto, the medical freedom movement-- I may never drink bottled water again-- and then purges of government agencies.

So what could go-- I think you have to ask the question, what could go wrong here? And in the outlook, I concluded that if they stick to this-- now, you've seen over the last 48 hours that the President can change his mind. The tariffs have already been pulled on Mexico and Canada depending upon what concessions may happen.

But if they stick to this policy, we'll have a 10% to 15% correction this year. And now, those happen a lot. I mean, roughly a third of all the years over the last 100 years, you've had those kind of corrections. But if they stick unrelentingly to this course, the markets aren't priced for it.

Now, that's not to say that they won't achieve some important national security objectives and re-solidifying supply chains. The pendulum had swung way too far, and there was always going to be a price to pay to dial back the impact of globalization, which has been devastating for American communities. But the markets aren't really priced for that.

(DESCRIPTION)

Text: What are the Alchemists trying to change? Other US data with sharp inflection points: Rising manufacturing job losses. Falling labor share of gross profits. Rising suicide rates. Rising non-metro poverty rates. A graph on the slide is titled, The Silence of the Plants. It has years on the x-axis, from 1930 to 2020, with intervals of 10, It has Index (100 equals 2,000) on the y-axis, with a scale that goes from 0 to 180 with intervals of 20. Three lines on the graph, representing Us industrial production, US GDP, and US consumer spending trend upward in a concave slope, until China joins WTO in 2000, marked by a dashed vertical line on the graph. At this point, the other lines continue an upward trend, but US industrial production begins to plateau. Text: Source: BEA, Federal Reserve, JPMAM, 2024.

(SPEECH)

And this is what, if you strip out the politics and the culture stuff and the gender stuff, this is what the Trump people are trying to fix from an economic perspective, which is that for 80 years, the manufacturing and production side of the economy grew at the same pace as consumption. And then, all of a sudden, in a bipartisan move, both the Republicans and the Democrats supported China joining the World Trade Organization and completely underestimated what the impact would be and what their intentions were.

And with the benefit of hindsight, we can now see that China abrogated the letter and the spirit of almost every single obligation that they were presumably going to adhere to as part of the World Trade Organization. And because of that, since then, consumer spending has been fine and industrial production has been flat and has had some terrible consequences in terms of manufacturing job losses, falling labor, share of gross profits, suicide rates, rural poverty rates, et cetera. So that's what the Trump people are trying to fix.

Will that weird mix of policies fix it? It'll fix some of it, but it's going to do some other stuff along the way.

(DESCRIPTION)

Text: Early verdict from bond markets is not great. A graph on the slide is titled, 10 year Treasury yield change after the first Fed cut. The graph has Business days since first Fed rate cut on the x-axis, with a scale from 0 to 100 and intervals of 10. The y-axis has a scale from negative 1.2% to 1.2%, with intervals of 0.2%. Numerous colored lines on the graph are labeled in a key with months, including the Septembers of 2007 and 1998, the Julys of 1995 and 2019, June of 1989, and January of 2001. Another line, in bolded blue, is labeled, September 2024. Most of the lines oscillate around 0, with many stooping below 0 for much of the graph. The thick September 2024 line soars above all the others, never reaching 0, and trending upward until around 0.8% at 100 days. Text: Source: Bloomberg, JPMAM, January 30, 2025.

(SPEECH)

And so, so far, the markets are a little nervous because, normally, when the Fed is easing, the 10-year Treasury, which is the foundation of all of the investments we all do, either stays the same or rallies. So all those squiggly lines on the bottom of the chart are what the 10-year Treasury does from the moment that the easing cycle begins.

The thick blue line is what's happening now. So the bond markets are clearly nervous about this mix of policies, and we'll see how it turns out. But so far, they're nervous.

(DESCRIPTION)

Text: Same from Fed fund futures markets. A graph is titled, Fed funds target rate: current vs prior market expectations. It has days from January 24 to December 25 of the following year on the x-axis, with intervals of three months. It has percent on the y-axis, with a scale from 3.50% to 5.50% and intervals of 0.25%. A gold line on the graph labeled December 27, 2023 drops precipitously from January to February the following year, going from 5.25 to around 3.50%. A red line on the graph labeled January 30, 2025 drops much more shallowly from February of the second year to December, from 4.25 to just above 3.75%. Text: Source: Bloomberg, JPMAM, January 30, 2025.

(SPEECH)

Same thing with the Fed funds futures markets. A few months ago, the markets were pricing in maybe that we'd bottom out at 3.5 on the funds rate. I think it'll be a miracle if we get to 4.

(DESCRIPTION)

Text: Equity investors are more optimistic. A chart is titled, Share of US households expecting higher stock prices in 12 months, Percent of respondents. The chart has years from 1987 to 2022 on the x-axis, with intervals of 5. The y-axis has a scale from 15 to 60% with intervals of 5%. A jagged line on the graph oscillates around the 30s and 40s, dips during the period from 2007 to 2012, then rises. In 2022, it dips to around 25%, then rises rapidly until the end of the graph, ending above 50% past 2022. Text: Source: Conference Board, Bloomberg, JPMAM, January 31, 2025.

(SPEECH)

Now, equity investors are still very optimistic. They're notoriously backward-looking, but equity investors are very optimistic over the next 12 months. Part of that can be justified by-- and this is one of the weirdest charts I've ever seen, and I've worked at JP Morgan since 1987.

(DESCRIPTION)

Text: Tight equity supplies are supportive. A segmented bar chart is named, Net equity supply globally. It has years from 99 to 23 on the x-axis, with intervals of 2. It has US dollars in billions per year on the y-axis, with a scale from negative $400 to $1,400. A key identifies the colors of the bar segments as Non-US SPAC, US SPAC, US, and Non-US. For most of the graph, the majority of the bars consists of Non-us, reaching above 0. At different points, US and non-us both reach slightly below 0. The only points where US SPAC and Non-US SPAC are visible in the segments are 2020 and 2021, where they make up very small parts of the bars, which are otherwise mostly Non-US and in the positive direction. Text: Source: J.P. Morgan Global Markets Strategy, November 2024.

(SPEECH)

The equity supply has just disappeared. Other than the ill-advised SPAC garbage from 2021, there's just-- there's no new equity supply.

So institutions, corporate pension plans, government pension plans, endowments, foundations, and high-net-worth individuals are deploying money in a market where buybacks are actually outweighing secondary issuance. So that's why some of the numbers for US have actually been negative. So this is a positive for on the supply and demand fundamentals of the equity market.

(DESCRIPTION)

Text: But US equity markets are expensive. A bar chart is titled, Current market valuation percentile vs history; 100% equals most expensive, 0% equals least expensive. The chart has years Market variable with year of inception on the x-axis. Years vary across the axis, with many repeating. The earliest year represented is 1990 and latest, 2014. The colored bars across the chart are divided into three categories, identified with headings across the top of the chart. They are, US equities, representing the largest share of bars, non-US equities, representing the smallest share, then fixed income, representing a share with a size between the two others. The y-axis goes from 0 to 100%, with intervals of 10%. Under US equities, two of the bars that reach almost 100% are S and P 500 EV/sales in 1991, and S and P 500 price/book in 1990. One that reaches just past 80% is S and P 500 price/FCF in 1991. Under non-US equities, a bar that reaches past 90% is MSCI India Forward P/E in 2006. One that hovers near 0% is Europe P/E discount vs US in 1999. In fixed income, a bar that reaches near 100% is US HY spreads in 1995. One that reaches just above 20% is CMBS BBB spreads in 1997. The bars are coded according to height, with shorter ones in greens, mid-height ones in yellows, and taller ones in orange and red. Text: Source: Bloomberg, JPMAM, January 31, 2005.

(SPEECH)

But as I mentioned, the markets are expensive. So one of the things that you can do is-- and you have to have the discipline to do this, even though it's a pain-- you have to look at a bunch of different valuation metrics, the whole market, the median one, the equal-weighted one.

And then you look at the whole history and you take that valuation metric and say, OK, what percentile am I in. And on the left side of this chart, you can see that most of the US equity valuation metrics we follow are in the 90-something-th percentile. So that, again, the point that the administration doesn't have a lot of wiggle room to experiment if they are worried about the midterm elections.

(DESCRIPTION)

Text: There have only been ten instances of multiple 20% years in a row since 1871: Only during the 1990's bull market and the Roaring Twenties did the good times continue. A table is titled, S and P 500: Two consecutive years of 20% plus total returns and what followed. It reads: 1879 48%, 1880 24%, 1881 8%, 1882 2% 1924 25%, 1925 29%, 1926 14%, 1927 35%, 1928 37%, 1929 negative 3%, 1930 negative 23%, 1935 46%, 1936 36%, 1937 negative 31%, 1938 20%, 1950 28%, 1951 26 %, 1952 17%, 1953 1%, 1954 47%, 1955 34%, 1956 6%, 1957 negative 9%, 1975 38%, 1976 23%, 1977 negative 6%, 1978 8 %, 1985 31%, 1986 24%, 1987 0%, 1988 19%, 1995 31%, 1996 23%, 1997 33%, 1998 29%, 2023 26%, 2024 27%, 2025 question mark question mark, 2026, question mark question mark. Text: Source: Shiller dataset, BBG, JPMA M, December 29, 2024.

(SPEECH)

And some of you may have noticed this in some of the stuff we've published. Since 1871, there have only been 10 times when, two years in a row, you got 20% returns. And I'm not a market technician and I generally dislike them. They're kind of weird people.

But only twice in-- since 1871 did you have two 20% years in a row where the good times kept rolling. One was the late '20s, and we know what happened after that. And the other one was the 1990s, which was this incredible constellation of events that was great for multiple years, mistakenly known as the Clinton recovery. He had nothing to do with it.

(DESCRIPTION)

Text: Many companies responded positivelv to election results. A graph is called, Small business capex plans, % of NFIB Survey respondents planning any capex in next 3-6 months, seasonally adjusted. The x-axis has years from 2021 to 2024 and intervals of 1. The y-axis has a scale from 18 to 32%, with intervals of 2%. A jagged line on the graph peaks around 33% in late 2021, drops to its lowest point near 19% in early 2023, then climbs again. It drops again toward 19% in late 2024 before climbing to reach 28% at the end of the graph. Text: Source: Bloomberg, JPMAM, November 30, 2024.

(SPEECH)

So now CEOs are generally optimistic as well. And I can understand why. The deregulatory push from the Trump administration is probably the piece of it that business people and investors like the most. Jamie talks about it all the time. I don't know if I brought some of our deregulatory stuff with us, but the pace of regulation under the Biden administration was like every morning someone woke up and said, something is moving. How can I regulate it to stop moving.

(DESCRIPTION)

Text: This era's US exceptionalism can be backed up with numbers. A bar graph is titled, Free cash flow margins by decade for the ten largest stocks, 1952 to 2024, Percent. The x-axis has decades, from the 1950s to 2020s, with one bar for each decade. The y-axis goes from negative 5 to 30%, with intervals of 5. In the 1950s, the bar reaches about negative 3%, then around negative 2 in the 60s, then positive 5 in the 70s. The bars continue to grow in height each decade, until the 2020s when they reach past 25%. The 2020s bar is labeled, AAPL, MSFT, GOOG, AMZN, META, TSLA, BRK.A, LLY, V. Text: Source: Empirical Research, September 2024.

(SPEECH)

You actually can track the number of economically significant bills. And they lapped the Biden administration-- the Obama administration, which I didn't think was possible, but they did. So the business community is very anxious to take a sigh of relief and just to slow down the pace of regulation for a little bit.

Now, let's just talk about-- I want to talk about this-- the US exceptionalism. Because for all of you that manage money, and for us that manage money, you've had to discard 60 years of postwar lessons learned on the benefits of diversification. Because the last 15 years, diversification has been a curse rather than a blessing with respect to regional equity allocations and portfolios. And there are some remarkable things happening.

And this is one way to think about it. If you take the 10 largest stocks in each decade in the United States and you say, what were their margins, the current crop of the largest stocks are not-- don't make less money than the rest of the market, which is what they used to do, they make more money than the rest of the market, which is incredible. It's like the heaviest guy on your block is also the fastest guy. It's hard to make that much money when you're that big, and yet they are.

And so, these are some incredible, incredibly profitable behemoth companies that we're dealing with here, and you've just been obliterated if you've been trying to invest in an equal-weighted approach.

(DESCRIPTION)

Text: But the rest of the S and P 500 have been treading water. A graph is titled, Earnings per share growth Rolling 2 year growth. It has years on the x-axis from 2017 to 2024 with intervals of 1. The y-axis goes from negative 50% to 250% percent, with intervals of 50%. A gold line on the graph, S and P 500 ex Mag 7, hovers around 0, dipping below 0 just before 2021, then peaking around 75% just before 2023, then sloping down to 0 again. A blue line, Mag 7, hovers around 75%, then climbs steeply, peaking at almost 250% in 2021. It slopes down just as steeply, until it reaches about 25% in 2023 and remains around there, though beginning to slope up, for the rest of the graph. Text: Source: Bloomberg, JPMAM, September 30, 2024.

(SPEECH)

And here you could look at the same thing. The Mag Seven have been generating really substantial and impressive profits growth continually over the last seven years or so, whereas the rest of the market has been kind of flat, went down during COVID, had a post-COVID recovery, and is now back to flat again.

So some of the most optimistic outlooks I've seen for this year is that the S&P 493, which is the gold line, is going to start to pick up. And I believe that they will. I think we'll get 5%, 10%, 12% profits growth from that cohort this year, which would be welcome.

But this is the part that there's-- this chart and the one that follows it were the ones that I think got the most amount of attention, certainly amongst the technology investors in our outlook.

(DESCRIPTION)

Text: Which is why it's so important to understand the AI boom. A bar graph is called, AI capital spending in context, Share of market wide capital spending. The x-axis of the graph has years from 2020 to 2026 E, with intervals of 1, and 1969 and 2000 shown after. The y-axis extends from 0% to 18% with intervals of 2%. The blue bars from 2020 to 2026 E measure NVIDIA data center revenues. They grow modestly from 2020 to 2020, staying below 2%, then grow rapidly from 2023 to 2026 E, where the bar reaches about 13%. The 1969 bar in gold measures IBM peak revenues, and reaches about 17%. The 2000 bar in red measures Cisco, Lucent, Nortel peak revenues, and reaches about 13%. Text: Source: Empirical Research, August 2024.

(SPEECH)

We're heading for something that's only happened twice before, which is a single company or small handful of companies capturing around 12% to 15% of all the capital spending taking place in the entire economy. I mean, that's kind of an amazing thing. It happened during the mainframe era. It happened in the fiber era, and now it looks like it's happening again.

(DESCRIPTION)

Text: The 400 billion-dollar hyperscaler revenue gap. A bar chart is titled, A 400 billion-dollar gap: AI infrastructure spending vs revenues, US dollars billions. The y-axis of the chart goes from $0 to $500. A bar labeled, NVIDA Y25 revenue, reaches just past $100. An arrow from the top of this bar to the next is labeled, GPUs equal 50% of data center operating cost. The next bar is labeled, Implied data center spending, and reaches around $250. An arrow pointing from the top of this bar to the next says, 50% hyperscaler gross margin. The next bar is labeled, Required AI revenue, which reaches $500 at the very top of the graph. An arrow from here points down to the top o the next bar. The arrow reads, 400 billion dollars revenue shortfall: Google, Microsoft, Meta, Apple, Oracle, Bytedance, Alibaba, Tencent, X, Tesla, Other. This last bar is multicolored, with thin horizontal segments. Reaching about $100, this multicolored bar is labeled, Projected AI revenue. Text: Source: Sequoia, Bloomberg, JPMAM, October 2024.

(SPEECH)

And this-- so this is the chart that has been sparking a lot of online conversation, debate, and strong feelings. I adapted it from a piece from David Cahn at Sequoia. And this is what I'm trying to figure out. And whenever people tell me, don't worry, it'll all work itself out, I start having these bad dreams that I'm stuck someplace in a profitless metaverse. So hang with me here while I walk through what's going on.

Let's take NVIDIA's order book for this year, $125 billion. The number-- the estimates differ, but for a new Greenfield data center, GPUs are roughly 50% of the total cost. And if anybody wants the backup for that, I can give it to you. So that means that there's about $250 billion of spending on data centers taking place.

The hyperscalers, which are spending that money, have gross margins above 50%, but let's round down to 50%. So to maintain their gross margins, the hyperscalers, which are the companies mentioned on the right, would have to earn $500 billion from these new capital outlays in order to maintain their gross margins. That's a big amount of money when we can only identify around 100 billion or so of what they're earning right now.

So some people think, well, Mike, it's way too early to look for a killer app within the GenAI space. We're just in the early innings here. All the other signs of AI adoption are pointing in the right direction. Yeah, I wouldn't disagree with that. But that's a lot of money. And essentially, what it's relying on is JP Morgan to say, OK, we have-- I don't know what the current number is, 275,000 employees?

310,000.

310,000? Did you hire a lot of people?

340,000.

340,000. 340,000? Oh, OK. So we desperately need some GenAI applications.

[LAUGHTER]

So let's say we can get that number down to 300. How much would we pay? We'd probably pay a lot for that, particularly if the productivity of the company could grow from where it is now. But the hyperscalers are betting that the corporate sector, including-- with specifically large companies like JP Morgan-- are going to make up that gap. And there are signs that we're moving in that direction. But that's the big question.

And nobody could-- people can't even imagine that there's a chance that the hyperscalers might be making a mistake, which is what worries me. They're probably not. But the fact that people can't even imagine a world in which they would do that is-- whoop-- is an issue.

Now, again, remember how I told you how profitable those companies were?

(DESCRIPTION)

Text: Hyperscalers are profitable but there are limits, even for them. A graph is called, Hyperscaler capex and R and D as a share of revenues, Percent. The graph has years from 2017 to 2025 on the x-axis, with intervals of 1. The y-axis has a scale from 15% to 70%, with intervals of5%. Three colored lines correspond to Amazon, Alphabet, and Microsoft, which all converge around 2021, between 20 and 25%. They hover in the 20s and 30s for the rest of the graph. Another line, Metal, hovers above all the others, around the high 30s and low 40s from 2017 to 2022, where it peaks dramatically above 65%, then drops down toward 45% where it remains. The peak on this Metal line is labeled, Metaverse money incineration. Text: Source: Bloomberg, JPMAM, September 28, 2024.

(SPEECH)

Their CapEx and R&D as a share of revenues has gone up, but only a little. That's very different than the late '90s where companies were spending massive amounts of their free cash flow and revenues.

So this isn't approaching any kind of breaking point here. But it's something that you have to monitor, particularly in a world where it's getting more expensive, until a couple of weeks ago, to train some of these models in order to get better performance out of them.

(DESCRIPTION)

Text: As models become more expensive and energy-intensive to train. A dot plot is titled, Estimated training cost and compute of select AI models. Training cost (US $, thousands, log scale). The chart has Training computer (petaFLOPS, millions, log scale) on the x-axis, with a scale from 0.001 to 100,000, with 0.1, 10, and 1,000 marked in between. The y-axis goes from $0.1 to $1,000,000. 10 different models are plotted with dots on the chart, with Google Transformer in the lower left and Google Gemini Ultra in the extreme upper right. Other Google, Meta, Open AI, and Microsoft models are plotted between these. From low left to upper right, they are, Google BERT-Large, UW/Meta RoBERTa Large, Google LaMDA, OpenAI GPT-3 (davinci), Metal Llama 2, Microsoft NVIDA Megatron-Turing NLG, Google PaLM, and Open AI GPT-4. Text: Source: Stanford Human-Centered Artificial Intelligence, Epoch, 2024 The next slide reads, Surveys are supportive of return on investment in Gen AI. A segmented bar chart is titled, Gen AI projects: return on investment vs expectations, % of survey respondents. The chart has bars titled, Tech, Consumer, Healthcare, Financials, MFG, and Industrials along the x-axis. The y-axis goes from 0 to 100% with intervals of 20%. The bars are divided into large blue segments at the bottom, which correspond to, Exceeded expectations, larger gold segments in the middle labeled, Met expectations, and thin red segments at the top, corresponding to Missed expectations. The largest Missed expectations segment is on the Industrials bar. The Exceeded expectations segments reach into the 30s and 40s, with the Met expectations segments reaching from the tops of these into the 80s and 90s, and the Missed expectations segments at the tops of these, filling the rest of the way to 100%. Text: Source: AlphaWise, Morgan Stanley Research, 2024.

(SPEECH)

And as I mentioned, there's lots of surveys. Hey, CIOs that have done these GenAI applications, how do feel about them? People are generally saying, yeah, they met or exceeded my expectations. So that's positive across these different industries.

(DESCRIPTION)

Text: Tim capsule. A chart is titled, Corning optical communications revenue, US $, billions. The x-axis has years from 2000 to 2024 with intervals of 3. The y-axis goes from $1 to $5 with intervals of 1. Two lines on the graph, Nominal and Real (2000 US $), plunge from 4 to near 1 from 2000 to 2003. Then they separate and slope up, showing the same jagged pattern but growing further apart, as nominal peaks at 5 around 2022, while real peaks just above 3 that same year. From here, both lines slope down, with nominal ending at 4 just before 2024 and real ending around 2.5. Text: Source: Bloomberg, company filings, JPMAM, 2024.

(SPEECH)

But I just want to remind everybody that everybody here that's been around for a long time, we've seen weird things. At the end of the '90s, Corning made $4 billion laying fiber. Every single mile of that fiber ended up getting used, but they outran demand, and it took 20 years for Corning to get back to the level of demand that they had in 1999.

So even in a situation-- so I believe that all of the AI infrastructure that's being built, I have 100% confidence that it's going to be utilized one day. Is it going to be utilized in three years? I don't know. That's the big question. Because investors have to care about the path, not just the end, because we're investing in the companies that are doing all of this. So this is my nightmare for all that capital. I don't think it's going to happen, but it's a useful object lesson to think about.

(DESCRIPTION)

Text: DeepSeek. A dot plot is titled, Cost vs performance of select AI models, MMLU Redux ZeroEval Score (multi-subiect performance). The x-axis is labeled, Input API price, US $ per million tokens (log scale). It has a scale that goes from $0.1, to $1.0, to $10.0. The y-axis goes from 78 to 94, with intervals of 2. 14 different models are plotted as dots on the graph, with GPT-4o-mini in the lower left, Claude 3.5 Sonnet in the upper mid-right, and GLM-4-Plus in the middle and far right. DeepSeek-V3 is plotted in the upper mid-left, with a score between 88 and 90, and price about midway between $0.1 and $1.0. DeepSeek-R1 is plotted very near the top of the graph, still left of center, with a score between 92 and 94, and cost still under $1.0. Text: Source: DeepSeek, JPMAM, January 2025.

(SPEECH)

Now, we're still trying to unravel what happened with DeepSeek. Just as a reminder, China is a place where we have such little confidence in the official statistics that we have to track things like port traffic, electricity consumption, and other kinds of metrics like that to try to figure out what's going on in China. So when I look at the reports of what DeepSeek said it did, it could be like my kid's middle school report where he just made half of it up.

[LAUGHTER]

So, we'll try to figure out what's going on. They appear to have taken a bunch of innovations that existed separately that had already been done, and very proficiently, with less than 200 people, do all of them at the same time, in part because they were under the squeeze of sanctions, and develop a model whose performance is as good or better than the existing models, and offering it at a fraction of the cost.

Notice that I didn't say, doing it at a fraction of the cost. Are they loss-leading here? The x-axis is what they're charging. That's not an assessment of the cost because none of us-- I mean, this is a company where we don't even know how many GPUs they have. So but what we know is, right now, they're willing to offer this thing at a fraction of the price for very good performance. And even Sam Altman came out and said, look, on paper, what they've done is pretty impressive.

So I agree with the people that have said, this is a sign that we're probably headed for a deflationary shock in GenAI applications, which is great for the adopters and corporate adopters of it. It's kind of neutral for the hyperscalers building all the infrastructure, and in the long run could be negative for the chip companies like NVIDIA. But we're a long way from actually having an accurate set of auditable facts from DeepSeek to figure out exactly what's gone on.

(DESCRIPTION)

Text: Dr. Seuss Goes to Europe. A chart is titled, Creation of new public companies in the 21st century, Cumulative market cap in trillions of US dollars. It has an x-axis titled, Number of Companies, that goes from 0 to 200 with intervals of 50. The y-axis goes from 0 to 18 with intervals of 2. A small red sphere, representing Europe, sits around 1 on the y-axis and just past 50 on the x-axis. A trail of red leads up to the sphere, sloping up from around 0. The sphere is labeled, Industria de, Diseno Textil, Airbus, Equinor, Porsche, Adyen. A large blue sphere, representing the US, spans a space between 12 and 18 on the y-axis, at about 200 on the x-axis. A trail of small blue sphere segments leads up to it, sloping upward toward it. The sphere is ablated, Alphabet, Meta, with other labels applying to the sphere segments in the trail. From top to bottom, they are, Tesla, Visa, Mastercard, Netflix, Salesforce, T-Mobile, ServiceNow, and Accenture. Text: Source: Bloomberg, JPMAM, December 17, 2024.

(SPEECH)

I, a lot of times, people ask me about Europe, and I have this really sad, gloomy slide deck on Europe. I've decided, I'm not going to use that anymore. I'm just going to show this one chart, and that answers all of my responses to your questions about Europe. This is the creation of public companies in the 21st century. And if this strikes you as a little bit culturally insensitive, I adapted it from a report that Mario Draghi wrote on the problems with European competitiveness. I took it from there. Mine's better.

So, yeah, normally we talk about the EU has lower labor productivity, and energy prices are two to three times what they are in the United States. The monetary policy challenges are much bigger than in the US, because the inflation and labor mobility differentials across the European countries are massive relative to what they are in the US.

If you go across every single sector, not just tech, but you also go into financials, and utilities, and telecom, and durables, ROE and ROA is higher in the United States than it is in Europe. So when you look at European underperformance versus the US, it's not just multiples, or sentiment, or stuff like that. The actual profitability of European companies is much higher-- of US companies is much higher and et cetera. So we have more stuff on Europe, but you don't need it.

(DESCRIPTION)

Text: What nuclear renaissance? Wake me when we get there. A line chart is titled, Nuclear pure play and exposed stock returns, Total return index (100 equals June 2021). The x-axis goes from 2021 to 2025 with intervals of 1. The y-axis goes from 100 to 450 with intervals of 50. Two lines, Nuclear exposed, and, Nuclear pure play, crawl along near the bottom of the graph, then slope steeply up, beginning around 2023. Nuclear exposed ends at 400 in 2025, with Nuclear pure play ending at 300. Text: Source: Bloomberg, JPMAM, January 30, 2025.

(SPEECH)

The outlook is a reflection of the questions that we're getting, and we were flooded with questions about nuclear. Over the last 18 months, there seems to be a nuclear renaissance. I would describe this as a renaissance of interest in nuclear rather than a renaissance of actual nuclear power. [LAUGHS] And, now, the markets are super optimistic. The nuclear exposed and nuclear pure play companies are soaring.

A lot of those nuclear pure play companies are working on small modular reactors whose designs haven't been approved, that have never been built, and there's not even a shovel in the ground in the United States on an SMR. Everything is in the planning stage. So will it work out? We'll see. But this has a very spacky feel to me in terms of how quickly the stocks have risen relative to what's actually taking place on the ground, which is nothing.

(DESCRIPTION)

Text: The collapse of OECD nuclear power. A line graph is called, During the 1980's, nuclear development shifted from developed to developing countries, Developed market share of global MW of nuclear starts, 5yr avg. The x-axis is labeled, Year of construction start, and goes from 1950 to 2020 with intervals of 10. The y-axis goes from 0 to 100% with intervals of 10%. A line slopes down from 100% in 1950 to around 30% in the 1980s. It dips and peaks modestly, before plummeting to 0 in 2000. It rises and crawls along the bottom of the graph, oscillating into the 20s and 30s and dipping around 0, from the 2000s to 2020. Text: Source: Power Reactor System Database, JPMAM, September 2024.

(SPEECH)

The Western world were the leaders of nuclear power in the '50s, and '60s, and '70s. But really, since the turn of the century, almost all nuclear development is taking place in emerging markets. So it takes a long time to resuscitate a capital-intensive industry where the expertise is gone.

(DESCRIPTION)

Text: The cost gap. A bar chart is titled, Capital cost of nuclear plants in developed countries vs emerging markets, US dollars millions / MW. The y-axis goes from $0 to $20 with intervals of $5. Blue bars, representing selected plants from developed countries, descend in size across the graph. In order from left to right, in descending height, they are, UK, Hinkley Point at around $17, USA, Georgia Vogtle at around $14, France, Flamanville around $9, and Finland, Olkiuoto, around $7. A wide red bar corresponds to the 4 most recently completed EM plants. Reaching around $3, it is labeled, China Taiwan 6, Pakistan Kanupp 3, China Fangchenggang 4, China, Fangchenggang 3. Text: Source: JPMAM, 2024.

(SPEECH)

And here's a look at some recently completed plants in the West, versus the cost of those plants in China and Pakistan. Now, I wouldn't want to live near any of those red plants. I don't think I'd even want to fly over them, but that's how much they cost to complete.

And a lot of really thoughtful, well-meaning people, including people that JPMorgan knows well, that work at something called the Energy Futures Initiative that's run by Ernie Moniz, who's the former Energy Secretary, are really hot on nuclear. And they believe that the learning curve can bring down some of those blue bars.

A learning curve is something that you get when you produce thousands or tens of thousands of something. Building four units and growing it to six is not a learning curve. So I'm really, really suspicious of claims that learning curves can bring down these nuclear costs.

(DESCRIPTION)

Text: Nixon could only dream of energy independence... A line graph is named, US energy independence, Net imports of oil, natural gas and coal as a share of total primary energy consumption. The x-axis has years from 1970 to 2020, with intervals of 10. The y-axis goes from negative 10% to 35%. A line runs horizontally across the graph at 0%. The area above 0 is labeled, Net importer, while the area below 0 is labeled, Net exporter. A blue line begins around 5% in 1970 and climbs, peaking at 30% around 2005. Then it slopes down steeply, ending near negative 10% in 2020. Text: Source: El Statistical Review of World Energy, JPMAM, 2024.

(SPEECH)

And then on energy, so Nixon used to dream of two things. One, he would dream of how to use the IRS to persecute his political enemies. [LAUGHS] That's a real thing. But he also used to dream about energy independence, because he was president, he was unlucky. He was there during the wage and price controls, and the OPEC embargo, and things like that.

We have finally reached-- and broadly speaking-- energy independence in the United States if you combine oil, gas, and coal, and you look at imports versus exports. I mean, this was an unimaginable dream as recently as 20 years ago when we were at the peak energy dependence in the history of the United States. It's been amazing. Now, the bitter pill to swallow here is that hydraulic fracturing is responsible for about 90% of this decline. It is what it is.

(DESCRIPTION)

Text: ...which has finally arrived. An area graph is titled, US production of crude oil, natural gas and natural gas liquids, Trillions of BTUs per month. The x-axis goes from the year 2000 to 2024, with intervals of 2. The y-axis goes from 0 to 8,000 with intervals of 1,000. An area of blue, Crude oil, grows from a height of 1,000 in the year 2000 to around 2,000 in 20204. A gold area, Natural gas, grows from a height of 3,000 to about 6,000 over the same period. A red area, Natural gas liquids, grows from just above 3,000 to about 7,500. The top right corner of the graph reads, Bessent 3-3-3 target. Text: Source: ElA, JPMAM, September 2024.

(SPEECH)

Now, Scott Bessent-- one of the things that the Trump administration is saying is we're going to bring down energy costs and we're going to ramp up production.

One of the weird things about the Biden administration was, while they were very pro-renewables, they were not anti oil and gas. They were in sentiment but not in action. Oil, natural gas, oil, and gas production hit all-time highs under Biden. So yeah, he said bad things about them and there was this toothless LNG export ban and stuff like that. But broadly speaking, the energy production numbers thrived under Biden.

So there's not a ton of room for Trump to come in and say, watch me now, what is it, drill, baby, drill. They may end up increasing production by 5%. And let's say half of that gets exported. I don't think there's going to be any benefit on domestic energy-related inflation.

So and the other challenge is-- Mary, this is the non-toothless version of this chart.

(DESCRIPTION)

Text: Moving energy is just as important as finding it. A line chart is titled, US transmission line and pipeline growth, Miles added. The x-axis has years from 2013 to 2023, with intervals of 2. The y-axis goes from 0 to 4,000, with intervals of 500. Another y-axis on the right side of the graph is labeled, Project, and goes 0, 5, 1, 1, 2, 2, 3, 3. A blue line corresponds to the label, Transmission, which has a leftward arrow under it. A gold line corresponds to the label, Pipelines, which has a rightward arrow under it. The two lines intersect and overtake each other for most of the graph, between 1,500 and 4,000. Transmission peaks above 3,000 around 2018 and Pipelines peak at almost 3,500 around 2019. Then both lines slope steeply downward, with transmission ending at around 500, at the Project level 5, in 2023, and Pipelines ending near 0, at Project level 0. Text: Source: S and P Global, ElA, JPMAM, 2024.

(SPEECH)

The other challenge is, you can increase extraction activities, but if you can't move power, and if you can't move liquids, and if you can't move gas around, you're not going to get any benefits in the choke points of Massachusetts, and California, and places like that.

And let's see. I mean, the Trump administration is probably better positioned to change this dynamic than any administration over the last 30 years. But boy, are they up against it because it's a states' rights issue. A lot of these projects are blocked at the state level.

Is anybody here from Massachusetts? You have my condolences. So Massachusetts is running out of power. The ISO New England is talking about how they might have to be brownouts and power rationing in the next 10 years. You're a stone's throw from the Marcellus Shale, but Massachusetts won't permit the building of any natural gas pipelines. I get that.

But you know what else failed in Massachusetts? For 10 years, Massachusetts tried to get a hydroelectric project with Hydro-Québec going that would have required the construction of a high-voltage direct current line. But it had to go through your friends, the progressives, in New Hampshire, Vermont, and Maine. They blocked the project every time. After 10 years and about $300 million, the project died.

So and that's why Massachusetts is now spinning its wheels, trying to get offshore wind built, which is running into another set of problems. So we have so many states' rights issues and so little power in federal eminent domain that this is a tough one to solve. I'll be interested to see whether the Trump administration can do anything about that.

(DESCRIPTION)

Text: A very gradual and linear transition. A line chart is named, Decarbonization is a linear industrial transition, Renewable share of final energy consumption. The x-axis gas from the year 1990 to 2025, with intervals of 5. The y-axis goes from 2% to 16%, with intervals of 2%. Four colored lines on the graph correspond to Europe, China, US, and Asia ex-Japan/China. Europe sits above all the other lines for vast majority of the graph, starting around 4% and ending around 15%. The other lines also show an upward trend, each starting below 5% and climbing. China ends just above 10%, the US ends around 8%, and Asia ends just above 6%. Text: Source: El Statistical Review of World Energy, JPMAM, 2024.

(SPEECH)

And just as a-- so you're from Massachusetts. Have you been to Harvard Square? There's a bar called the Hong Kong Bar, and you can get this thing called a Scorpion Bowl. And it's-- oh, you're nodding. You've had a lot of them.

[LAUGHTER]

So it's good that you're sitting in the front for auditory processing.

[LAUGHTER]

So a Scorpion Bowl has rum, and gin, and tequila, and everything in it. This chart has everything in it, wind, solar, batteries, EVs, deep geothermal storage, replacing peaker plants. Everything you've ever read about in the renewable transition is in this chart. And the share, the renewable share of final energy consumption, is growing about 0.5% a year. This is a long journey.

And the thing that annoys me the most when I read buy-side and sell-side research on the renewable transition is when they describe it as having this S-shaped adoption curve that you see in the tech sector, because it's just not. This is a linear industrial transition that's going to take a long time. This is not a kind of accelerating geometric adoption curve the way that you see in Silicon Valley. And your policy decisions have to reflect these realities.

(DESCRIPTION)

Text: DOGE Quixote: the billionaires quest to reduce US government spending. An illustration shows two men on horseback headed the same direction, one hunching and charging in front of the other while holding a long weapon. Text: "Don Quixote and Sancho Panza", Honoré Daumier, 1865-1867.

(SPEECH)

DOGE, Quixote, good luck to them. I hope they succeed. I mean, every American, right, left, center, wants the government to stop wasting money, but it's been tried before. And the question is, can they do this without violating too many constitutional norms.

There's a bunch of things the administration is doing that people immediately assume is unconstitutional that aren't. I spent the day yesterday making this giant table of all the things that Trump has done so far, and putting them into two buckets. Number one, maybe unwise, maybe unpopular, maybe unprecedented, maybe unpleasant, but within the broad powers given to the executive branch by the Constitution and by the Congress.

That's a long list and includes a lot of the things that people assume are unconstitutional. It's a much shorter list of things that are almost certainly headed for Supreme Court constitutional adjudication, ending birthright citizenship and stuff like that. But even something like providing a bunch of teenagers with access to the federal payment system falls under the discretion of the executive branch unless they're violating things related to security clearances. But a lot of the data they've been sharing don't require those.

So is it unwise? Yeah. Has it ever been done before? No. Is it clearly unconstitutional? No, it isn't. The GAO was set up by Congress to improve government efficiency, but Congress never said that's the only entity that can do it. And it's fully within the president's discretion to set up an alternative commission approach to trying to figure out how to cut spending.

So we'll see. I mean, I don't want to go through

(DESCRIPTION)

Text: Vava con Dios. DOGE Quixote! There are savings to be had from reversing Biden executive actions. • Non-defense discretionary spending has already been cut, and is at its lowest level in decades. Entitlements are a political third rail. SEC, DOL, Education and EPA staff is very small at approximately 1% of all Federal workers. Expanding weight loss drug coverage may offset DOGE savings elsewhere. Impounding or redirecting Congressionally appropriated spending will face judicial challenges. Most defense spending is people and operations rather than weapons procurements. Concentration of defense contractors makes it difficult to cut procurement costs. Reducing improper/fraudulent payments will take substantial investment in systems and people.

(SPEECH)

all of this. I want to draw your attention to stuff like the fourth bullet. If you add up all the people at the SEC, the Department of Labor, the Department of Education, and the EPA, it's like 1% of federal workers.

And I have a rule in terms of people that I take seriously in life and people I don't. I don't take people seriously when they spend a lot of time jumping up and down and barking about stuff that ends up being 1% of the thing they say they're talking about. So right now, I don't take these people seriously. Maybe that'll change.

(DESCRIPTION)

Tex: DOGE. A line graph is called, Defense contracts 2020 US dollars, billions. The x-axis goes from the year 2001 to 2019 with intervals of 2. The y-axis goes from $75 to $300. Two lines on the graph, Competitive and Non-competitive, both slope up until about 2008, then dip into a valley around 2015, then climb for the rest of the graph, neither reaching their former peak. The competitive line sits above the non-competitive line for the duration of the graph, reaching a much higher peak. Competitive starts at $125 and ends just above $200. Non-competitive starts at $75 and ends just shy of the Competitive line above $200. Text: Source: Reilly (George Mason University), 2023.

(SPEECH)

This is-- I'm hoping they do something about this. Imagine you're running a company or a household, and you find out that 50% of all the expenditures that you're doing are not competitive. You're just, they're non-competitive, fixed priced, no risk. And that's what happens within weapons procurement. That's kind of amazing. I think, obviously, that's something that people should look at.

(DESCRIPTION)

Text: Reconciliation bill. A segmented bar chart is called, What might a 2025 reconciliation bill look like? US dollars, trillions. The y-axis goes from $0 to $5 with intervals of 1. A bar labeled, Deficit increase, has a large blue area from 0 to 4$ labeled, Extend individual tax cuts, and a thin gold area from 4 to just below $5 labeled, Extend corporate tax cuts. A bar labeled, Deficit reduction, has a gray area from 0 to around $1 labeled Possible spending cuts, a thin red area from 1 to around 1.5 labeled, Reduce IRA/infra tax credits, and a dark blue area from 1.5 to around $2 labeled, Dynamic scoring: tax revenue. Text: Source: Piper Sandler, JPMAM, 2025.

(SPEECH)

So this is the outline of the reconciliation bill. It costs somewhere between $4 and $5 trillion relative to the CBO baseline to extend all the tax cuts. And that doesn't even include raising the SALT cap and stuff like that. The administration doesn't have to offset all of that in order to get a reconciliation bill passed. They just have to offset enough of it so that the fiscal hawks in both parties will support it. Remember, Trump has an advantage of three in the House. So they have to bring a lot of people along for the ride.

And right now, it looks like they're going to try to grab the tax credits from the infrastructure and the energy bill, which is very disruptive, but they need money. A bunch of spending cuts we've listed on the right, getting some benefit from the CBO in terms of dynamic scoring revenue, and that's what this reconciliation bill looks like so far. But there's a long way to go, and there's going to be a lot of horse trading.

And anybody here from Texas? Chip Roy is interesting. He's a true believer in fiscal conservatism. And he's going to attack a Trump reconciliation bill from the right if he feels like it's not offsetting enough of the costs. So there's a really, really intense balancing act here that has to happen.

(DESCRIPTION)

Text: Deregulation. A bar chart is titled, Total cost of final rules, US dollars, billions. The y-axis goes from 0 to $1,800. A blue bar labeled Obama extends to around $500. A red bar labeled Trump reaches around $100. Another blue bar labeled Biden reaches $1,800. Text: Source: Doug Holtz Eakin, American Action Forum, January 17, 2025.

(SPEECH)

And here's just one of the deregulation charts that I mentioned earlier. The first Trump administration was like an oasis of low regulatory intensity compared to what happened around it. And that Biden number is just kind of amazing. We were fined for having this slide under Biden.

[LAUGHTER]

So anyway, let me just wrap up here.

(DESCRIPTION)

Text: Bond markets are nervous, so far. Supply side growth benefits, deregulation, increased capital spending, productivity gains, lower energy prices, M and A activity, versus, Shrinking labor supply, tariffs, risk of inflation, large budget deficits, precautionary increase in household savings due to govt cost-cutting / OMB directives. A line graph is called, 10 year Treasury yield change after the first Fed cut. The x-axis is labeled, Business days since first Fed rate cut, and goes from 0 to 100 with intervals of 10. The y-axis goes from negative 0.4% to 1.2% with intervals of 0.2%. A horizontal line stretches across the graph at 0%. A gold line, Average of six prior easing cycles, hovers around 0 and negative 0.2%. A blue line, Sep '14, rises steadily toward 0.8%, dipping briefly around 60 on the x-axis to around 0.5%, then rising again, ending the graph around 0.9% at 100 on the x-axis. Text: Source: Bloomberg, JPMAM, January 30, 2025.

(SPEECH)

This is-- because this is really the big issue. The 10-year Treasury is really the foundation of global markets. And I collapsed all of those other easing cycles into one line so you can see the average. The verdict so far is nervous in the bond markets.

And so, this is the battle for investors. Do we get those supply-side growth benefits, deregulation, capital spending, productivity gains, lower energy prices, a boost in M&A activity after changes at the FTC? I believe all of that. But we're also going to get the stuff on the bottom, a shrinking labor supply, tariffs of some kind, risks of inflation, budget deficits.

And then the big issue is, the way this all translates is you're going to have households increasing their precautionary savings because of all the uncertainty. And so, that's the battle that's being fought here. And that's why I'm a little nervous how this plays out.

We'll know a lot in a month. We'll see the full list of who gets confirmed in the Senate and who doesn't. And we'll get the outlines of a reconciliation bill and what the priorities are, and then we'll know a lot more.

But I think, after two 20%-plus years, I think it pays to have some patience here with respect to holding on to some liquidity and seeing how this plays out. So thank you very much for coming to the event and enjoy the rest of the day.

[APPLAUSE]

(DESCRIPTION)

Michael waves and departs.

Logo: J.P.Morgan Private Bank.