Semiquincententacles

(DESCRIPTION)

Text, JP Morgan, Eye on the Market. June 2026, Semiquincententacles. The US grip on markets on the 250th anniversary of the Declaration of Independence. An eagle with tentacles wrapping around the globe above a cloudy blue sky.

(SPEECH)

Good afternoon, everybody. This is Michael Cembalest with the June Eye on the Market podcast. I wanted to do something special this time in commemoration of the 250th anniversary of the Declaration of Independence.

So I put together a long-form piece called Semiquinten-- let me get this right. Semiquincententacles. The

(DESCRIPTION)

Text, Behold the Aquilaceph, half-bald eagle and half-octopus On the semiquincentennial 250th anniversary of the US Declaration of Independence, this imaginary beast is a metaphor for the continued US grip on financial markets. In this special issue we look at the details: US reserve currency status, capital flows, the much anticipated but still unprofitable "Sell America" trade, US corporate profitability and productivity in the age of Al, investing in Security & Resilience, equity market concentration, energy independence and the revival of the US IPO market. The biggest medium-term concerns for investors in US assets, other than the sustainability of the US Federal debt and cyclical inflationary pressures: the increased unpredictability in the rule of law, and government defunding of science and sidelining of scientific expertise.

(SPEECH)

semiquincentennial is the 250th anniversary of the United States, and the image that we created for this Eye on the Market is an Aquilaceph, which is a name I made up. But it refers to this beast that you can see here, that is a combination of a bald eagle and an octopus. And it's a metaphor for the tight grip that the US, on its 250th anniversary, still has on global markets.

And so in this issue and in this podcast, we look at the details. I'm doing a webcast that will go through more details. This is going to be a more abbreviated version in this podcast of the materials. And, of course, to get the whole shebang, you got to look at the actual PDF itself.

And I'm going to cover issues like the US Reserve currency status, capital flows, this much anticipated but so far on profitable sell America trade, corporate profitability and productivity at a time of AI, investing in security and resilience, equity market concentration issues, energy independence, and then the revival of the IPO market.

And we also take a look at some of the medium-term risks that I'm concerned about for investors, which is the increased unpredictability in the rule of law and government defunding and sidelining of scientific expertise. So again, this is an abbreviated version. Let's get into it.

Let's

(DESCRIPTION)

A chart of government budget deficits as a share of GDP from 2006 2026 and projected into the near-future for Canada, Italy, the UK, Japan, Germany, France, the US, and China shows big negative share dips in the 2007 financial crisis and 2020 pandemic to under 12.5%, with respective gradual and sharp rebounds to less negative values.

(SPEECH)

start with Reserve currency status. The usual suspects on this are, the US has a large budget deficit. That's true. Although note that China's, at least the way the IMF computes it, is even worse.

(DESCRIPTION)

Another chart shows general government gross debt as a share of GDP over the same period. China shows a notable rate of rise compared to the others, with Japan’s in decline from a highly elevated relative level.

(SPEECH)

Rising government debt-- the US has a debt ratio that's basically doubled from 60% to 120% of GDP over the last 20 years. Again, not that different from China.

And

(DESCRIPTION)

A third chart shows total household, nonfinancial corporate and government debt as a percent of GDP, with varying trends. The US and Germany are relatively neutral, versus China, which rises throughout the whole period. And all show marked bumps at the pandemic in 2020, with declines since then with the exception of China.

(SPEECH)

then overall, total economy-wide debt, which is usually looked at as household, non-financial corporate, and government debt together, here the US has been kind of flat. Because what was happening was that for a while the government debt was offsetting declines in household and corporate debt. And here China's much worse.

(DESCRIPTION)

A chart of sanctions, specially designated nationals and blocked persons (SDN) list designations and entity list additions, by number per year from 207 to 2025. There’s a tripling in SDN list designations from 2022 to 2024 up to 3,000, before a drop in half into 2025. A similar increase in entity list additions, but at much lower numbers under 1,000.

(SPEECH)

And then the other thing that dollar doomsayers typically point to is the increase in US sanctions applied to different entities and individuals. The idea being more and more sanctions would make people less interested in holding dollar denominated assets wherever they were because they could be seized.

(DESCRIPTION)

A chart of the US dollar trade weighted exchange rate from 2009 to 2026 shows a volatile stepwise rise from 75 to 85 to a peak over 110,, with a downturn to under 100 since Trump’s 2025 inauguration.

(SPEECH)

So despite all of this, apparently the dollar can do everything but read. Because the dollar has been remarkably stable, ignoring all of these negative fundamentals. The dollar took a 10% hit when Trump was elected, but has been pretty stable since then.

And I'm not really surprised.

(DESCRIPTION)

A table titled Reserve Currency Tracker: US dollar share of debt, FX, reserve, and trade. Cross-border loans, international debt securities, FX transaction volume, official FX reserves, exports invoicing, and SWIFT payments are broken down by years 2020 to 2026 in percentages, with the latest as of date, source, and next release date to the month or quarter provided. Most of the precentage data remains stable ar around 50% for every year, with exports invoicing consistently lower at 40%, and SWIFT payments ranging from 39% in 2020 to 51% in 2026. FX transaction volume is missing for all years except 2022 at 88% and 2025 at 89%.

(SPEECH)

We have this Reserve Currency Tracker. I would expect the dollar to come under pressure if any of these metrics or some combination of them started to decline. And what we looked at is the dollar share of a bunch of things that are going on in the global economy like the dollar share of cross-border loans and the dollar share of international debt securities denominations, and the dollar share of FX transaction volumes, of FX Reserve investments by central banks, of exports invoicing or of SWIFT payments.

And as you can see in our Reserve Currency Tracker table that we're showing, through the end of 2025 all of these metrics were pretty stable and none of them really experiencing much of a way of substantial declines. So as long as the widgets driving the usage of the dollar as the world's reserve currency aren't really changing, I differ from some of the dollar doomsayers on this issue.

(DESCRIPTION)

Text, Not much movement elsewhere. A chart of select currency shares of global FX reserves since 2020 in percentages. The Swiss Franc near 0%, Chinese yuan, slightly above to under 2%, British pound just under 5%, euroo just under 6%, and Japense Yen just under 6%.

(SPEECH)

And the dollar has gone down in terms of its share of global FX reserves by like 3%. But what's interesting is, that hasn't been ground that has been gained by the yen, the pound, the Chinese currency, or the Swiss franc.

As a matter of fact, all of those numbers are lower than they were two, three years ago. What

(DESCRIPTION)

A chart of other currencies as share of global FX reserves rising from under 3% in 2018 to 6% in 2025.

(SPEECH)

has gone up is this other category, which is a cats and dogs collection of Singapore dollar, Swedish krona, Korean Yuan, Norway, and other currencies where the IMF doesn't break out the individual numbers.

(DESCRIPTION)

A chart of gold shares of world FX reserves from 2009 to 2025. The actual shows a sharp rise from 2019 to 30%. And assuming market prices and no change in gold holdings since March 2009 matches that trend very closely, while assuming actual holdings and no change in gold price since March 2009 trends neutral to slightly down.

(SPEECH)

Then what about this issue that, oh, central banks have more and more and more of their reserves in gold? That's a price issue, not an allocation issue. As we wrote last year, if you hold market prices the same, and you look at the actual gold holdings, gold allocations have gone down over the last 15 years or so. If we don't change allocations, but we just look at the price change, that's basically the entire thing.

So what's happening is the central banks are getting a markup benefit on their existing holdings. But those allocations aren't really changing much in troy ounce terms.

(DESCRIPTION)

Text, China is larger than the US, but size is not all that matters. The US economy was 3.5 to 5 times larger than the UK when the US dollar replaced the pound as the world’s reserve currency. A chart of US real GDP divided by UK real GDP from 1850 to 1960 showing a rising trend, with the 1929 Eichengreen paper marked just before a backslide from 3.5 times to 2.5 times, and then the 1945 standard historical assessment at a local peak of nearly 5 times UK GDP. Dollar overtakes British pound as dominant currency. Another chart asks how much Chinese liquidity would leave if it could? Showing M2 money supply as a percent of GDP for various countries, with China at the highest with nearly 250%, overtaking Japan at 200%, and well above most others between 25 and 125%.

(SPEECH)

OK. And China, as the world's reserve currency is frankly an absurd concept. We write about it. It has the highest money supply as percentage of GDP in the world.

If they ever decided to open the capital account, that giant sucking sound, as Ross Perot used to mention that you'd would hear of renminbi rushing out of the country could clobber both their equity markets and their real estate markets.

(DESCRIPTION)

Text, One of these things is not like the other. Three charts show pegged currencies from 2009 to 2025 with M2 money supply, central bank assets, and FX reserves all rising in tandem on a US dollar index. A second chart shows Mexico (floating currency) from 2000 to 2025 with a similar upward trend across the three types. And then a chart of China (controlled/ restricted) with only M2 money supply rising over 600, but central bank assets and FX reserves stagnating under 250 on the index form 2009 to 2025.

(SPEECH)

And we get into some research showing just how different China looks from a macroeconomic perspective than other pegged currencies and floating currencies like the Mexican peso, for example. China's system is basically totally inconsistent with the concept of being world's reserve currency.

OK. So let's talk about capital flows.

(DESCRIPTION)

A chart shows US net debtor nation measures as percent of GDP with current account and net international investment position both declining from positive in 1982 to deeply negative in 2024.

(SPEECH)

The US for sure is a debtor nation and needs a lot of money from abroad. The two ways you can look at that are the current account deficit and also the net international investment position. Both of those things are telling you that the US needs a lot of money.

The good news, as far as we can tell, it's still rolling in.

(DESCRIPTION)

Four charts show foreign holdings share of US treasuries declining from 50% to 25% from 2012 to 2024, foreign holdings of US treasuries in absolute dollars, rising over the same period, foreign holdings of US long term non sovereign bonds rising, and foreign holdings of US equities normalized by S&P returns volatile but slightly higher over the same period.

(SPEECH)

And there's a lot of hand-wringing about the falling foreign share of US treasuries. And it's true that foreign holdings of treasuries, excluding T-bills, have gone from 45% to 25% just over the last decade or so. But that's because the debt's growing faster than foreign holdings.

If you look at the upper right here, foreign holdings of treasuries and T-bills are actually going up. They're just not going up as fast as the debt. And when we look at foreign holdings of corporate bonds and of equities, those are still rising. So we don't see any evidence of a buyer's strike on US assets.

And as it relates to this kind of sell America trade, it's very anticipated and rarely profitable.

(DESCRIPTION)

A chart of “Sell America” instances since 1970 showing S&P minus 5%, US trade weighted dollar minus 5%, US equities vs. ROW minus 5%, 10 year treasury yield plus 10 basis points, over a 30 day period. Marks are made in 1970, ‘74, ‘79, ‘82, and 2025.

(SPEECH)

So last year when Trump was elected, shortly afterwards, maybe two months later, there was this instance where the dollar went down on a given day by 5%, US equities underperformed the rest of the world, the Treasury bonds sold off, the 10-year treasury by 10 basis points. And this, by the way, is over a 30-day And it happened for the first time since the '70s and early 1980s. And so there was a lot of hand-wringing about this. I just didn't think that this was going to build on itself and it hasn't.

Right when this happened, I went on a Scott Galloway podcast, which I partially regret. Because before those podcasts begin, he goes on these bizarre r-rated rants that are kind of better suited for his midlife therapy sessions than a markets discussion.

But when we did get into the meat of the discussion, he was pushing pretty hard on the sell America trade. And all I can tell you is, ever since that podcast took place, Treasury yields have only gone up 10 basis points, US equities have outperformed the rest of the world, and the dollar is down by 1%.

So far, sell America, I'm not seeing it.

(DESCRIPTION)

A chart shows the discount for non-US stocks. MSCI world ex-US forward PE divided by US and PE, with a volatile downward trend from 130% in 2005 to a low of 60% in the 2020s and around 75% now.

(SPEECH)

And the longer story here is that this chart on the P/E discount to buy non-US stocks compared to US stocks has been the equivalent of a bug zapper for investors. And for the better part of 15 years, people kept diving in, diving in, diving in, buying new non-US stocks over US stocks and it hasn't worked out well. And

(DESCRIPTION)

A table of US equities vs. rest of the world from 1984 to 2022 showing outperformance every year of US stocks, with the US excess annualized returns, and decline in US excess returns since December 2024.

(SPEECH)

we have the full tally going all the way back to 1984. Depending upon when you invested, any long-term investor would just be much, much better off.

So let's use one example. If you start investing in 2012, US equities outperformed the rest of the world by almost 10% annually through the end of December '24. Now there was a catch up in non-US equities in 2025. But still, using that same 2012 starting point, the US excess returns annually are 8% instead of, like, 9.5%.

So even with the catch up that took place in non-US equities in 2025, for long-term investors going back several decades and all the way through recently, US equities have done much better.

(DESCRIPTION)

Text, Yes, US tech margins have exploded vs. other sectors. A chart of S&P 500 free cash flow margins: tech and interactive media vs. the rest in percent. Tech takes off from under 10% to 25% between 2002 and 2022, while the rest of the S&P, excluding financials and REITs grows only a couple percent.

(SPEECH)

And while tech margins have exploded versus other sectors, in other words tech margins have expanded a ton, they've gone from 5% to 25% since the early '90s, the US market profitability advantage versus the rest of the world is not just a technology story.

So

(DESCRIPTION)

Text, But the US is more profitable than other regions across all major sectors and most minor sectors as well. Four tables compare major and minor sectors returns on assets and returns on equities for the US, Europe, Japan, and China. The US leads in all major sector returns on assets, except for Europe in healthcare, and in minor sectors it leads in real estate and energy, but lags Japan in utilities, and China and Europe in materials. It leads all major sectors in returns on equities, and real estate and energy in minor sectors, but lags the Euro in utilities and China in materials.

(SPEECH)

again, this is one of my other favorite pages. We look here at the major sectors and at the minor sectors, and we look at the US, Europe, Japan, China. And for the major sectors, the US basically has higher return on assets and higher return on equity across the board. And for most of the minor sectors, which is utilities, materials, real estate and energy, they also have higher ROA and ROE. So this is not just a valuation story that's been driving the US out-performance. Its actual underlying profitability. So enough about that.

(DESCRIPTION)

Text, Investor exuberance: plenty of signs. Four charts show the moving average of the Philadelphia semiconductor index with big spikes from a baseline in 2000 and recently. Another chart shows hedge fund next exposure to semis and hardware, with a huge increase from under 15% in 2025 to 35% in 2026. Another chart shows Korean margin loans outstanding in trillions of won, with a bump up in the pandemic, a consolidation, and then another spike higher in 2026 to 40 trillion. The fourth chart shows the average daily semiconductor options premium traded by month, indexed to the 2020 average, with a gradual rise overtime, and punctuated by big spikes in 2024 and 2026 up to 5 times the average.

(SPEECH)

Now there are signs-- look, there are plenty of signs of exuberance. I had an acid flashback recently when I was looking at some technicals on the Philadelphia Semiconductor Index. The last time I saw the 200-day moving average chart looked like this was sometime in the year 2000. Hedge fund exposures to semis and hardware is flying off the page. Korean margin loans, which always gives me a nightmare when that goes up a lot, and then just retail options trading in semiconductors is skyrocketing. So there's plenty of investor exuberance right now. That's to be sure.

(DESCRIPTION)

Text, 4. Concentration: market cap, earnings, and belief.

(SPEECH)

I wanted to talk also in this piece about the issue of concentration because there's a lot of concern about concentration.

(DESCRIPTION)

Text, Equity market concentration. Three charts. The first shows that the US is actually at the lower end of the top 10 company share of country equity index market capitalization at only 35%, while the majority of countries are higher, going up towards 100% with only Japan and India less concentrated. Another chart shows MSCI all country world index region weights with the US rising from 30% to near 60% from 1988 to 2024, with Europe and Japan getting squeezed out, and emerging markets growing slightly along with other. A third chart shows the top 10 company share of S&P market cap and earnings, rising from under 25% in 2015 to over 35% by 2025.

(SPEECH)

And what do people mean by that? Well, they mean different things. The concentration that I'm not that worried about is the fact that the top 10 stocks in the United States now account for around 30% to 40% of the market cap. And it was half that for most of the 1990s. So we now have a much more concentrated equity market. And by the way, the earnings of the top 10 companies are also very concentrated, about 35% of all earnings. And the US is now 60% of global market cap, up from 30% at the end of the 1980s.

Now first, amazingly, that 30% to 40% concentration of the top 10 stocks, the US has the third lowest number in the world. Only India and Japan are better. And we have a chart in here showing that every other equity market, you pick it, no matter how big or small, actually has greater market cap concentration of the top 10 stocks. Hard to believe but true.

(DESCRIPTION)

A bar chart shows AI accelerator revenue by company in billions of dollars for Nvidia, AMD, custom ASICs (Google, AWS, Meta, Microsoft), Intel, and other, with overall growth under $25 billion in 2022 to $200 billion in 2026, with Nvidia making up the overwhelming majority and custom ASICs growing in share vs. the others. Text, The unit economics of Al increasingly favor ASICs (Google TPUS, AWS Trainium, Microsoft Maia, Meta MTIA) over NVIDIA GPUs. The hyperscalers using their own in-house chips report total cost of ownership reductions of 30% - 40% vs merchant GPU fleets given NVIDIA's 70% gross margins. Anthropic's commitment to run Claude on AWS Trainium for the next decade is the strongest third-party endorsement of ASICs to date.

(SPEECH)

The other concentrated issues that I'm more worried about have to do with things related to AI. One example is, NVIDIA has a very, very concentrated share of the accelerator market, which refers to the customized processors that are used for AI. But that share looks like it's shrinking, and I'm not sure that investors are paying enough attention to this.

And there was a research report from JP Morgan last week. And buried inside of all the discussion about hyperscaler debt financing was some research showing that the unit economics of AI are increasingly favoring companies like Google, Amazon, Microsoft, and Meta using their own chips rather than NVIDIA chips, and getting a 30% to 40% total cost of ownership reduction for doing it.

And so that's why this chart here shows some projections of the overall accelerator market, which is referring to both GPUs and the ASICs, which are all of the other PUs that NVIDIA's share of-- the NVIDIA revenues are rising, but its share of the total is falling. And I'm not sure that people are paying enough attention to this. The fact that NVIDIA has 70% gross margins has certainly made it tempting for those other hyperscalers to develop their own processors.

(DESCRIPTION)

Text, Frontier lab revenue is growing at a very rapid clip. A chart shows Anthropic and Open AI annualized revenue in billions of dollars from January 2023 to July 2026. Both are tracking at an exponential rate, with Anthropic taking off significantly steeply and quickly in early 2026, overtaking Open AI.

(SPEECH)

And then there's the issue of the frontier labs. So the frontier lab revenues like Anthropic and OpenAI are going geometric at this point. That's clear. They are very profitable, reportedly between 40% to 45% gross margins, unsurprisingly. They're not super labor intensive businesses.

But the questions revolve around their operating margins, right, net of capital spending, essentially, depreciation, things like that. And that part is less clear because these revenues that are growing can't grow unless they continue to acquire more and more computing resources to support all the modeling that people are doing and all of these agentic AI programs.

And so there is a stratospheric revenue growth here, but I'm still grappling in the dark to understand the issues about their eventual operating margins, which we'll learn more about when and if they decide to go public.

(DESCRIPTION)

Text, Gross margins vs. operating margins. Two charts show estimated compute commitments as of June 2026 in gigawatts, with Open AI leading with 18 between Stargate, AMD, and AWS, and Anthropic trailing at under 15 between AWS, Google and Broadcom, Google, and a sliver of Space X. A second chart shows cloud provider revenue backlogs in billions of dollars for Microsoft, Oracle, Google, and Amazon, each broken down by Open AI, Anthropic, and other spending commitments. Microsoft has $280 billion from Open AI, $30 billion from Anthropic, and $317 billion from other. Oracle has $300 billion from Open AI, and $253 billion from other, none from Anthropic. Google has $200 billion from Anthropic, $268 billion from other, and none from Open AI. And Amazon has $138 billion from Open AI, $100 billion from Anthropic, and $226 billion from other.

(SPEECH)

We do have some data showing just in the last year, these are big numbers. 15 to 20 gigawatts of new compute resources for OpenAI and somewhere around 12 to 15 gigawatts for Anthropic. And just to put some numbers on it, one gigawatt is the equivalent of a very large, nuclear power plant, like the ones that were shut down in California or Indian Point in New York. One gigawatt has more or less been the standard size of nuclear reactor for the last 30, 40 years.

And when you look at the cloud revenue, the cloud provider companies, the big hyperscalers, around half of their entire revenue backlog is just OpenAI and Anthropic. And so the concentrated risks related to these frontier labs is real and something we have to keep looking at, particularly since now the token prices are going up, some enterprises may seek to avoid paying them by migrating to cheaper models, whether they're cheaper US models or cheaper Chinese models.

(DESCRIPTION)

A table shows AI model Elo score leaderboard with Claude Fable 5, Claude Opus 4.6, and Opus 4.7 int he top 3, followed by Gemini 3 Pro, Gemini 3.5 Flash, Claude Opus 4.8, Qwen 3.7 Max Preview, GPT 5.5, Gemini 3 Flash, Claude Sonnet 4.6, GPT 5.4, Qwen 3.5 Max Preview, Kimi K2.6, Qwen 3.6 Max Preview, Grok 4.1, and DeepSeek V4 Pro, with Chinese models highlighted. Source: LMArena Al Leaderboard, June 15, 2026 Best Chinese open-weight mid-sized models (Qwen, DeepSeek V4, Kimi) scored within a few dozen Elo points of closed frontier models and also cost 10x - 50x less per token The founder of Lindy Al, a productivity assistant chatbot company, announced that they moved their entire app Al service from Claude to DeepSeek, claiming savings of millions of dollars and improved performance Brian Armstrong from Coinbase tweeted that "80% of workloads will be running in 99% cheaper models" within a year, and mentioned Coinbase was actively moving workflows to cheaper models.

(SPEECH)

And let me just put some factoids in front of you. Bryon Armstrong from Coinbase-- I have mixed feelings about crypto, as you all know. But he tweeted the other day that 80% of workloads will be running with 99% cheaper models within a year.

The founder of Lindy AI, which is a productivity assistant chatbot company, announced that they had moved their entire AI process from Claude to DeepSeek and will be saving millions of dollars, and it's also going to improve performance.

And

(DESCRIPTION)

Text, Most tokens consumed in the future may not come from frontier models but from smaller open models that are up to the tasks at hand. A dot plot shows Artificial Analysis Intelligence index score by cost to run by model. Score on composite benchmark of 10 evaluations across math, science, coding, and reasoning. A green quadrant is highlighted in the upper left labeled maximizing intelligence per dollar with the Chinese models and Grok 4.3 concentrated there. Claude Opus 4.8 costs $3,700 to run the entire Artificial Analysis Intelligence Index task set for a score of 56, while DeepSeek V4 Pro (Max) scores at 44 for just $186, which is approximately 20 times cheaper

(SPEECH)

so what gets interesting is this chart that we have in here from a company called Artificial Intelligence, where they measure the cost of doing things and the effectiveness of the models that you're using. And there's this quadrant at the upper left of the chart for cheap models whose scores are kind of OK or just good enough, where you really have to ask yourself, for an extra bit of performance, is it worth spending 20 times more?

And there's one example here where we're looking at DeepSeek versus Claude Opus 4.8, where you're picking up a very marginal score improvement and paying 20 times more to do it. And so there is a lot of questions that this kind of thing raises as it relates to these open models and what they do.

(DESCRIPTION)

Text, For three of the last five quarters, tech equipment and software has contributed more to GDP than the rest of the economy combined. A chart shows quarterly contributions to real GDP growth from information processing equipment and software vs. the rest of the economy in billions of dollars between Q3 2021 and Q1 2026, with information spiking in 2025 to over $100 billion and then again in 2026, with subsequent troughs in the rest of the economy.

(SPEECH)

And then getting back to some of the economic stuff, for three of the last five quarters, tech equipment and software has contributed more to GDP than the rest of the economy combined. Which is why I've said for over a year now, the Trump administration is extremely fortunate that this AI boom is happening on their watch. Because without it, the economic data would more accurately reflect some of the growth dampening aspects of some of their policies.

And so, just take a look at the chart here that looks at the GDP contribution from information processing, equipment and software compared to everything else in the economy is pretty striking.

OK.

(DESCRIPTION)

A chart shows hyperscaler net debt to trailing 12 month EBITDA, multiple, including bonds, loans, and SPY triple net leases for Alphabet, close to 0, rising to just above for Q1 2027, Microsoft, slightly above 0, Meta, slightly higher with a boost up close to 1 times in 2027, Amazon similar, and IBM at 3 times, breaking the S&P 500 median of 2, and Oracle at 4 times, with similar amounts respectively for 2027.

(SPEECH)

So let me just keep going. Hyperscaler debt, everyone's concerned about it. We've written about it a lot. It's very manageable other than Oracle, and there's plenty of numbers in here to back that up. You can take a closer look.

We get into productivity gains.

(DESCRIPTION)

Two charts show growth in labor productivity per hour worked from 2010 to 2025 and total factor productivity growth for the same period with similar distribution across G10 economies and emerging markets, with productivity increases in emerging markets outperforming, except in Saudi Arabia and in Brazil on a total factor basis. Netherlands, Canada, Belgium, UK, France, and Italy also saw negative total factor productivity growth over the period, with US leading the G10 on both measures.

(SPEECH)

The US has the highest productivity gains over the last 15 years or so within the G10. That's pretty consistent. So it's, by the way, another supporting factor in terms of the US dollar remaining the world's reserve currency,

(DESCRIPTION)

A table shows US productivity gains pre-COVID vs. post GPT. Real GDP from Q1 2016 to Q1 2020, Q4 2022 to Q4 2025, and Q3 2023 to Q4 2025. Non-financial corporate was 1.1%, 2.9%, and 3.3% respectively. Information sector, 5%, 9.9%, and 7.8%. And data processing was 8%, 15.3%, and 14.4% respectively.

(SPEECH)

US productivity has actually picked up from pre-COVID levels TO post GPT levels. You have to cut the COVID thing out of your productivity calculations for obvious reasons. We expect to get another productivity report at the end of June that continues to point in that direction.

And then it's hard to talk about the US grip on global markets without getting into some of the issues related to China and Taiwan.

(DESCRIPTION)

Four charts on China innovation show China breaking into the top 10 most innovative economies on a global innovation index, rising to close a gap with the US between 2012 and 2026. Another chart shows export complexity, with China overtaking the US after 2015 and the US declining somewhat from a neutral trend. Another chart shows frontier language model intelligence for both countries rising geometrically. And a chart shows total cost of ownership (TCO) and per token in millions per milliwatt for Nvidia, Huawei, and Cambricon, with Nvidia about twice as much, up to $9 compared to the others. Cost per token is about in line, with Nvidia H200s costing the least, and one Chinese outlier Cambricon costing significantly more.

(SPEECH)

So I know I'm running a little bit long. Just bear with me.

China has been breaking into the top 10 in terms of the most innovative economies and export complexity, the capabilities of its frontier models compared to the US narrowing the gap, and then the cost of the models that they're providing to the world.

China innovation is moving pretty quickly.

(DESCRIPTION)

Text, Chip trade now larger than oil trade. A chart shows oil and semiconductor trade share of global GDP, with oil more volatile while semiconductors and integrated circuits rises gradually from 1990 to 2022 to near 2%.

(SPEECH)

And as we start to think about the US over the next decade, there's been an interesting convergence recently. Trade in semiconductors and integrated circuits is actually higher than oil trade. So while everybody is concerned about the Strait of Hormuz, as am I, I'm also concerned about the Taiwan Strait. Because at this point there looks to be more global GDP at risk from things that affect semiconductor trade than oil trade.

(DESCRIPTION)

A chart shows US AI related computer imports in billions of dollars, monthly annualized, including large compute, GPUs, and other parts from various countries, with Taiwan overwhelming in the lead at $200 billion, and Mexico in second under $150 billion, and all other sources under $50 billion.

(SPEECH)

And then when you look at AI-related related imports in the US, by far and away it's all about Taiwan.

(DESCRIPTION)

A bar chart shows US might reach 30 to 35% of advanced node (less than 5 nanometer production by 2028 to 2030. A stack of bars list various US domestic plant projects and their proportional production contribution.

(SPEECH)

And a lot of you have heard me talk about this before. The US is now scrambling to rebuild what it once had, which was more domestic semiconductor fabrication. And after running through the numbers, looking at the TSMC projects in Arizona and some Intel production in Arizona and Oregon, and some projects that may happen for Samsung in Texas, the US could, let's say, get to 30% to 35% semiconductor self-sufficiency by 2030.

But at the end of the day, when you're talking about these advanced node semiconductors, the US is still going to be highly reliant on Taiwan.

(DESCRIPTION)

A chart shows net imports of fossil fuels as a share of primary energy consumption by country, with Singapore, China Hong Kong in the lead with nearly 100%, and Taiwan in third at nearly 90%.

(SPEECH)

And that gets into this issue that Taiwan is, by far, in my opinion, the most blockadable country in the world. Forget about invadable. Blockadable is easy. And Taiwan imports 90% of its primary energy consumption as fossil fuels. So that puts them up there with Singapore and Cyprus and places like Morocco, where they are just highly, highly, highly dependent on importation of energy.

(DESCRIPTION)

Text, Chinese military assets are mostly focused on one thing. A chart shows share of Chinese assets deployed near the Taiwan Strait (in the PLA Eastern and Southern Theater Commands), as a percentage. 100% of amphibious assault ships, coastal patrols, and nuclear ballistic subs, over 80% of medium landing ships, 80% of Corvettes, around 70% of frigates and destroyers, over 60% of attack subs, 60% of special mission aircraft and bombers, around 50% of cruisers, 40% of air fighters, 30% of aircraft carriers and nuclear attack subs, and 10% of transport.

(SPEECH)

And China's military assets, if you look at where they're located, are mostly focused on one thing. So the vast majority of Chinese military assets are deployed near the Taiwan Strait, whether that's in the Eastern and Southern theater commands when you're talking about surface vessels, submarines, air force, you name it. So that's something that we have to keep paying attention to.

Now

(DESCRIPTION)

Text, 7. Security and resilience investing.

(SPEECH)

related to that issue is the US is desperately trying to rebuild certain industries. And there's a security and resilience initiative going on with inside JP Morgan that is focused on this. And

(DESCRIPTION)

Text, Industrial policy is changing as the US aims to catch up. A chart shows industrial policy spending in key economies, 2022, percent of GDP. China leads with over 1.5%, with South Korea a little over .5%, followed by France, Japan, Germany, Taiwan, the US, and Brazil under 0.5%.

(SPEECH)

so the US never really had much of an industrial policy, which is something that we've shown the chart on a lot in the past.

Now the US is looking to catch up.

(DESCRIPTION)

Text, Key Themes. A chart of security and resilience stocks by industry indexed to December 2024. Energy independence and resiliency has risen to 160, along with aerospace and defense, advanced manufacturing and supply chain, and a late catch up by frontier and strategic technologies. With pharma and healthcare resiliency treading under 100. The S&P market cap weighted index very closely tracks the frontier and strategic technologies.

(SPEECH)

And there are four or five main themes for investors to take advantage of this new industrial policy-- energy independence and resiliency, aerospace and defense, frontier and strategic technologies, advanced manufacturing and supply chains, and then to a lesser extent, pharma and health care resiliency.

The first four of those themes are doing quite well and have outperformed the equity markets last year and through today for all the obvious reasons, with the emphasis being put on repatriating a lot of these things and providing incentives for everything from shipbuilding to drones to aerospace, you name it.

(DESCRIPTION)

Text, Pure play defense contractors showing signs of supply chain stress. A chart of US defense prime contractor capacity constraints, book-to-bill ratio. Raytheon, Lockheed, and Northrup have risen from 1.5 to 2 to up to 3 times capacity between 2010 and 2025, while Honeywell and Caterpillar are close to at capacity after a later rise.

(SPEECH)

And the other thing that's interesting too is the new policies are putting a lot of emphasis on regular commercial contractors instead of the pure play defense contractors, in part because the pure play defense contractors are showing a lot of signs of supply chain stress in terms of their book to bill ratios, which is one way of looking at how much they're billing versus how much they're booking.

And

(DESCRIPTION)

Text, Security and resilience: commercial-first approach The 2026 National Defense Authorization Act (NDAA) enforces a commercial-first approach that requires research and formal documentation before opting out of commercial products, and reduces compliance burdens by limiting defense-unique clauses in contracts. Other similar pro-commercial efforts: • The DoD is increasingly using non-traditional acquisition pathways like Commercial Solutions Openings and Other Transaction Agreements, bypassing the bureaucratic Federal Acquisition Regulation framework The Modular Open Systems Architecture (MOSA) acquisition criteria requires new programs to maximize use of modular components that are more easily produced by commercial manufacturing processes On software, the FY26 NDAA creates more flexibility to pay contractors with a commercial-style subscription model, adding more flexibility to refine the goals of the product throughout the life of the contract The Joint All Domain Command and Control (JADC2) initiative seeks to connect sensors across land, sea, air, space and cyber domains into one network that automatically matches threats to the best available response, opening the door for new software, Al and modular hardware

(SPEECH)

so there's this new commercial first approach that's coming out of the National Defense Authorization Act that is designed to really broaden out the number of companies that are going to be participating in this Defense Initiative. You can read more about it in the piece.

(DESCRIPTION)

Text, I-Robot. A chart shows the amount of private capital invested in US robotics and drones in billions of dollars from 2015 to 2025, with a gradual growth staying under $10 billion until a big spike in 2025 up to $21 billion.

(SPEECH)

Lots to hear. We have a lot of information on investment and robotics and drones.

And

(DESCRIPTION)

A complex flow chart shows drone component technology origin countries on the left and countries of manufacture on the right, with the US dominating the technology origin and China dominating the manufacturing overwhelmingly.

(SPEECH)

it's going to take a lot of effort to reverse this image. This is a pretty terrifying image from the MIT Technology Review. Most drone technology, when you look at the origin of it, most of it was the US with little bits from Japan, the UK, and Europe. And then when you look at current country of manufacture, it's almost all China. And so it's going to take a lot of effort to reverse some of that.

(DESCRIPTION)

Text, 8. The rule of law.

(SPEECH)

I want to close with a couple of things, because this is important as well for investors. The rule of law issue gets talked about a lot.

(DESCRIPTION)

Text, Maybe unpopular, maybe unprecedented but not unconstitutional: repealing Biden Executive Orders imposing increased duties on certain foreign imports terminating federal employees not subject to federal civil service protections allowing states to use federal educational block grants for public school choice programs Restructuring/eliminating agencies not created by federal statutes and whose existence is not statutorily required abolishing DEl offices, trainings and programs in federal agencies temporarily suspending refugee admissions and temporarily barring admission to foreign nationals from countries with screening procedures determined to be insufficient from a national security perspective designating foreign drug cartels as foreign terrorist organizations pausing implementation of government programs (such as renewable energy subsidies) to develop new priorities, as long as funds are spent within the broad boundaries of the related congressional appropriation Withdrawing from organizations such as the WHO, particularly in the absence of Congressional legislation

(SPEECH)

Some of the things the Trump administration has done, some of them are unpopular. A lot of them are unprecedented, but that doesn't make them unconstitutional.

And we have a long list of things here, repealing Biden executive orders and duties on foreign imports and terminating federal employees not subject to certain protections, allowing states to change what the block grants get used for, suspending certain refugee admission admissions, withdrawing from the WHO. These things are not unconstitutional.

And I think sometimes people are sloppy about the things that they describe as being a deterioration in the rule of law. We list a bunch of things that are, again, unpopular, unprecedented, but not unconstitutional. And

(DESCRIPTION)

Text, Also not unconstitutional: redistricting. A chart shows new 2026 House seats that may result from state redistricting and Voting Rights Act decision. 16 potential House seats to Republicans and 6 to Democrats, with 7 of the Republican redistrictings subject to judicial change or voter referendum. The US Constitution and most state constitutions do not contain bans on gerrymandering for political purposes, while the Supreme Court has held that creation of majority-minority districts for racial purposes does violate the constitution.

(SPEECH)

also the Congressional redistricting issue that's going on, that is not unconstitutional either.

The US Constitution and most state constitutions do not contain bans on gerrymandering for political purposes. Now you may not like it, and if you don't like it, your legislature would be more than prepared to pass laws against it if people vote for that. But there are no constitutional and few state constitutional bans on gerrymandering for political purposes. And that's why these gerrymandering things are taking place.

The GOP may pick up anywhere from 12 to 16 seats for the midterms this year versus six for the Democrats. Some of that will probably reverse in 2028 when the Democratic states can catch up and figure out how to play the same game.

So I think this is not a good example of deterioration in the rule of law.

(DESCRIPTION)

Text, UCLA/ Bright Line survey of 21 Article III Federal Judges, 113 lawyers, and 193 law professors. A chart shows legal expert ratings of the rule of law, with their mean rule of law rating with a marked decline in all groups from 2014 to 2026.

(SPEECH)

There are other examples one could use, and UCLA did a survey of federal judges, law professors, and constitutional lawyers, and they do express a lot of issues. We have a chart in here that show a substantial deterioration in the rule of law since 2024.

(DESCRIPTION)

Text, Substantial concerns about constraints on Executive Branch power, compliance with court orders, politicized law enforcement and fear of reprisals UCLA, Bright Line survey of 21 Article III Federal judges, 113 lawyers and 193 law professors Retaliation. Nine in ten say Trump administration has used the DoJ to go after enemies and provide benefits to allies Presumption of good faith. Only one in five agree the federal government still merits the "presumption of regularity" (courts should presume government officials act lawfully and in good faith absent evidence to the contrary) Compliance with judiciary. Eight in ten report that federal officials fail to comply with court orders somewhat or very often, and nearly nine in ten say Trump DoJ appointees mislead federal judges "somewhat" or "very often". Legal representation and harassment. Nearly one in five lawyers report that representation decisions by their firms have often been affected by fear of adverse action by government officials or agencies. Almost half of federal judges are concerned about harassment if they rule against the federal government Core legal principles. Only 25% believe in the following principles: that US government agencies do not punish political opponents; that law enforcement is not exploited for political purposes; that investigations of public officials are not compromised; that government officials face sanctions for misconduct; and that government officials do not use public office for private gain.

(SPEECH)

And they talk about substantial concerns related to executive branch power, compliance with court orders, politicized law enforcement, fear of reprisals. And then we have some specific information here on retaliation, compliance with the judiciary, legal representation and harassment, abandonment of core legal principles. If you want to find it, you can find it and you don't have to look very far.

(DESCRIPTION)

A chart shows legislative checks on government, an index based on legislative/ government agency ability to question, investigate, and oversee the executive for various countries, with China way below others under .2, but marked declines in Hungary, India, and the US to around 0.6 since 2010. With Poland and Brazil showing dips, but recoveries during that time, and Germany remaining flat near 1.

(SPEECH)

And there's also something called the Varieties of Democracy project, hosted by the University of Gothenburg that shows a pretty substantial decline in what they describe as legislative and government ability to question, investigate the executive branch. The US has had the largest decline of all major countries in that index over the last couple of years.

(DESCRIPTION)

Text, 9. Federal government defunding of science.

(SPEECH)

More on all of this on the webcast or in the piece itself.

We also have a big piece on the federal section in here on the federal government defunding of science.

(DESCRIPTION)

Text, NIH and NIH grant declines. Four charts show percent declines for different subject areas in 2025 vs. the 2015 to 2024 average Declines in competitive NIH grants (aging, diabetes, cancer, strokes, heart disease, infectious diseases) and the NSf [technology. STEM disciplines, engineering. biology) We estimate that roughly half of 6,000 terminated NIH grants fell afoul of Executive Order 14151 and other directives which seek to terminate government activities related to LGBTQ+ health, vaccine hesitancy, environmental justice and DEl; the remainder were terminated for other reasons entirely.

(SPEECH)

I don't want to go through it. You know what's going on. There are very, very, very large declines in grants from the National Science Foundation and the National Institutes of Health.

(DESCRIPTION)

Text, Science related staffing. A chart shows a change in agency staff levels from previous year from 2016 to 2025 for the Centers for Disease Control and Prevention, the National Institute of Standards and Technology, National Oceanic and Atmospheric Administration, National Institutes of Health, National Science Foundation, Department of Energy, Food and Drug Administration, NASA, and the Environmental Protection Agency, all showing sharp 15 to 25% cuts in staff levels for 2025.

(SPEECH)

Massive cuts to agency staffing levels on anything related to science.

(DESCRIPTION)

Text, OMB proposal for the NIH and other agencies Political Appointees Take Control of Grant Awards (200.205) Peer Review Is No Longer Binding ($200.205(d)) Active Grants Can Be Terminated at Any Time, for Any Reason ($200.340) Applicants Can Be Denied Based on Organizational Affiliations (§200.206) Program Goals Must Align with Administration Policies and Priorities (§200.202) OMB Gains Direct Oversight of Which Institutions Receive Grants.

(SPEECH)

There's an OMB proposal that would do some pretty bad things to the entire scientific process within the federal government. And

(DESCRIPTION)

Text, Vaccines are arguably among the greatest achievements in biomedical science and public health. A table of vaccine preventable diseases in the US showing pre-vaccine vs. post vaccine cases and deaths, with 100 or near 100% case declines after vaccines.

(SPEECH)

then you have all the stuff that RFK is doing.

(DESCRIPTION)

A chart of life expectancy by year of birth showing a steady rise since the 1800s with various medical advances marked.

(SPEECH)

We have some charts here that go back to the Bronze Age in 1,200 BC in terms of life expectancy and the benefits of antiseptics, vaccines, and antibiotics.

(DESCRIPTION)

Text, RFK Jr. October 2025 OpEd entitled "Six surgeons general: It's our duty to warn the nation about RFK Jr was written by surgeons general serving under Bush I, Clinton, Bush II, Obama, Trump and Biden. Their concern: RFK's "profound, immediate and unprecedented threat to public health"..."'science and expertise take a back seat to ideology and misinformation* In June 2025, Kennedy fired all 17 members of the Advisory Committee on Immunization Practices FDA officials delayed or blocked publication of several studies supporting safety of widely used vaccines against Covid and shingles. The studies, conducted by scientists at the FDA that analyzed millions of patient records, found serious side effects to be very rare Scientists directed to withdraw Covid vaccine studies accepted by medical journals, including one by career CDC scientists that Covid vaccine sharply cut odds of hospitalizations and emergency room visits Aaron Kesselheim, Harvard University professor who studies FDA regulation: "at any other time in history, this would be a major scandal that would lead to congressional hearings and resignations of leadership.

(SPEECH)

I'm just going to leave you with this. Six surgeon generals appointed by Bush, Clinton, Obama, Trump himself, and Biden wrote an article recently called, It's Our Duty to Warn the Nation about RFK Jr. And they describe his profound, immediate and unprecedented threat to public health and how science and expertise are taking a back seat to ideology and misinformation. You can read more about it in the piece itself.

We have some stuff here on energy.

(DESCRIPTION)

Text, Appendix. IPO calendar, supply, demand, free float timelines, and performance. A chart showing creation of new public companies in the 21st century, by cumulative market cap in the trillions, color coded by economy, with Alphabet, Broadcom, Telsa, and Meta leading in market cap for the US, but China slightly outcompeting the US in pure numbers. A general trend upward for each country at a geometric growth rate, indicating more quantity and higher valuations over time, with European number of companies lagging the US and China by about half, with under 100. And Japan trailing significantly behind that.

(SPEECH)

I want to close just with some IPO stuff, because that's top of mind. I love this chart. It shows the creation of new public companies in the 21st century. It was drawn from a Draghi Report that's trying to boost European entrepreneurship. Good luck with that.

And as you can see, in the 21st century, the US market has really been the leader in value creation with new public companies. China is actually a distant second, and then distant behind them is Europe and Japan.

(DESCRIPTION)

A chart shows market cap of IPO companies as percent of US total market cap. IPO market cap at first close divided by US total market cap. There’s no trend, but an all time high spike near 2000 above 4%, matching where Space X, Open AI, and Anthropic are estimated.

(SPEECH)

There was a lot of hand-wringing for many years about the decline in IPOs, and some people thought it was Dodd-Frank, and some people thought it was changing tax policy that allowed companies to stay private longer with more shareholders.

The bottom line is, then we have the SPAC boom. That was regrettable. But now with some of these big mega IPOs, as a share of total market cap, we're getting close to some pretty high numbers, even going back to the late 1990s.

(DESCRIPTION)

Text, Enough demand? A chart shows US equity mutual fund cash balances, percent of total assets, with a general downtrend from 2009 to 2025 from 6 to 2%.

(SPEECH)

And so we have a section in here on IPOs that's interesting. There are questions about whether or not mutual funds have enough cash to handle all this stuff, because their cash balances are low.

(DESCRIPTION)

Text, Probably. A chart shows US equity net demand share of Russell 3000 market cap by gross buybacks less equity issuances and shares from expiring lockups from 1995 to 2005. Buybacks stayed under 0 until after 2000, and then ranged from 0 to 3%, and are at a trough near 0% but still positive. Text, M&A may be another source of demand for IPOs: US M&A activity has been resilient with approximately $900 billion of announced transactions Year to Date (plus 48% vs 2025), and cash has represented minus 70% of consideration for announced M&A this year.

(SPEECH)

If you look from a big picture perspective though, I think there is. Because the gross buybacks, less equity issuance and shares from expiring lockups is still a positive number. It's a smaller positive number, so there's less money sloshing around than there used to be, but it's still a positive number. Again, that looks at buybacks less what you need to absorb new and secondary issuance and also for the expiring lockups.

And remember, M&A may be another source of demand for these IPOs. M&A activity has been pretty resilient, up 50% versus last year. And cash is around 70% of what gets paid. So there should be plenty of money to absorb some of these large IPOs.

(DESCRIPTION)

Text, Free float share timelines. Two charts show the average free float of the 10 largest IPOs since 2010, free float share of publicly traded share class, starting at around 45% with a sharp increase up to 80% after 100 days. Another chart shows free float shares of publicly traded share class of the 10 largest IPOs since 2010, Meta, Air BNB, Coinbase, Salesforce, Palantir, DoorDash, Lyft, Rivian, Uber, and Snap, with a broad range, but a move up from 20 to 40% up to above 50% after 100 days for all IPOs.

(SPEECH)

The bigger question is not how they get absorbed when they get issued, but how do they do after they get issued? And JP Morgan equity research did a report recently where they looked at the free float share timelines for the 10 largest IPOs since 2010.

And the average-- now there was a very wide dispersion around it-- but the average one floated around 40% of the shares. But before the end of the year, 80% of the shares were public. And for some of them, they went from maybe 10% or 20% of the shares to 60% to 80% of the shares being public by the end of the year.

So the question is less about how does it trade during the first week, although that's of interest to people that participate in the syndicate and are going to flip their shares. For longer term investors the question is, how do they do after the first close, heading into the lockup expiry.

And that's particularly the case for a company like SpaceX.

(DESCRIPTION)

A chart shows the Space X share release timeline, percent of total equity market capitalization, with standard lockups building from late summer to December of 2026 to 1%, then extended lockups into 2027, and a huge block of Elon Musk lockups occurring in June 2027 up to above 2.5%.

(SPEECH)

As a share of total market capitalization for the whole equity market it's a very small number up front. But then you've got the standard lockups expiring, the extended lockups expiring, and then Musk's lockup expiring. And you could have, all of a sudden, 3% of the entire equity market capitalization get released over the next year or so.

(DESCRIPTION)

Text, What happens after the first day IPO pop? A chart shows the price performance relative to Nasdaq 100 index of the 10 largest IPOs since 2010, with a general downtrend after an early spike up to 0,8 relative to the Nasdaq’s 1.0 at 350 days after first close.

(SPEECH)

And so when we looked at what happens after the first IPO pop, the first day IPO pop at the largest IPOs since 2010, it doesn't look so good. They're, on average, a 20% decline heading into the lockup expiry. And then we have some additional-- and so we can look at that actually by company by company. We do that in here.

And

(DESCRIPTION)

A busier chart shows the same performance broken down for each major IPO. Text, Impact of lockup expiry. Another chart shows cumulative return around IPO lockup expiration, percent, US IPOs greater than $25 million in value since 2017 with the average IPO during the last 10 years dipping into the lockup expiration, and dipping even more so from as early as 180 days prior down to minus 10% at lockup expiration for the average IPO since 2023.

(SPEECH)

then just broadly speaking, when you look at the average IPO over the last 10 years and since 2023, you'd get 5% to 10% declines heading into the lockup expiry and then improving thereafter. So for people that didn't participate in the original syndicate, and you're trying to think, when should I build a position, there are some clear signals from history that you should wait until further in closer to the lockup period in order to start acquiring this stuff.

So

(DESCRIPTION)

Text, June 2026. Semiquincententacles. The US grip on markets on the 250th anniversary of the Declaration of Independence.

(SPEECH)

that is the abbreviated-- oh dear, 30 minutes. That is the abbreviated version of the semiquincentennial piece. See the printed piece for all the details, or listen to the webcast where I will not be speaking so quickly. Thank you for dialing in and I hope you guys are enjoying your summer. Goodbye.

(DESCRIPTION)

Text, JP Morgan.