I also see a potential head & shoulders top on US Treasury Yield 30 Years chart, and expecting the price action to confirm my bullish tone soon.

This stems from the financial system’s hurdle rate (coined by Stanley Druckenmiller, his full 2016 Sohn speech around this subject is below), which I believe is around 8% (preferred return) that drives the private, public, real estate, and every other imaginable market. Let’s assume most assets are ~50% to 100% levered (private equity and credit for sure are, most hedge funds run ~400% to 600% gross, and real estate is usually ~400% levered), and you could do the math but the conclusion is, borrowing <= 3% is fantastic, 3% to 5% is dangerous, and 5%+ is impossible unless your investments clear the 8% hurdle.

I asked ChatGPT “what are the bearish arguments for bonds again? Give me short bullet points and find historical parallels - do the narratives remind you the top around 2000 and 2007” - below is the response:

• Heavy Treasury issuance keeps forcing the clearing yield higher.

• Persistent deficits push term premium structurally upward.

• Inflation settles around 3% rather than returning cleanly to 2%.

• Oil/geopolitics create another inflation impulse.

• AI/data-center capex creates huge competing demand for capital.

• Japan offers attractive domestic yields, reducing Japanese demand for Treasuries.

• Fed stays tighter because growth/employment refuse to break.

• 5%+ long yields may simply be the new equilibrium rather than an extreme.

• Foreign buyers demand more compensation for duration and fiscal risk.

• Strong equities/credit tell the bond market that financial conditions still aren’t tight enough. Those are essentially the arguments being made in the current selloff.

Let’s run the exercise now, I will be giving examples from publicly traded banks, insurance companies, and private equity / private credit firms. In my opinion, almost none of them clear the return hurdle and that’s one of the reasons why I don’t think yields globally, at these levels, are not sustainable.

I won’t get into historical IRRs, and will be using ROE as the back of the envelope metric.

Here are the numbers (from worst to best) trailing 12-months:

They don’t look great, except for asset managers (private equity mostly) and insurance companies. Of course looking solely on ROE is not great due to its biggest blind spot; how much leverage was required to generate it.

If you look at the balance sheets of the firms with high ROE, you will see my point; AMP -1.27%↓ is 100% levered, BX -1.71%↓ is 150% levered, AON 2.51%↑ is 145% levered et al.

In 2008, years of cheap money and leverage inflated housing and credit until higher rates exposed the weakest balance sheets, housing rolled over, credit tightened, and the growth engine broke; in 2026, the economy is again absorbing unusually high long-term rates largely because AI investment, equity wealth and expectations of a productivity boom are keeping growth resilient despite tightening financial conditions, but if the AI productivity story is interrupted, things are not looking great, and that’s why the risk/reward makes sense to go long 30-year U.S. Treasuries.

Below is a cleaned-up transcript of Stanley Druckenmiller’s 2016 Sohn speech. Its core idea is simple: the risk-free rate is the hurdle rate for the financial system.

The 21st Sohn Investment Conference

Stanley Druckenmiller
May 4, 2016

When I founded Duquesne Capital in February 1981, the risk-free rate of return, the five-year Treasury yield, was 15%. Real interest rates were close to 5%.

That period became the starting point for one of the greatest bull markets in financial history. Asset prices were extraordinarily cheap because they had to compete with a 15% risk-free rate. At the same time, Volcker’s brutal monetary tightening forced badly needed structural reforms at both the macroeconomic and microeconomic levels.

It was no coincidence that soon afterward, the politically divided Tip O’Neill and Ronald Reagan came together to enact the last major reforms to Social Security and the tax system.

Moreover, a 15% hurdle rate forced corporations to invest their capital wisely and undertake their own structural reforms.

If that environment produced one of the greatest investment opportunities in history, how can its mirror image, the environment we have today, also be considered a great investment environment?

Every week, we hear people extolling the virtues of the stock market because “interest rates are zero and there is no alternative.” This view has become so widespread that it has even earned its own acronym: TINA, “There Is No Alternative.”

[Slide 3]

Not only were valuations low in 1981, but financial leverage was less than half what it is today.

There was therefore considerable room to generate growth through credit expansion. The policy response to the global financial crisis was so aggressive, and more importantly has remained so aggressive, that it prevented any meaningful deleveraging from taking place.

In fact, leverage has increased globally.

Ironically, from my perspective, that is exactly what most policymakers are deliberately trying to achieve today.

[Slide 4]

Let me focus on two major policies that have not only failed to address the excessive leverage accumulated in developed markets, but have also contributed to an explosion of leverage in emerging markets.

The first is the policy pursued by the Federal Reserve.

By almost any objective measure, we have experienced the longest period of excessively accommodative monetary policy in history.

During the Great Recession, interest rates were reduced to zero and the Federal Reserve expanded its balance sheet by $1.4 trillion. More importantly, after the Great Recession had ended, the Fed expanded its balance sheet by another $2.2 trillion.

Today, with unemployment below 5% and inflation approaching 2%, the Federal Reserve remains extraordinarily dovish.

If the Fed responded to economic data in a manner consistent with the average behavior of the Volcker and Greenspan Feds, the standard Taylor Rule would imply a federal funds rate of roughly 3%.

So, rather ironically, this is probably the least “data-dependent” Federal Reserve we have ever had.

In short, this is the largest and longest deviation from historical monetary-policy norms that I have witnessed during my career.

The Federal Reserve is borrowing more from future consumption than ever before.

Despite the relatively strong performance of the U.S. economy, the United States currently has the lowest real interest rates in the G7.

At the 2005 Ira Sohn Conference, facing a similar but far less extreme policy deviation, I argued that the Greenspan Federal Reserve was sowing the seeds of a historic housing bubble driven by reckless subprime lending, and that it would end badly.

Today’s policy excess is far greater, both in duration and magnitude.

The obsession with short-term stimulus stands in stark contrast to the structural-reform mentality of the early 1980s.

Volcker was willing to accept short-term pain in order to eliminate inflation and force reform. The turmoil he created ultimately contributed to a productivity boom, strong real growth, and a 25-year bull market.

Today, the short-term thinking of central bankers is producing the opposite result. It is encouraging reckless behavior at both the corporate and government levels.

Five years ago, one could perhaps have argued that these policies were necessary to achieve economic “escape velocity.”

But after eight years of radical monetary experimentation accompanied by sluggish growth, and after more than two decades of stagnation in Japan, that argument has become increasingly difficult to sustain.

Smoothing normal cyclical fluctuations should never be confused with continuously borrowing from the future in order to consume today.

The Federal Reserve has no endgame.

Its objective increasingly appears to be to make it through another six months without allowing the S&P 500 to fall 20% and without allowing the U.S. economy to enter recession.

But in doing so, policymakers are encouraging exactly the kinds of behavior that prevent necessary reforms. They are increasing, rather than reducing, the probability of the economic tail risks they are trying to avoid.

At the government level, suppressed market signals have allowed politicians to continue postponing long-overdue entitlement and tax reforms.

Look at the slide behind me.

The doves continue to ask: “Where is the evidence of overinvestment?”

As you can see, operating cash flow peaked five years ago and has recently turned negative year over year, while net debt continues to increase at an alarming rate.

We have not seen anything like this since World War II.

In the cycles preceding the Great Recession, these two measures generally peaked at approximately the same time.

Even during that cycle, they diverged for only about two years.

Historically, when EBITDA turns negative year over year, as it has today, net-debt growth has already been declining for more than two years.

This five-year divergence is unprecedented in modern financial history.

If that is not alarming enough, consider how debt has been used during this cycle.

In the 1990s, debt financed the buildout of the internet.

Today, much of that debt is financing financial engineering rather than productive investment.

The distinction is very clear on this slide.

The purple portion represents share repurchases and acquisitions. The green portion represents capital expenditures.

In the 1990s, the green portion dominated.

In the current cycle, the purple portion dominates.

Think about that.

Last year, share repurchases and acquisitions totaled approximately $2 trillion.

All spending on research and development and office equipment combined totaled roughly $1.8 trillion.

After eight years of virtually free money, this behavior has accelerated nonlinearly.

In 2012, share repurchases and acquisitions totaled approximately $1.25 trillion, while spending on R&D and office equipment totaled about $1.55 trillion.

Since then, as asset valuations have risen, R&D and office-equipment spending has increased by only about $250 billion.

Financial engineering, meanwhile, has increased by approximately $750 billion, three times as much.

You can only live off your savings for so long.

Despite a $1.7 trillion increase in net borrowing without a corresponding increase in interest expense, the corporate sector’s share of profits peaked in 2012.

Today, the corporate sector is trapped in a vicious cycle of earnings management, questionable capital allocation, weak productivity, declining profit margins, and continuously rising debt.

And investors are paying 18 times earnings for these assets.

[Slide 7]

The second source of short-sighted policy comes from China.

In response to the global financial crisis, China launched a roughly $4 trillion stimulus program.

But after a decade of enormous infrastructure investment, which had become China’s primary mechanism for stimulating growth, additional investment only exacerbated the existing overcapacity in the economy.

Unsurprisingly, the stimulus generated only a temporary increase in nominal growth.

When we first became concerned about the ratio of Chinese banking assets to GDP in 2012, we could hardly have imagined what followed.

Over the subsequent four years, credit expanded by an amount equivalent to roughly 70% of GDP.

Put differently, since 2012 Chinese banks have effectively created the equivalent of Brazil’s entire GDP in new credit every year.

Imagine financing the equivalent output of an entire country each year through new housing, infrastructure, and related investment.

Remarkably, this enormous credit expansion occurred while nominal GDP growth declined from approximately 15% to 5%.

For companies borrowing at 10% interest rates while expecting 15% revenue growth, that change is devastating.

We therefore strongly suspect that a significant portion of this credit expansion has simply gone toward refinancing borrowers that were already insolvent.

Otherwise, how do we explain the apparent absence of significant non-performing loans in heavy industries that have simultaneously experienced collapsing prices and slowing sales?

The result is that before the stimulus program, China required approximately $1.50 of new debt to generate $1 of additional GDP.

Today, it requires roughly $7.

That is an extraordinarily dangerous development.

The most recent comparable historical example was the United States during the mid-2000s subprime boom, when the amount of debt required to generate $1 of additional GDP rose from roughly $1.50 to $6.

Two years ago, we hoped China would be willing to accept slower growth in exchange for structural reform.

Unfortunately, encouraged by the G7, Chinese policymakers chose another round of investment-driven fiscal stimulus.

That may buy them time, but it will ultimately make the underlying problem worse.

China does not need more debt and more houses.

As the chart demonstrates, this will remove one of the major cylinders from the global growth engine.

[Slide 8]

As I said earlier, short-sighted policymakers have no endgame.

They move from one short-term fiscal or monetary stimulus program to the next despite overwhelming evidence that these policies provide only temporary sugar highs while accumulating unproductive debt that weighs on long-term economic growth.

More importantly, the unprecedented stimulus of the past decade has coincided with a persistent decline in global growth.

That suggests we have borrowed too much from the future for too long, and the consequences are beginning to appear.

Three years ago, on this same stage, I criticized the logic of Federal Reserve policy but nevertheless reached a bullish medium-term conclusion.

At that time, the balance of evidence suggested that the global central-bank liquidity expansion would continue to push financial asset prices higher.

Today, I believe the balance of evidence has shifted in the opposite direction.

Higher valuations, another three years of unproductive corporate behavior, diminishing room for additional monetary easing, and excessive borrowing from the future all suggest that this bull market is running out of steam.

If we are mortgaging the future more aggressively than at any point in history, and financial markets are supposed to discount the future, then markets should trade at a discount to historical valuations, not at a premium.

It is difficult not to think back to 1982.

At that time, the market traded at roughly seven times depressed earnings, interest rates were about to be cut repeatedly, and productivity was beginning to rise.

Today, we are paying approximately 18 times elevated earnings while productivity is declining and policymakers have largely exhausted the interest-rate tool.

Over the past year, global equity markets have made little progress while volatility has increased, a combination that has often preceded major changes in market trends.

In my view, that leaves the risk-reward profile unattractive until either asset prices decline significantly or meaningful structural reforms take place.

Do not expect the latter anytime soon.

Policymakers may not have an endgame.

But markets do.

Let me leave you with one final question.

When I founded Duquesne Capital in 1981, what was the last asset you would have wanted to own?

Here is a hint.

It has traded for roughly 5,000 years.

Today, for the first time in many parts of the world, it effectively offers a positive interest-rate differential while central bankers experiment with the absurd concept of negative interest rates.

Some people consider it a metal.

We consider it a currency.

And it remains our largest currency allocation.

I am Stan Druckenmiller AND George Soros truther - they are my investment heroes. I will try to use this article to differentiate their approach towards the markets, and I am leaning towards George Soros’ take on “free markets”.

I would indeed listen to the Bond market. But I would never confuse what the bond market says with the truth. George Soros’ entire approach to financial markets begins with what I have called “a belief in our own fallibility.” What he means by that is; perceptions are imperfect, and those imperfections have tremendous impact because “misconceptions and misunderstandings that go into our decisions help shape the events in which we participate.”

Based on that, I have trouble assuming the 30-year Treasury yield as an objective measure of America’s fiscal condition. The bond market (for any nation) is not standing outside the economy taking its temperature - as we all know, the bond market is participant in the economy. Things across the market including mortgages, auto loans, student loans et al. are key’d off of where the yields stand on a given day.

George Soros has argued that “financial markets always present a distorted picture of reality.” More importantly, the distortion does not remain confined to price. “The mispricing of financial assets can affect the so-called fundamentals” those assets supposedly reflect. Soros calls this reflexivity, which I try to apply to the markets and think about all the time.

Apply reflexivity to the long bond dilemma of 2026 - really post 2022. Investors become concerned about deficits and Treasury supply (so am I, it is a global problem). Therefore, investors demand a higher term premium causing long yields rise. Those higher yields increase federal interest expense, increase mortgage rates, weaken housing and investment, tighten credit, pressure asset prices and eventually slow economic activity. The deficit then becomes larger and the debt dynamics worse. The market’s original fear begins creating the conditions that validate the fear.

Stan Druckenmiller describes this phenomenon as price discovery. Based on my understanding of reflexivity, George Soros would describe it as a feedback loop.

George Soros’ interpretation of markets has always been “built on the twin pillars of fallibility and reflexivity.” Positive feedback can reinforce “both the prevailing trend and the prevailing bias.”

In such scenario, as of August 24, 2026 right before the Fed’s Jackson Hole meeting, a rising Treasury yield may be revealing deteriorating fundamentals. It may also be worsening those fundamentals. Once those two things start feeding into each other though, I dont know how you can really seperate what is the signal anymore vs what the signal itself is causing.

This is why George Soros has always rejected what he calls market fundamentalism, the belief that markets tend toward equilibrium and should generally be left to correct their own excesses. The financial crisis is the latest extreme example that demonstrated the weakness of market fundamentalism belief. Markets can move very far away from equilibrium precisely because perceptions and fundamentals can reinforce one another.

This is where I have a little bit of an issue with Stan’s argument though. He, alongside George Soros, has spent an extremely successful career making money PRECISELY because markets become mispriced, narratives become extreme, policymakers change incentives and trends overshoot fundamentals. Yet here in the article Druckenmiller treats one particular market, the Treasury market, almost as an infallible institution.

There is also an important distinction between saying intervention may be necessary and saying policymakers are correct. I have never believed the latter. The same fallibility applies to governments. In discussing financial regulation, Soros explicitly warned that “regulators also operate with a bias.” So Treasury can be wrong too.

If Bessent is simply uncomfortable with a 5.5% 30-year yield and begins buying duration whenever markets challenge that level, Stan’s warning becomes powerful. Investors will learn the reaction function. They will test Treasury. What begins as liquidity management becomes price management, and what begins as price management can become an increasingly expensive commitment. That itself would be reflexive.

But that danger also doesn’t mean the government should just always stand aside. Soros’ point is basically that governments can be wrong, and markets can be wrong too, at the same time. Both sides are fallible.

And in this scenario Druckenmiller’s argument is that higher yields impose discipline on Washington. Maybe they do. But what happens if higher yields themselves increasingly become part of the fiscal problem? Higher term premium means higher interest expense, higher interest expense means bigger deficits, bigger deficits means more issuance, which can then push yields even higher. At some point doesnt the so called fiscal disciplinarian itself start making the fiscal situation worse?

This is where Soros’ idea of “the illusion that fundamentals and values are independent” becomes very relevant to me. Druckenmiller’s argument kind of puts fiscal policy on one side and the bond market judging it on the other. But thats not really how it works.

Deficits influence yields, yields influence deficits, Treasury policy influences positioning, positioning influences yields, and then those yields feed back into the economy and politics again. There really is no clean independent variable here.

So I am not saying Bessent should suppress the long bond, and I also dont think a 5.5% 30-year yield necessarily means there is some crisis. Druckenmiller is ultimately right that Washington actually NEEDS pressure from the bond market before politicians deal with the primary deficit and entitlements.

For me the more interesting question is what misconception is driving the trend, whether the trend itself is reinforcing that misconception, and what eventually breaks the loop. Maybe Washington is wrong to think it can run 6% deficits at full employment without eventually paying a materially higher real rate. Maybe the bond market is wrong to think deteriorating fiscal arithmetic must mean continuously higher yields regardless of what those yields themselves do to the economy. And maybe Treasury is wrong to think buybacks can interrupt the whole thing without actually changing fiscal policy. All three could literally be true at the same time, which is my main point.

To conclude, none of this takes away from how much I admire Stan Druckenmiller. In 2013, when most people on Wall Street were happy to complain about Washington privately, he actually went around college campuses with Geoffrey Canada trying to explain the entitlement problem directly to younger Americans, basically telling them that the current system was stealing from their future. I have enormous respect for that. You can disagree with parts of someone’s framework and still recognize when they have spent decades putting their reputation, money and time behind what they believe. That is a big part of why Druckenmiller remains one of my investment heroes - I love & admire him.

Below are the positioning charts from Sunday.

Since then Bitcoin is up ~15% and Ethereum is up ~25%. Alright, so what now?

Like precious metals, cryptocurrencies may be the most reflexive asset class there is because price itself changes the underlying reality faster than almost anywhere else in markets.

Rising prices attract capital, attention, leverage, developers, institutional participation and new narratives, which then improve liquidity and perceived legitimacy and push prices higher still.

Falling prices reverse the entire process: liquidity disappears, leverage unwinds, projects lose funding, sentiment collapses and the deterioration in fundamentals becomes a consequence of the decline that began in price.

There is very little stable notion of intrinsic value anchoring the cycle, so perception and reality can reinforce each other for extraordinary periods in both directions. Starting next week I am expecting a lot of investors and market commentators to build fundamental stories behind the move such as

• US Dollar declined following Treasury Secretary Scott Bessent’s comments

• Stablecoins are having their Minsky moment

• Bitcoin traded with software, and is just catching up to price action

Maybe there is a fundamental reason, but I don’t care - I personally hate cryptos. I think it made a lot of people who don’t deserve the capital wealthy and that wealth did not enter the economy in a purposeful way. There are no Andrew Carnegie’s building libraries or John D. Rockefeller’s building schools and hospitals with cryptocurrency wealth. Anyway…

I care about the price action - there are many bagholders all the way up to prior highs for Bitcoin ($130k) and Ethereum ($5k). I think it will be a challenge to hit new ATHs, BUT I think it will happen before year ends - most likely post midterms.

Look at Bitcoin ratio to gold and S&P 500 - the underperformance has been impressive, and the recent spike to the upside has been a very unique move.

I think 126 day rate of change will hit 200%+ before this rally pauses for Bitcoin. What does that mean? I wouldn’t be surprised to see Bitcoin >$250,000 over the next 24 months.

I’ll finish this article with a quote from legendary Bruce Kovner - as a crypto hater, adhering to this helped me catch the bottom in the asset class with size.

“I’m not sure one can really define why some traders make it, while others do not. For myself, I can think of two important elements. First, I have the ability to imagine configurations of the world different from today and really believe it can happen. I can imagine that soybean prices can double or that the dollar can fall to 100 yen. Second, I stay rational and disciplined under pressure.

[Successful traders are] strong, independent, and contrary in the extreme. They are able to take positions others are unwilling to take. They are disciplined enough to take the right size positions. A greedy trader always blows out.” - Bruce Kovner

The direction of the trade is weaker Japanese Yen versus flat to strong US Dollar - a month of so ago we had intervention where the US Treasury sold Euros to buy Yen and today the treasury made a big announcement. See below.

Treasury Announces Increased Sizes of Nominal Long-End Liquidity Support Buybacks Beginning September 9

The Treasury announcement today looks small in nominal terms. Buyback operations in the 10-20 year and 20-30 year sectors are being increased from a maximum of $2 billion to at least $4 billion per operation. So, what does that mean? What is happening?

The 30-year yield pushed above 5.3%, its highest level since 2007, and Treasury responded by increasing purchases specifically in the part of the curve under the most pressure. It also did it between Quarterly Refundings rather than waiting until November.

There is actually historical precedent for the tool Bessent is using BUT there is less precedent for the environment in which it is now being used.

Treasury ran a major buyback program between 2000 and 2002. The mechanics were similar: buy older, less-liquid securities and improve the functioning of the Treasury market. But the macro backdrop could hardly have been more different. The federal government was running budget surpluses and paying down debt. Treasury was dealing with the problem of too little issuance. Those were the good days in the US where positioning of the government’s balance sheet was so so strong that America had to issue more debt on the long end.

Today the government is running large deficits. Every dollar Treasury uses to buy back long-dated bonds ultimately has to be financed somewhere else, which makes Operation Twist a more useful analogy for thinking about the economic effect.

In 1961, and again under the Federal Reserve in 2011, policymakers tried to influence the shape of the yield curve by reducing the amount of duration held by the private sector. Buy longer-dated bonds, leave more short-duration paper outstanding, and long-term yields should be lower than they otherwise would be.

What is NOT happening that I see articles and some respected economist say is quantitative easing (QE). Bessent is NOT running QE. Treasury cannot create reserves, and these purchases do not eliminate the government’s financing requirement.

Prior Treasury secretary Janet Yellen created the modern Treasury buyback program in 2024. Bessent doubled the frequency of long-end buybacks in 2025. Now Treasury is doubling the size of those operations. Bessent actually does not like this, but the spike in long-end of the yield curve will cause something to break if it keeps trending the way it did since June 2026.

At the same time, Treasury has been relying heavily on bills to absorb incremental financing needs while keeping coupon auction sizes relatively stable.

Put those together and the government is effectively shortening the duration of debt being supplied to the market at the margin. That is not Operation Twist in the formal sense, but economically the resemblance is becoming harder to ignore.

The part I find most interesting is the emerging reaction function.

A few years ago, Treasury described buybacks primarily as a liquidity-management tool. Today, after a violent selloff in long-duration bonds, Treasury increased its buying specifically in the 10-30 year sector.

That does not mean 5.3% on the 30-year is a hard ceiling. Fiscal deficits, inflation, oil, foreign demand and term premium can overwhelm a few billion dollars of Treasury purchases very easily.

But markets care about the marginal buyer and, even more importantly, about how policymakers respond when pressure builds.

If the November refunding brings another increase in long-end buybacks, continued restraint on long-dated issuance and heavier reliance on bills, I think the interpretation becomes much clearer.

George Soros’s broader point about markets has always been that they do not naturally tend toward equilibrium. Trends can reinforce themselves until the consequences of the trend force a change in behavior.

The long end had increasingly started to look like that kind of process to me: higher yields tightening financial conditions, worsening fiscal arithmetic and creating the possibility that eventually something would have to break.

I am not convinced that outcome was inevitable, or that today’s announce-ment solves the underlying problem. But for at least one day, the dangerous trend reversed rather than accelerated.

The market spent the last couple of years rewarding bottleneck solvers within the AI buildout theme such as memory names like , GPU/CPU semiconductor designers and manufacturers like , independent power producers like , and many many other names et al. It is a long list.

However, the market not only didn’t reward but also penalized the ultimate owners of the infrastructure for the cash cost of that buildout - I believe we are quickly approaching a unique point where the platform companies either convey investors that they can monetize (positive ROI on AI CapEx spend) or decrease the growth rate of capital spending (burning less capital). Both can improve earnings and free-cash-flow optics, but only if underlying demand remains intact.

What I am trying to say is, mega-cap names like and we can include here also did not perform as well as they did in the 2010s AND are entering a period where they can’t lose. They will either generate some real productivity from all the money they are spending on AI buildout or they will slowdown and restart their historical approach to spending free-cash-flow, which is share buybacks. Risk/reward to seems extremely asymmetric to me.

I asked ChatGPT historical performance of these names - see below:

Look at the three year CAGR - , and underperformed S&P 500. The only company that outperformed Nasdaq 100 over the past three years is

So, what is going on?

What is going on is, the mega-cap companies haven’t spent their excess capital on anything but buybacks for a long-time and artificial intelligence changed that pattern. One of the modern day investors I admire alongside George Soros, and Stanley Druckenmiller is Peter Thiel whom I have written about in the past - let’s bring him into this conversation: Peter Thiel’s view is skeptical of share buybacks when they become the default use of corporate cash. He talks about this on his famouse book Zero to One. Thiel describes modern finance as an “indefinite” system in which capital keeps circulating because nobody has a sufficiently concrete idea of what to build. Entrepreneurs give cash to banks, banks give it to institutional investors, institutions buy stocks, and successful companies generate free cash flow and then “issue dividends or buy back shares,” sending the money back through the system. His criticism is that capital has become an end in itself rather than a means of creating something new.

He gave example often in his speeches saying Apple is using enormous free-cash-flow for buybacks because the management team does not have good ideas anymore to spend the capital on.

So, In Thielian terms, the hyperscalers’ current CapEx boom is almost the opposite of the buyback/stagnation critique. Peter Thiel has been preaching for mega-cap tech companies to use their free-cash-flow for something other than share buybacks for almost two decades.

The real question becomes whether the AI investment is definite optimism backed by genuine monopoly economics, or simply another consensus investment cycle with poor returns.

Now back to CapEx numbers. Amazon entered 2026 expecting about $200 billion of CapEx. Alphabet now guides to $195–205 billion. Microsoft expects roughly $175 billion of calendar-2026 CapEx investment after the accounting change around leases. Meta guides to $130–145 billion. Those four alone are therefore running roughly $710 billion of 2026 capital investment at the midpoint.

I believe there are two paths now for these companies. If revenues generated by that installed compute base begin accelerating faster than depreciation and operating costs, earnings estimates rise. If capacity catches up with demand and CapEx growth falls from extraordinary rates toward normal rates without destroying revenue growth, free cash flow inflects sharply. In either scenario and + win.

What I do not want is CapEx slowing because management discovers that customers will not pay enough for the compute. That scenario hits both the numerator and the multiple. In short, my binary (optimistic) scenarios boil down to:

• Good outcome 1: CapEx stays high because ROI is demonstrably high, and revenue/EPS growth accelerates.

• Good outcome 2: CapEx growth decelerates because the infrastructure catch-up is largely complete, while AI/cloud monetization continues, creating a major FCF harvest period.

• Bad outcome: CapEx remains elevated while incremental revenue, utilization or margins weaken. That is stranded/low-return capital rather than normalization.

And the latest numbers increasingly favor outcome one.

I wouldn’t rank these names, and pick long/short themes. What I will do next is, try to summarize each names’ upside in a sentence and drop couple of charts that make me bullish these names.

: the cleanest proof that ROI has arrived (Azure growth of 43%)

When 126-day rate of change turns negative ~10% and reverses to positive you can see that the stock rallied hard. The chart is now at a clear make/or break level - if it crosses prior highs, I think the stock doubles over the next 24 months. If it fills the earnings gap, the chop continues

: the purest CapEx-normalization equity

is super cheap, and despite weak price action since Q4 2025 - I’d argue it is chopping close to highs, has a clear falling wedge forming and if it breaks above the 21-DMA and sustains that as MACD turns positive, the stock can rally hard - much more than investors anticipate.

: the P&L is already showing the return before FCF does

’s uptrend is messy, but it is an uptrend - there is one major factor that is damaging the stock often. Jeff Bezos sells this thing whenever it crosses key technical levels, and that kills the reflexive positive bid in the short term. The founder himself kills the momentum! The market sees no FCF because is reinvesting essentially all of its enormous operating cash generation. But recent earnings’ AWS growth and operating income suggest the assets are already monetizing.

: less catch-up, more “the bear case isn’t happening”

Google had the weakest chart amongst these names, and now has the strongest chart. What makes me bullish these names in the first place is the reaction to following a prolonged period of chop, this was a $2Tn+ company before the name doubled in twelve months. If Google can, why can’t other $1Tn+ companies? They are also less woke now that the founders are steering the ship again, so models should be decent and also there is YouTube… infinite amount of visual data to feed the machine that is proprietary.

: highest convexity, OpenAI bet

has the highest chance to blow up amongst these names, but if you think OpenAI will generate the revenues it is targeting (which I think they will) and need the extra compute to achieve their goals (which I think they need) then the highest convexity bet is . Also, even if you think you have a really high IQ and are a savvy businessperson - do you really want to bet against Larry Ellison? The literal samurai who is one of the last surviving members of the old Silicon Valley guard? I don’t - I love the man.

:it is what it is

has been out the game for a while, but lack of AI CapEx spend has been positive to the price action as market participants been going long this name while shorting other mega-cap names. I think Apple will be just fine, the fact that the stock is close to ATHs during a generational CEO transition keeps me optimistic with this name.

Fundamentals

I’ll keep this part short, because it is a joke. In modern market structure, fundamentals in the short term (36 months - so not that short) do not matter because you can go long and short, beta adjust and/or delta adjust the positions in a way where you can make a portfolio valuation insensitive. And a lot of marginal buyers/sellers (hedge funds especially) do this.

Long way of saying, fundamentals of these companies are unbeliviable - I mean who in their right might would buy Costco at 50x P/E when they can buy at 20x P/E? Who would buy at 50 P/S when they can buy at 6.5? The answer is a lot of people, it is a sad reality of public investments - mark to market gets you and fundamentals don’t really matter. What matters is are fundamentals improving or deteriorating from current levels (rate of change)?

Conclusion

I think I am right, and if I am wrong - heading into the most challenging part of the year in a Midterm year, the risk/reward is not that crazy to the downside. These names are not performing that well against the indices year-to-date. If you look at the latest filings, you’ll see all the legendary investors loading up on these names and earnings are objectively quite strong.

The only real concern I have is, if I am right and if these names work alongside semiconductor names + software names then we will see Nasdaq 100 up 50%+ and S&P 500 up 30%+ in 2027, and maybe for the remainder of the decade. That creates its own reflexive risk, which will then translate to a new boom/bust cycle. I don’t want to get bearish before all the price action I anticipate starts but yeah - the end of this move will probably also lead to a 30%+ index level haircut (not next year or the year after I don’t think).

The constraint is increasingly obvious. Look even if you are bear on AI CapEx buildout, customers exist, GPUs exist (in high demand), and capital exists. Getting several hundred megawatts energized quickly is becoming the gating factor.

That is almost a direct advertisement for Bloom Energy. Nebius itself selected Bloom because traditional grid timelines were too slow. Their first deployment is 328 MW of behind-the-meter Bloom fuel cells expected to be operational this year, with Nebius explicitly citing fast time-to-power and reduced dependence on transmission construction.

is becoming the bridge between the AI-capex boom and the electrical infrastructure that has to arrive before that compute can generate revenue. CRWV/NBIS therefore strengthen the argument that Bloom’s current growth isn’t a one-customer anomaly.

I think every additional neocloud announcing gigawatts of capacity increases the value of shortening time-to-power. Price action also is confirming my claim so far. This takes us to independent power producers, stocks that ran a lot in 2024-2025 but have been chopping around for almost two years now.

, and had an enormous first rerating in 2024-25 when the market discovered that nuclear and dispatchable generation were AI assets. What has not worked recently is the second leg. The stocks became waiting trades while memory and optics started producing immediate quarterly revenue revisions.

Investors could see HBM shortages and 800G/1.6T optical demand in , and earnings today, while the power thesis depended on PPAs, interconnection reform, capacity markets and contracts whose economics don’t show up for years. Lumentum, for example, just reported revenue up 109% and its shares are roughly six times higher over the past year.

Power has consequently lost relative momentum, with Vistra Corp/Constellation Energy Corp in particular pressured recently by uncertainty around PJM/ERCOT rules and how much of the AI scarcity rent generators will actually capture.

I think that sequencing is what can reverse over the next several quarters. Compute demand is now being proven so aggressively that the market has to move from asking whether AI needs more electricity to asking who owns the megawatts that can actually be delivered.

Talen already has roughly 4 GW of powered-land/data-center contracting options, raised 2026 EBITDA guidance to $2.0-$2.2 billion and is buying back stock. (Talen Energy Corporation)

Vistra has locked in 20-year Meta PPAs covering more than 2.6 GW of nuclear capacity, while its 2027 EBITDA opportunity is now $7.4-$7.8 billion before further data-center contracts. (Vistra Corp. Investor Relations)

Constellation just raised 2026 earnings guidance, signed another 920 MW of 15-20 year nuclear PPAs and cleared major regulatory milestones for the Crane restart. (Business Wire)

The story is finally migrating from “someday data centers need power” into contracted MW, earnings revisions and longer-duration cash flows.

Fundamentally all three names look good.

Technically, they have been super weak in 2026 but I feel like following neocloud earnings we may finally see some positive traction as the names recover 200-day moving averages.

Rate of Change

If you look at rate of change over 126-day moving period, you’ll notice that all three names are chopping around 0 - just sideways chop since January 2026.

Technicals

For technicals, I am showing 2-year timeframe, weekly candles. For moving averages I used 65, 130, 195, and 260 week moving averages that translate to quarterly moves. You can see that 130-week moving average (green line) has been decent support.

Monthly Seasonality Explorer

Finally, when you look at monthly performance, September/October/November/December period has been quite good for and since IPO.

Conclusion

In conclusion, most AI CapEx tangent sectors (even the ones who should not be tied to the theme) have run quite a lot in 2026. Independent power producers have not, because they are digesting the gains. I think offer serious upside potential over the coming months / quarters.

This article is from October, 2025.

This article is from March 2026.

Up until last week of July 2026, not much has happened in the USD/JPY trade that proved my mental model right - frustrating, but then again I built a thesis where Bank of Japan, Japanese Treasury, Japanese Prime Minister, Federal Reserve, the US Treasury, and POTUS need to align a certain way. BIG global macro trade, and I am in it big (100%+ NAV size) because I have the dream to be one of the persons who has participated in a generational currency trade with AD Fund Management LP.

At 8:25 p.m. tonight (August 2, 2026) EST, USD/JPY was trading close to 157.50. Within twenty minutes, the dollar fell through 156, briefly traded near 155.30 and then recovered toward 156.25. The chart looks like an isolated air pocket. It is better understood as the latest stage of a much larger regime change.

For most of the past four years, the yen’s depreciation was rational. The Federal Reserve raised rates aggressively while the Bank of Japan kept its policy rate near or below zero, maintained unusually accommodative financial conditions and moved slowly away from yield-curve control. Even now, the Fed funds target remains 3.50–3.75%, against a 1% overnight rate in Japan. The spread pays investors to borrow yen, sell it and own higher-yielding dollar assets. As long as volatility remained contained, the carry trade generated its own demand.

The weakness became reflexive. A declining yen increased the value of Japanese companies’ overseas earnings and improved exporters’ price competitiveness. Japanese institutions earned higher returns abroad, while corporations had little reason to repatriate foreign profits immediately. Traders learned that every attempted yen rally could be sold because the fundamental interest-rate structure remained intact. Japan also had deep institutional reasons to tolerate depreciation.

The generation that built modern Japanese monetary policy lived through the consequences of the yen’s earlier appreciation. Under the Plaza Accord of September 1985, the United States, Japan, West Germany, France and the United Kingdom agreed that the dollar needed to decline. The objective was to correct an increasingly unbalanced global system in which high American interest rates and expansive U.S. fiscal policy had driven the dollar higher and placed enormous pressure on American manufacturers.

The agreement worked with extraordinary force. The yen appreciated approximately 46% against the dollar by the end of 1986. Japanese exporters were squeezed, growth slowed and policymakers responded with easier money and fiscal support. Plaza alone did not cause Japan’s subsequent asset bubble and lost decades, but the sequence became embedded in Japan’s governing memory: external pressure produced yen appreciation, yen appreciation damaged industry, and the policy response contributed to the financial excesses that followed.

That history shaped three decades of behavior. A strong yen became associated with factory relocation, declining export competitiveness, falling prices and economic stagnation.

Yen weakness, by contrast, was treated as reflationary. It raised corporate profits, supported the stock market and helped Japan escape the psychology of deflation. The mistake was assuming that this relationship had no limit.

Beyond 150 and especially beyond 160, the weak yen stopped functioning primarily as a stimulus to exporters. Japan imports much of its energy, food and industrial raw materials. Currency depreciation raised the domestic cost of those imports, reduced real household purchasing power and made ordinary voters feel poorer. Large multinational exporters benefited, but households, smaller companies and import-dependent businesses absorbed the cost.

The political distribution of the trade changed. Yen weakness became evidence that Japan had lost control of its currency.

I believe we are living through history with respect to USD/JPY - the price action behind the currency pair has global ramifications that will impact everybody. Japan has intervened repeatedly since 2022, but unilateral intervention could only interrupt the trend while the rate differential continued to favor the dollar.

Recently, Japan and the United States jointly bought yen after USD/JPY approached 164, the weakest level since 1986. Japan reportedly committed about ¥8.45 trillion, and U.S. officials indicated that they were prepared to participate again if volatility returned.

That changes the market’s reaction function. Yen shorts are no longer trading only against Japan’s Ministry of Finance. They are potentially trading against the combined policy interests of Tokyo and Washington.

The American involvement is important because the weak yen had begun to threaten markets beyond Japan. To finance large-scale intervention, Tokyo can sell dollar assets, including U.S. Treasuries. Japan holds roughly $1.14 trillion of Treasury securities. At a moment when long-term American yields are already under pressure from fiscal deficits, issuance and term premium, Washington has an incentive to prevent disorderly Japanese reserve liquidation. The yen, Japanese government bonds and U.S. Treasuries have become part of the same policy problem.

The move from 164 toward 156 therefore reflects more than intervention. The BOJ is slowly normalizing policy, one board member has argued for a higher rate, and the United States has signaled that unlimited yen depreciation will not be treated as benign. Once that signal changes, positioning begins to change with it. Investors cover shorts, volatility rises and the expected return on the carry trade falls. A position that looked safe at 150 can become unstable at 160 because the authorities’ tolerance is no longer the same.

Could the yen eventually strengthen beyond 100 per dollar?

It is possible, although intervention by itself will not produce that outcome. Moving from 156 to 100 would require roughly 40% appreciation and a reversal of the macroeconomic structure that produced the weak yen.

The cleanest path would involve a U.S. recession or financial shock that forces the Federal Reserve to cut rates sharply, while the BOJ continues tightening because Japanese wages and inflation remain firm. The U.S.–Japan rate differential would compress, the dollar carry advantage would disappear and Japanese pensions, insurers and corporations would have greater reason to hedge or repatriate foreign assets. A global deleveraging event could accelerate the process because investors would need to buy yen to repay yen-funded positions.

The move would then become reflexive in the opposite direction. Yen strength would force additional short covering. Rising volatility would make the carry trade less attractive. Japanese institutions would increase currency hedges as the dollar declined, creating further demand for yen. Below 130, the market would begin discussing 120. Below 120, the institutional memory of 100 and 80 would return.

That would resemble the direction of the Plaza Accord, though today’s intervention is still far smaller in scope. Plaza represented an explicit multilateral decision to depreciate the dollar. The current operation is officially aimed at excessive volatility in one currency pair. It becomes a modern Plaza only when Washington actively prefers a weaker dollar, the Fed supplies monetary easing, the BOJ continues tightening and intervention is sustained rather than episodic.

A move toward 120–130 can occur through ordinary policy convergence. A move below 100 requires a U.S. downturn, a major unwind of global dollar positioning and durable cooperation between Washington and Tokyo.

The fall on the chart is therefore meaningful, but it is not yet the full move. The real change is that yen depreciation has crossed the point where Japan’s historical fear of a strong currency outweighs the present economic and political costs of a weak one. The market spent years learning that Tokyo would tolerate almost any level. It may now have to learn where that tolerance ends.

What does this mean for professional investors? A global margin call is looming. What should one do about this? Now is not the time to max-out margin in either long or short direction across any asset classes.

Eighteen years later, Griffin was on the other side of the trade.

Citadel Buys Bulk of Situational Awareness’ AI Stock Wagers

Citadel bought most of the public-equity portfolio of Situational Awareness after the fund was hit by losses across its leveraged AI positions. Leopold Aschenbrenner had built the firm at extraordinary speed. Assets grew from less than $1 billion to more than $20 billion. The public portfolio reportedly reached roughly $16 billion, and the fund returned 439% net in the first half of 2026.

The trade had worked so well that its own success created the conditions for more risk. Higher prices increased collateral values. Strong returns attracted capital. Lenders became more comfortable. The fund could add leverage and increase the size of the positions that had already made it successful.

Once the AI trade began to weaken, the same process ran in reverse. Falling prices reduced collateral. Lower collateral created margin calls. Margin calls forced sales into a declining market. Those sales pushed prices down further and tightened the financing pressure across the rest of the portfolio.

Within 24 hours, end of July 2026, Situational Awareness sold the leveraged portion of its public book. Goldman Sachs and JPMorgan helped arrange the transaction. Citadel ultimately bought the assets after other firms, including Millennium, had looked at the portfolio.

The long-term view on AI may not have changed much during those 24 hours. The financing conditions changed completely. The fund’s investment horizon was no longer determined by how long the AI buildout would last or how much earnings power the companies would eventually generate. It was determined by collateral requirements and how much cash the prime brokers wanted the next morning.

Soros built his entire investment framework around these feedback loops. Prices do more than reflect reality. They influence behavior, financing conditions and capital allocation, which then change the reality investors are trying to value.

A rising stock price can make a company stronger. It can lower the cost of capital, support acquisitions, attract employees and allow management to spend more aggressively. In a leveraged portfolio, rising prices also improve collateral and increase the amount of risk the investor can carry. The strength of the market begins to validate the thesis, and the thesis brings more money into the market.

The process can continue far beyond what traditional valuation work would suggest. It can also unwind much faster than the fundamentals themselves.

Ken Griffin learned this in 2008. Citadel had significant exposure to convertible arbitrage and other relative-value strategies. These positions were designed to carry limited market direction. They relied on financing, liquid dealer markets and the ability to hedge the equity exposure by shorting the underlying shares.

After Lehman failed, funding markets froze and relationships that had held for years broke apart. Securities that appeared unrelated began moving together because everyone was selling what they could. The SEC then banned short selling across a large group of financial stocks. Convertible-arbitrage funds that owned the bonds and shorted the common shares suddenly could not maintain the same hedges.

Citadel’s positions became more volatile as the market around them became less functional. Investors asked for roughly $1.5 billion back. Griffin suspended redemptions rather than sell into a market where many assets had no credible bid. He later said that Citadel might not have survived if Morgan Stanley had failed and created another wave of forced deleveraging.

Citadel recovered. The funds gained roughly 60% in 2009 and eventually returned to their previous high-water marks. Griffin also came out of the crisis with a permanent understanding of what happens when leverage, regulation, liquidity and investor behavior all turn against the same portfolio.

That experience is probably one reason Citadel moved so quickly on Situational Awareness. Griffin had already owned positions that needed more time than the financing market was willing to provide. He knew how quickly a sophisticated trade could become a liquidation once counterparties and investors started demanding cash.

Market structure sits inside every investment thesis, whether the investor acknowledges it or not. Who owns the asset matters. How much leverage they use matters. The terms of that leverage matter. Redemption rights, dealer balance sheets, short availability, options positioning and the concentration of ownership can become more important than earnings estimates during stress.

A portfolio can own dozens of stocks and still be one trade. If every position depends on AI capital spending, momentum, cheap financing and the willingness of prime brokers to extend leverage, the diversification is mostly cosmetic.

Situational Awareness may still be right about the scale of the AI cycle. The private Anthropic stake it retained may ultimately be worth more than the public portfolio it sold. Citadel may also make a large return from the assets it bought.

Griffin nearly lost Citadel in 2008 because the market stopped giving him time. In 2026, he was the buyer across the table from a manager who had run out of it. I don’t like the sequence of events, but it is what it is. What’re you gonna do?

Lindsey Graham was a patriot, and a good man. May he rest in peace.

"This action will deny Assad and Iran access to critical resources to fund malign activities in the region as well as provide a source of revenue to our Kurdish partners in the ongoing fight against ISIS." - 11.15.2019

S2 Capital fund closed in September 2022, held a roughly $400 million value-add apartment portfolio, and, according to the investor letter reported by The Real Deal, was hit by a brutal combination of higher expenses, higher interest costs, cap-rate expansion, new supply, and falling rents. That is the entire real estate private equity cycle in one clean autopsy: buy late, finance floating, underwrite lease trade-outs as if the macro regime is irrelevant, then discover that the debt owns the asset before the equity ever did.

S2 Capital Dissolves $400M Multifamily Fund With No Returns

S2 Capital dissolves $400M first fund with “no return of capital”

I know this space, and I know it well. Before launching AD Fund Management LP, I was an investment associate at Greystar, covering multifamily, student housing, build-for-rent, and active adult across the US. People I worked for are best-in-class, I would argue you cannot find better real estate investors than East Coast Greystar investments team. I sat inside the models, I built models from scratch. I looked at rent rolls, renovation premiums, concession trends, capex budgets, debt quotes, exit cap assumptions, insurance inflation, taxes, payroll, turnover, bad debt, and the small operating deltas that decide whether a deal is a real estate investment or just a levered rate bet with granite countertops.

The outside world (non-institutional) thinks value-add multifamily is about buying tired apartments, renovating kitchens, pushing rents, and selling to the next buyer. Inside the machine, you learn very quickly that the biggest variable is usually the moment in the cycle when you buy.

The great founders in real estate understood this - if you look at the legendary real estate shops, you will notice that inception periods of almost all of them follow historical boom/bust cycles. Stephen Ross did not begin Related as the developer of Hudson Yards. He founded the company in 1972 as a financier and developer of affordable housing, using a $10,000 loan from his mother and building the business around the financing, subsidy, and policy architecture of low- and moderate-income apartments. Related’s early history was affordable housing, government programs, complicated capital stacks, and knowing how to sit at the intersection of public need and private capital.

Jorge Pérez’s (Related Florida originally) origin story is equally interesting. In 1979, before he became Miami’s condo king, he rehabilitated a small government-subsidized apartment building for low-income residents in Little Havana after working for HUD (The U.S. Department of Housing and Urban Development). The project needed plumbing, electrical work, and elevator access for elderly tenant. He made around ~$200,000 on that deal. Pérez then built Related Group into one of Florida’s dominant residential developers, starting from affordable housing before moving into the luxury skyline.

Barry Sternlicht (Starwood co-founder) and Bob Faith (Greystar founder AND Starwood co-founder) came out of a different wound in the system. After the S&L crisis, capital had disappeared from apartments. Assets were available because lenders, regulators, and broken owners needed balance sheet relief more than they needed perfect price discovery. Sternlicht formed Starwood Capital in 1991, and Faith, his Harvard Business School classmate, was part of that early apartment-buying machine. They bought into distress when the sellers were forced and the next buyer did not yet exist. In August 1993, Starwood contributed about 6,400 multifamily units into Sam Zell’s Equity Residential at its IPO, turning distressed private real estate bought after the crisis into public REIT currency just as the market was beginning to institutionalize apartments.

Sam Zell (Equity Residential - the Grave Dancer) understood that public market timing could be as powerful as asset-level underwriting. Equity Residential’s predecessor went public on the NYSE in August 1993, and Zell helped turn the REIT structure into an institutional ownership vehicle for real estate at exactly the moment when public investors were ready to own stabilized property cash flows. The genius was understanding when the ownership form itself was changing, when the private asset could be recapitalized by public equity, and when apartments were becoming an institutionally acceptable asset class.

Blackstone then scaled the entire idea into the modern era. After the financial crisis, Jon Gray and team bought into distress, aggregating real estate while capital was still impaired, then later used vehicles like BREIT to bring private real estate income to a much broader individual investor base. BREIT launched in January 2017, deep into the era of suppressed rates, when the world wanted yield, private marks, monthly income, and the illusion that stabilized real estate could behave like a bond with upside. That was not an accident.

This is the structural point most real estate private equity people still do not want to admit. The last forty years gave the industry a falling-rate tailwind so powerful that it laundered timing into track record.

Cap rates compressed, because US Treasury yields kept compressing. Refinancings rescued mediocre basis. Exit caps fell while models had cap rate expansion assumptions. Floating-rate debt looked intelligent because the Fed kept rewarding anyone who believed money would get cheaper again. Then the regime changed, and suddenly the same model that looked institutional became fragile.

Real estate private equity is especially dangerous because the leverage is institutionally acceptable - you are going levered long (long only), and keeping that leverage with private marks. In public equities, 2:1 leverage sounds aggressive. In real estate, buying an asset with one to two dollars of debt for every dollar of equity can still get described as prudent. That leverage makes timing cut both ways.

If you buy when capital is scarce, sellers are forced, and financing is about to improve, the debt magnifies skill.

If you buy when capital is abundant, lenders are loose, and the exit buyer is already underwriting perfection, the debt becomes a slow liquidation mechanism.

Carl Icahn has made the point for years that corporate America is full of buffoons, and the same disease hits every asset class after a long enough bull market. At the beginning of a cycle, the people in the room are operators, distressed buyers, lenders with scars, and investors willing to look stupid before they look right. Later, after the returns are visible, the asset gatherers arrive. They flood the market with capital, compress returns, hire people who have never seen a real down cycle, and mistake spreadsheet fluency for judgment. In real estate terms, they burn investor capital because they do not understand the violence embedded in lease trade-outs, floating-rate debt, tax resets, insurance shocks, and exit-cap math.

Today is not a great time to invest in real estate through a traditional private equity fund. The marks have not fully cleared. The maturities have not fully forced capitulation. The institutional bid is still too anchored to the last cycle. The time is coming over the next decade, but it will come through distress, restructurings, lender fatigue, broken capital stacks, and sellers who need certainty more than price.

That is why, under AD Fund Management LP, I intend to launch an open-ended real estate private equity fund designed to capitalize on the real estate crisis as it actually unfolds - BUT NOT YET. Soros Fund Management’s Quantum Fund was not a real estate fund, but George Soros understood something most allocators still miss: the best capital is flexible, reflexive, and patient enough to wait for the forced trade. That is the structure I want.

The next great real estate vintage will not belong to the people who raised the most money in 2021. It will belong to the people who understand that real estate is duration, leverage, liquidity, politics, credit, and human behavior tied together in physical form. S2 Capital is not the final nail in the coffin, we’re just getting started with liquidations.

Then, I rewatched several interviews of Stanley Druckenmiller - the one below especially may shine light on how Chair Kevin Warsh will run the Federal Reserve throughout his tenure.

As we know now, Kevin Warsh’s first real act as Fed chairman was to take away the market’s script. At the first FOMC, the funds rate stayed at 3.5% to 3.75%, the statement stayed focused on inflation, and the Committee reaffirmed ample reserves, but the substance was in the institutional posture, because Warsh walked into the first press conference and effectively told the market that the old game was over: the Fed was no longer going to spend every meeting telling investors what it wanted them to believe about the next meeting, the next quarter, and the next dot plot.

It’s important to note that, The Federal Reserve’s official biography now lists Warsh as chairman and notes that, after leaving the Board in 2011, he was a partner at Stanley Druckenmiller’s Duquesne Family Office, which matters because this is not a normal central-bank résumé and this is not a normal moment for monetary policy. I believe this is the first time we are going to be having a real disciple (not an operator) of a market operator governing the Fed.

The first Warsh statement was short, almost aggressively stripped down, with the Committee holding rates steady, keeping the target range at 3.5% to 3.75%, and reaffirming that policy would maintain ample reserves, while the press conference did the more important work of signaling that the Fed’s communication regime itself was now under review. Warsh said the 2% target remained the objective, said he saw no reason to revisit it before the Fed had reestablished its ability to deliver price stability, said “inflation is a choice,” and then, when pressed to say what came next, said the part the market should remember: “We’ve dropped forward guidance.”

In the Druckenmiller interview, the legend says the same thing verbatim.

In December 2018, Druckenmiller was being asked why a man who had spent years complaining that the Fed had been too easy suddenly wanted the Fed to stop hiking, and the answer was not that he had become dovish, or that he wanted a Powell put, or that he had forgotten his own argument about QE, bubbles, and the risk curve; the answer was that monetary policy, in his mind, should be regime-sensitive, market-aware, and risk-reward based, and the Fed’s mistake was allowing a previously announced path to dominate the facts in front of it. He had hated QE3 because it pushed investors into financial assets, inflated corporate leverage, canceled the bond market’s disciplinary function, and produced the same boom-bust pattern that eventually forces even more radical monetary policy, but by late 2018 the inside of the market, the curve, credit, commodities, housing, autos, banks, small caps, and cyclicals were all telling him that the tightening had already moved from theory into the body of the market.

This is the first thing to understand about a Warsh-led Fed if Warsh is actually operating with a Druckenmiller-shaped mental model: the hawk-dove language becomes useless. Druckenmiller’s framework was never “raise rates” or “cut rates” as a permanent identity; it was tighten when financial conditions are booming, pause when markets begin to signal policy is biting, reverse when credit and liquidity say the economy is breaking, and never let the Fed’s public commitment to an old path become more important than the market’s new information. The man who attacked free money was the same man who said, in that interview, that he was not calling for cuts but wanted the Fed to pause, take stock, and stop hiking into a market where the S&P was down, cyclicals were down far more, the yield curve had flattened into warning territory, credit was deteriorating, and balance-sheet runoff was already tightening liquidity.

Now the question about Kevin Warsh is; is he simply a more hawkish Powell, or is he trying to rebuild the Fed around the idea that the central bank should stop narrating the future and start listening harder to the present?

I am leaning toward the second. Warsh said financial markets work best when they react to incoming data, and less efficiently when they spend their time trying to infer how the Fed will react to that data; he went further and said financial market prices are probably the most important source of information for central bankers, but when those prices are only reflecting back what the Fed has already said, the central bank has blinded itself to its own best source of information, which is almost pure Druckenmiller.

Markets are supposed to be the economist. Markets are supposed to warn before payrolls, before GDP revisions, before the staff model, before the official recession language, and if the Fed spoon-feeds the market every six weeks, it corrupts the signal it later claims to be reading.

This is makes me think that ending forward guidance is a change in the distribution of risk. Forward guidance made the Fed an author of the curve, an author of volatility, and an author of leverage, because every additional sentence about the likely path of policy reduced the cost of carry, gave risk-parity and vol-control books something to lean against, allowed credit investors to underwrite duration with more confidence, and trained the equity market to convert every FOMC meeting into a volatility-suppression event. Warsh seems to be saying that this arrangement became self-defeating. If markets are always asking what the Fed will say next, they stop asking what the economy is saying now.

A Fed built this way will probably feel much harsher than the Powell Fed during expansions, because there will be less comfort language, less protection for multiple expansion, less sympathy for financial markets that are already loose, and less desire to pre-validate the next easing cycle before the data have earned it. If inflation is still above target, productivity is strong, AI capex is powerful, fiscal deficits are large, the labor market is not breaking, and the stock market is still loosening financial conditions on the Fed’s behalf, Warsh has no reason to rescue anyone. In that branch of the world, the Fed’s removal of forward guidance is effectively a tightening, because it raises the uncertainty premium around the front end, lifts the hurdle rate for duration assets, and forces risk to trade without the old metronome.

But the same framework can become very aggressive in the other direction once the market internals turn. Druckenmiller’s entire 2018 argument was that the Fed should have tightened earlier in the cycle when financial conditions were booming, then stopped when the forward-looking indicators moved to amber, because the cost of pausing and being wrong was small, while the cost of hiking into a liquidity break was five to ten times larger. He was explicit that balance-sheet runoff was tightening policy even without another rate hike, because deficits required ongoing Treasury issuance into a market already absorbing Fed runoff, and he mocked the idea that the Fed could pretend runoff was not a sale simply because the Fed was not actively dumping bonds.

A Warsh Fed could therefore be structurally hawkish and tactically violent.

It may not cut because the S&P is down 5%, because MAG7 sells off, or because real estate wants lower mortgage rates.

It may cut hard if credit freezes, banks underperform in a way that signals balance-sheet stress, small caps and cyclicals collapse while defensives hide the index damage, housing weakens beyond affordability arithmetic, commodities roll over in a way that confirms demand destruction, and the two-year starts telling the Fed that policy is already too tight.

Although the market wants a put, Warsh may offer something closer to a stop-loss.

The balance sheet is the second place where Warsh could matter more than the first twenty-five basis points of rates. He has already put the balance-sheet framework under review, alongside task forces on communications, data, productivity and jobs, and inflation, and the press conference made clear that the Fed wants to look again at how policy is transmitted through both rates and the balance sheet. The old Fed treated the balance sheet as a technical backdrop once reserves were ample. Druckenmiller treated the balance sheet as liquidity itself. Those are not compatible worldviews.

If Warsh moves the Fed toward a smaller, cleaner, less permanent balance sheet, the long end should price more supply risk and less central-bank absorption, especially in a fiscal world where Treasury issuance is not a background variable but the macro variable. This does not mean he will run reckless QT or invite a repo accident, because a Fed chair who wants to restore credibility does not get there by breaking funding markets, but it does mean that the market should probably stop assuming that every downturn automatically brings back the full post-2008 playbook. QE may become a crisis tool again rather than a standing insurance policy. That would be a profound change for every asset whose valuation was built under the assumption that the Fed balance sheet would always expand faster in busts than it shrank in recoveries.

This is also where Scott Bessent matters, although the claim should be stated carefully. Bessent’s official Treasury biography says he was previously chief executive and chief investment officer of Key Square, chief investment officer of Soros Fund Management, and managing partner of Soros’s London office from 1991 to 2000, which means Treasury and the Fed are now both led by people formed, directly or adjacent to, the macro-trading world that thinks in currencies, liquidity, term premia, reflexivity, and policy mistakes rather than the academic model of a smooth output gap converging toward equilibrium.

I do not know whether Warsh and Bessent “see Druckenmiller as their hero” in the literal personal sense (I think they do, as I also do alongside George Soros), but intellectually the overlap is obvious enough: both understand that markets are not just discounting machines, they are transmission mechanisms, political constraints, and sometimes the only honest real-time economic data policymakers have.

The political problem is that this type of Fed will be accused of being political no matter what it does. Druckenmiller addressed the same problem in 2018 when Trump was publicly attacking Powell. His point was sharp: it would be terrible if the Fed paused because Trump bullied it, but it would be just as political if the Fed refused to pause because it was afraid people would think Trump bullied it. The correct answer was to remove Trump from the equation and decide based on the market, the economy, credit, liquidity, and risk-reward. Warsh will face the same trap, except with a larger fiscal deficit, a more politicized Fed, and a market that will over-interpret every silence because he has deliberately reduced the amount of formal guidance.

The market implication is that the two-year Treasury becomes the cleanest expression of whether Warsh is winning or losing. If the economy holds, inflation stays sticky, and the Fed refuses to guide cuts, the front end can stay heavier than investors want, and the entire equity market has to reprice the absence of pre-committed easing. If the economy cracks, the same front end can collapse because the Fed has not tied itself to a dot path that needs to be walked back over three meetings. This is why the Warsh Fed can be bearish for duration first and then violently bullish later. The absence of guidance creates a wider distribution, and the wider distribution itself becomes the trade.

For equities, the clean read is that multiple expansion becomes harder unless the productivity story is powerful enough to overwhelm the discount-rate story. Warsh spent real time on AI, describing it as possibly the most important general-purpose technology change of his adult lifetime, while also admitting that the Fed can count the demand side more easily than it can infer the supply side.

That makes the leadership question more interesting. The old easy-money market allowed long-duration assets, speculative growth, credit, private equity, and levered real estate to rise together because the same central-bank liquidity compressed everything. A Warsh market should be more discriminating. Secular growth with real cash flow can still work if rates ultimately fall because growth slows, and Druckenmiller himself made that point in 2018 when he talked about cloud software as long-duration growth that could be worth more in a low-growth, low-rate world. But the junk version of duration, companies that needed liquidity more than they needed earnings, should be less protected. The difference between real secular growth and liquidity-dependent growth becomes the whole game.

Credit is the place I would watch most closely because Druckenmiller’s 2018 interview treats credit as the bridge between market prices and the real economy. He did not care that the excesses were outside the banking system rather than inside it, because the economy still has to absorb a credit contraction whether the marginal lender is a bank, a CLO, a private-credit fund, or a hedge fund. That point is even more relevant now. The post-GFC system moved risk away from bank balance sheets and into private markets, asset managers, direct lenders, insurance structures, and vehicles that do not transmit stress through the same old channels. If Warsh is serious about reading market prices, he cannot just read the banks. He has to read private-credit marks, leveraged loans, HY spreads, refinancing calendars, small-cap interest expense, CRE maturity walls, and the equity prices of the firms that warehouse or originate that risk.

The dollar is more complicated. A Fed that refuses to promise cuts can support the dollar in the short run, especially if real rates remain high and other central banks are already easing, but if Warsh’s framework produces a later recognition that policy has gone too far, the repricing can move against the dollar quickly, particularly if the market begins to see the Treasury-Fed combination as one where fiscal pressure, reserve-management choices, and a more discretionary Fed all interact. The dollar trade under Warsh may be less about static rate differentials and more about whether U.S. policy credibility is being restored or whether the market starts to price a policy accident followed by a larger easing cycle.

The cleanest way to describe the regime is the following: Powell tried to manage expectations, Warsh is trying to restore price discovery. Powell’s Fed talked constantly because it believed communication was a policy tool. Warsh seems to believe too much communication turned into a market-distortion tool, and the Fed now needs to step back enough for market prices to become useful again. Which is a very Druckenmiller view, because it begins with humility toward markets rather than confidence in models.

The risk is that the Fed discovers markets have been so trained by central-bank narration that taking away the script creates more noise before it creates better signals. Druckenmiller himself said in the same interview that algos, passive flows, central banks, and suppressed volatility had damaged some of the old price-action signals he used to rely on, especially over short windows, even if the market still worked over longer horizons. That is a problem for Warsh. If he wants to rely more on market signals, he has to decide which signals are real and which are just positioning, mechanical flow, vol control, CTAs, option-market gamma, and reflexive liquidity.

My base case is that the Warsh Fed will be less predictable meeting to meeting, less willing to validate the market’s easing expectations, more focused on the left side of the inflation decimal point than the old market wants, more interested in balance-sheet reform than investors are pricing, and more willing to let financial conditions tighten if the economy and the tape are still absorbing it. But I also think that if the internal market signals turn together, especially credit, banks, cyclicals, housing, small caps, commodities, and the two-year, Warsh may move faster than Powell would have moved because he has explicitly refused to bind himself to a path.

That will be quite ironic if I turn out to be right, won’t it? The market hears “no forward guidance” and thinks hawkish uncertainty. It may get that first. But the deeper point is optionality. Warsh is taking optionality away from investors and giving it back to the Fed. He does not want the market front-running a promised easing cycle, and he does not want the Fed trapped by its own old words when the facts change. If that is right, the coming years will reward investors who stop trading the Fed’s sentences and start trading the same indicators Druckenmiller was laying out in 2018: liquidity, the curve, credit, the inside of the stock market, commodities, and the difference between what the official data still says and what the market has already begun to discount.

The Warsh Fed may not be a hawkish Fed in the simple sense. It may be a Fed trying to kill the autopilot, and for a market built on autopilot, that is a much bigger change than one meeting, one dot plot, or one twenty-five-basis-point move.

AD Fund Management LP finished the first half with extremely strong performance. I share gross returns with you every once in a while - I was turning cautious towards semiconductor names, but I think enough damage has been done to the true leaders such as - I would stay away from every other name within semiconductor names until Japanese Yen and Korean Won currencies make their move against the US Dollar.

Tech is done - the correction that occurred, and was needed in June had too fast of a recovery and we are just experiencing the ripple effects right now. / is at 50-DMA, and is down 7.5% over the past 30 trading days. That’s good enough for me, for a drawdown.

From the ratio charts above, you can see that semiconductor sector compared to technology and technology compared to rest of the market got crushed over the past couple of weeks. If this is THE top, obviously we are just getting started - however, at an index level I DO NOT THINK this is THE top.

From a rate-of-change perspective $SPX reset the momentum it had coming out of April 2026 bottom.

The main reason I am bullish US equity market is due to the fact that I am bearish US Dollar - I think DXY topped.

So, in this scenario - we should see US treasury yields drop, precious metals, bitcoin and other crypto stuff, and equities outperform. Market does not agree with me as 2-year US treasuries still price in ~3 rate hikes before end of the year, but I really don’t think that will happen.

Gold may be a good buy, I shorted the parabolic spike and lost in Q4 of 2025 but since the beginning of this year gold and silver shorts in the fund have yielded tremendous results. I wouldn’t short gold / silver anymore at these levels.

This equal weight outperformance is becoming a thing for via and va - I would fade it, the world works in market cap weight terms. There are short bursts of outperformance equal weight can capture, but I think it’s simply because MAG-7 names did not move since the beginning of the year.

Nasdaq-100 outperformance vs S&P 500 since May 1, 2026 vanished over the past two weeks.

has been a literal trash stock versus (semiconductor ETF). I think this changes in 2H-2026.

As I said, weak dollar should help - bitcoin got crushed against year-to-date.

I would stay away from emerging markets. I never understood it, the only growth channel in the world is via AI CapEx cycle led by the US.

July is a bullish month, and we are off to a terrible start. That said July is a terrible month for momentum names - this year, the same thing happened.

This momentum unwind coupled with flows (lack there of) also makes me more bullish equities.

As I said earlier, I anticipate yields to come down across the curve - the move started with the payroll numbers in the AM on July 2, 2026.

The next big events are CPI / PPI numbers - I anticipate them to further confirm the deflationary environment we have entered following crude oil pull back.

Happy 4th of July - I feel lucky and privilaged to be able to live where I live, do what I do, and not worry about many things individuals across the world worry about day by day.

I have been following saga for a long time now, and Bill Ackman owns ~20% of the float in commons. The stock is down 60% year-to-date, and was up 40% as recently as April 2026.

Hertz changed in the past week because the stock broke on real residual-value news and then absorbed a highly technical financing event that added hedge-related short supply without providing common shareholders equity proceeds. The market is no longer asking whether rental demand is okay, demand is good enough. The question is whether residual values and depreciation improve quickly enough for Hertz to earn through a levered balance sheet before new capital, secured creditors, and exchangeable-note dilution take most of the upside.

This past week, Hertz said Q2 fleet size, revenue, revenue per day, and rental days were broadly in line with or slightly above prior expectations, but used-car market softness caused the company to realize losses on vehicle sales in May after gains in April. Management now expects Q2 net depreciation per unit, or DPU, of roughly $300 per vehicle per month and Q2 Adjusted Corporate EBITDA of $50 million to $80 million, still within the prior range but toward the low end.

The problem is, Hertz launched a financing built around $350 million of 6.75% exchangeable senior first-lien secured PIK notes due 2030, upsized from $300 million, with proceeds intended mainly to repay revolver borrowings and for general corporate purposes. The notes pay 3.375% cash interest and 3.375% PIK, mature July 1, 2030, and are exchangeable at an initial price of roughly $3.58 per share, a 32.5% premium to the $2.70 borrowed-share offering price.

At the same time, Hertz loaned 37.0 million shares to J.P. Morgan at $2.70. The share borrower planned to sell those shares so investors in the exchangeable notes could hedge. Hertz gets no proceeds from the share sale other than a nominal lending fee, and the borrowed shares must eventually be returned. In the meantime, the structure creates immediate tradable supply and hedge-related short pressure.

Long story short, the stock’s collapse was partly fundamental because residual values matter directly to Hertz’s P&L and liquidity. But the extreme volume and price pressure were also mechanical. The borrowed-share sale and note hedging created a forced seller dynamic at exactly the moment the market was repricing depreciation risk.

1. Macro and industry regime

The macro regime is mixed for Hertz, but it has deteriorated for the equity because the company’s weakest variables are still residual values, credit costs, and liquidity, not top-line travel demand.

Rates remain a headwind. The Fed held the federal funds target range at 3.50% to 3.75% in June, leaving Hertz exposed to high fleet-financing and corporate interest costs even after recent refinancing activity. Hertz paid $136 million of vehicle-debt interest and $137 million of non-vehicle debt interest in Q1 alone, so higher-for-longer rates still flow directly into the capital structure.

Consumer travel demand remains supportive. TSA checkpoint data showed 2.93 million passengers on June 28, 2026, which indicates air travel volumes remain high heading into peak summer. U.S. Travel expects total U.S. travel spending to reach $1.37 trillion in 2026, with domestic travel spending back to 2019 levels in real terms and inbound travel helped by major events such as the World Cup. (U.S. Travel Association) That supports rental days, airport demand, and summer utilization.

Corporate travel is less clean. GBTA’s April survey showed only 30% of respondents expected business trips to increase in 2026, down from 35% in January, while 28% expected trip volume to decline, up from 16% in January. (GBTA) That does not kill the Hertz thesis, but it limits the ability to underwrite a broad corporate-volume rebound as the central driver.

Used-car pricing is the key pressure point. The Manheim Used Vehicle Value Index rose 0.3% month over month and 3.6% year over year in May, while non-adjusted wholesale prices fell 1.2% from April but remained 3.1% above year-earlier levels. (Cox Automotive Inc.) Mid-June Manheim data improved again, with the index up 0.6% from May and 2.6% year over year. The broad index is not collapsing, which makes Hertz’s May loss-on-sale disclosure more nuanced. Either the company’s disposal mix, channels, EV exposure, model-year composition, or rental-fleet-specific residuals are worse than the headline index, or May was a temporary pocket of weakness.

EV residuals remain a separate drag. Cox data showed used EV listing prices down 6.1% year over year in March, with the used EV premium over ICE and hybrid vehicles narrowing sharply. (Cox Automotive Inc.) Hertz has been shrinking EV exposure, but the prior EV strategy damaged depreciation, resale economics, and investor trust. Until EV residual pressure is fully washed through the fleet, the market will haircut management’s DPU targets.

Rental-car pricing is the positive side of the regime. Hertz reported Q1 revenue per day of $57.38, up 5% year over year, and revenue per unit of $1,353, also up 5%. (Securities and Exchange Commission) The market is not giving Hertz much credit for pricing because depreciation and financing overwhelm it, but the pricing data are not broken.

ABS and fleet-credit markets remain open but conditional. Hertz had $11.95 billion of vehicle debt and $6.25 billion of non-vehicle debt at March 31, 2026. (Q4 ) The company also had $1.86 billion of remaining vehicle-debt capacity, but none was available under borrowing-base limitations at quarter-end, which means fleet financing exists only if collateral and advance-rate conditions cooperate. (Q4 )

The regime is improving for revenue and utilization, stable-to-mixed for travel, but still adverse for equity because residual values, interest expense, and secured-credit availability dominate the equity math.

2. Business model mechanics

Hertz makes money by converting a financed vehicle fleet into rental days at prices high enough to cover depreciation, direct operating expense, vehicle interest, SG&A, and corporate interest. The model looks simple at the counter and highly levered in the accounts.

The revenue formula is:

Average rentable vehicles × utilization × rental days × revenue per day

In Q1, Hertz had 514,163 average vehicles, 493,359 average rentable vehicles, 79% utilization, 34.9 million transaction days, and $57.38 revenue per day. That produced $2.0 billion of revenue, up 11% year over year. (Securities and Exchange Commission)

The cost stack matters more than the revenue line. The major operating variables are:

Q1 makes the unit economics clear. Hertz generated $1,353 of monthly revenue per unit and incurred $312 of monthly net depreciation per unit. Direct operating expense was $38.52 per transaction day. At 79% utilization, a vehicle is rented roughly 24 days per month, implying about $913 of DOE per vehicle per month. That leaves roughly $128 per vehicle per month before vehicle interest, SG&A, corporate interest, and other below-the-line items.

That is why the equity is so sensitive to DPU. A $50 per month change in depreciation looks small at the vehicle level but is enormous at fleet scale. At roughly 514,000 vehicles, $50 per month equals more than $300 million of annual EBITDA sensitivity before considering second-order effects on liquidity, borrowing base, and lender confidence.

This is also why Hertz can look operationally better while the equity still trades like a distressed claim. RPD can improve, rental days can improve, and utilization can improve, but if depreciation, vehicle interest, and corporate interest absorb the incremental revenue, common equity gets little credit.

3. Capital structure and liquidity

The capital structure is the core risk. Hertz equity holders are financing a turnaround beneath a large secured and fleet-financed balance sheet. The business is not out of cash, but the equity does not control the capital structure.

At March 31, 2026, Hertz had $18.2 billion of total debt, split between $11.95 billion of vehicle debt and $6.25 billion of non-vehicle debt. (Q4 ) Vehicle debt funds the rental fleet and is tied to fleet collateral. Non-vehicle debt is the corporate stack that common equity must ultimately earn through.

The non-vehicle debt stack at Q1 included roughly:

Hertz ended Q1 with $837 million of corporate liquidity, consisting of $583 million of cash and cash equivalents and $254 million of revolver availability. (Q4 ) The company also disclosed it was in compliance with all covenants at March 31, including first-lien leverage and minimum liquidity tests. (Q4 )

The June exchangeable-note financing improves runway but does not delever the company. The $350 million of new first-lien secured PIK notes are intended mainly to repay revolver borrowings. That should restore revolver availability and improve liquidity optics, but gross secured claims rise and PIK interest accretes.

The financing also adds dilution risk. Existing exchangeable notes already represented up to 166.8 million shares issuable upon exchange, according to the prospectus. The new $350 million exchangeable notes are initially exchangeable at roughly $3.58 per share, equivalent to about 97.8 million shares before the 19.9% shareholder-approval cap.

Basic shares outstanding were 315.1 million at March 31. On a fully diluted stress view, common holders need to think beyond today’s basic market cap. Existing exchangeables, the new exchangeables, RSUs, PSUs, and potential stock-settled instruments can push effective dilution materially higher if the turnaround works. That is the usual distressed-equity bargain: if the business fails, equity goes down; if the business works, part of the upside funds the securities that kept the company alive.

The right framing is this: Hertz is liquid enough to attempt the turnaround, but common equity sits below a capital structure that can absorb a large share of the recovery if residual values do not normalize quickly.

4. Fleet depreciation and residual-value sensitivity

Depreciation per unit is the fulcrum. Hertz reported Q1 net DPU of $312 and guided Q2 to about $300 after May used-car losses. Management’s “North Star” has been sub-$300 DPU, revenue per unit above $1,500, and DOE per day in the low $30s.

Using Q1 average vehicles of 514,163, the sensitivity is:

This is the cleanest way to underwrite HTZ. At the current market cap, a $50 to $100 per month DPU swing can be worth more than the entire equity value, depending on whether the market believes the improvement is durable.

The breakpoints are:

The cash-flow timing is less clean than the EBITDA math because depreciation, estimated residuals, vehicle sales, and ABS borrowing bases do not all move at the same time. But the economic direction is direct. If the used fleet sells above depreciated value, Hertz gets gains and borrowing-base support. If vehicles sell below expected values, DPU rises, losses hit EBITDA, collateral value weakens, and liquidity tightens.

The long thesis needs DPU to move below $300 and stay there. A single quarter at $300 does not prove normalization. A second consecutive quarter with DPU below $300, positive vehicle-sale execution, and stable Manheim data would change the equity debate.

5. Operating turnaround

The operating turnaround is real in some areas and unproven in the one that matters most.

Q1 revenue rose 11% year over year to $2.0 billion, RPD rose 5%, RPU rose 5%, and adjusted corporate EBITDA improved versus the prior year despite still being negative. Utilization was 79%, and management said recall activity was unusually high, with a meaningful impact on rentable fleet, transaction days, revenue, and EBITDA.

The fleet refresh also is quite valuable. Hertz has been trying to reduce prior fleet mistakes, lower EV exposure, improve model-year mix, and normalize disposition economics. Q1 depreciation fell from the prior year, helped by lower disposition costs, lower capital costs on newly acquired vehicles, and stronger expected residual values at disposal.

The credibility problem is the June update. Management had been guiding investors toward sub-$300 DPU in 2026, then had to tell the market that May vehicle sales flipped from gains to losses and Q2 EBITDA would land toward the low end of the prior range.

Cost control is also not yet decisive. DOE per transaction day was $38.52 in Q1, while management’s target is low $30s. If Hertz can move DOE toward the low $30s while maintaining RPD and improving DPU, the equity torque is large. If cost savings are offset by recalls, utilization friction, higher maintenance, or weak residuals, the operating turnaround remains a guidance story.

In summary, the market is not being irrational in withholding credit.

6. Equity valuation

At roughly $2.30, Hertz’s market cap is around $722 million. Using Q1 non-vehicle debt of $6.25 billion and cash of $583 million, corporate net debt was roughly $5.66 billion before the June financing’s revolver repayment mechanics. (Q4 ) That implies a corporate enterprise value around $6.4 billion at the current stock price.

Vehicle debt should not be treated the same way as ordinary corporate net debt in an EBITDA multiple because it finances the fleet assets that generate revenue, but it absolutely matters for liquidity, collateral, and residual-value risk. The common stock is a residual claim on a highly financed operating system.

A practical valuation grid:

What does the market price at $2.20? It prices survival optionality, not durable recovery. At $2.20, basic market cap is about $725 million. With corporate net debt above $5.5 billion, the market is implying either that normalized EBITDA is far below historical bull-case expectations, that much of the recovery will be consumed by creditors and dilution, or that the probability-weighted equity value includes a meaningful restructuring or sub-$1 outcome.

What gets the stock to $1? A $1 stock is roughly a $315 million basic market cap. That happens if Q2 final EBITDA misses the $50-$80 million guide, DPU remains above $325 into Q3, liquidity falls toward covenant-sensitive levels, or the company needs another dilutive financing before the market sees real operating recovery.

What sustains $2.20? The stock can hold around entry if Q2 lands within guidance, June and July used-car data do not roll over, liquidity improves after the June note proceeds repay revolver borrowings, and the market concludes the borrowed-share pressure is technical rather than permanent equity impairment.

What gets the stock to $5? A $5 stock is about $1.6 billion of basic equity value. To support that fundamentally, Hertz likely needs a credible path to more than $1.1 billion to $1.3 billion of normalized Adjusted Corporate EBITDA, depending on the multiple and corporate net debt assumption. That requires DPU below $300, RPU moving toward management’s target, DOE improvement, and no near-term repeat financing.

What gets the stock to $10? A $10 stock requires more than $3.1 billion of basic equity value and much more on a diluted basis. With corporate net debt around $5.5 billion, that means the market must believe Hertz can generate roughly $1.3 billion to $1.5 billion-plus of durable EBITDA, avoid meaningful additional dilution, and regain enough creditor confidence that the equity receives a real multiple. That is not a 3-month base case. It is a full normalization case.

Basic shares were 315.1 million at Q1, but existing exchangeables represented up to 166.8 million shares, and the new exchangeable notes add another large contingent claim subject to shareholder-approval caps. Important to remember, a $5 stock on basic shares is not the same as a $5 stock on a fully diluted capital structure.

7. Positioning and technicals

The positioning setup has squeeze ingredients, but the squeeze needs a fundamental trigger.

Reported short interest was 54.4 million shares as of mid-June. Vendor estimates differ on the percentage of float because float definitions differ, but the share count sold short is large and days-to-cover was roughly 10 days on the cited average-volume base.

The borrowed-share financing complicates the tape. The 37.0 million borrowed shares create immediate hedge-related supply, and Hertz explicitly disclosed that sales, short sales, or derivatives activity by the share borrower and note investors could affect the stock price.

The setup can squeeze if four things happen together: Q2 results land at or above the $50-$80 million EBITDA guide, management gives better Q3 commentary, used-car data stabilize, and the market absorbs the borrowed-share hedging flow. Without those, high short interest is simply a reflection of an overlevered residual-value trade.

I won’t touch on technicals, this is a penny stock trading below every single imaginable moving average - if this goes, it’ll go due to fundamental reasons and the equity portion will rerate.

8. Catalysts over the next 3 to 6 months

The next hard catalyst is Q2 earnings. The market will compare actual results against the $50-$80 million Adjusted Corporate EBITDA guide and the roughly $300 DPU guide. The critical detail will be June commentary: if May was the low point and June improved, the stock can work; if May weakness continued, the stock likely trades as impaired.

Used-car data is important to track also. July Manheim data, mid-month updates, non-adjusted auction trends, and EV resale pricing will be directly tied to the DPU thesis. The broad index’s resilience in May and mid-June is currently one of the few external supports for the long case.

Financing updates matter almost as much as earnings. Investors need to know whether the $50 million additional note option is exercised, how much revolver capacity is restored, whether liquidity improves as expected, and whether any further secured financing or equity-linked financing is needed.

Finally, rating-agency actions matter because the equity is dependent on credit-market access. S&P had already downgraded Hertz to B- with a negative outlook earlier in 2026, citing weaker metrics and tighter liquidity. Any further downgrade would pressure refinancing confidence; stabilization would help the equity.

9. Conclusion

As I said, this is one of those opportunities that can double in a week or drop 50% in a week. The equity portion of is a gamble, but a potential lucrative gamble where I see asymmetric risk/reward to the upside. Bill Ackman himself has the capital pool to backstop this company to an extent.

This is not a balance-sheet value investment at today’s price. It is a levered equity option on DPU normalization, liquidity extension, and summer operating leverage. The June financing bought time, but it also showed that time is expensive. The stock can double if Q2/Q3 prove the business has stabilized, but it can also trade below $1 if DPU remains stressed and liquidity starts compressing again.

The stock is cheap because the market is treating common equity as a subordinated claim beneath an uncertain fleet asset base. That may be too punitive, but the burden of proof is now on Hertz to show that May was a residual-value air pocket, not the beginning of another capital-structure squeeze.

New York did not wake up overnight as a socialist city. It did not collectively decide that every complicated problem in American life can be reduced to rent, ICE, corporate greed, and Israel. What happened is more mechanical and, over time, more dangerous. A highly organized political faction found the weak point in the system.

1. Low-turnout primaries

2. Safe Democratic districts.

3. Closed-party rules.

4. A distracted, exhausted electorate.

A city where the general election often functions as ceremony, while the real transfer of power happens in June, among a small and unusually motivated slice of voters.

Mayor Zohran Mamdani’s endorsed candidates are now turning New York’s congressional primaries into a test of whether his movement can move from City Hall into Washington. Brad Lander’s projected defeat of Dan Goldman in NY-10 was the most visible result. Claire Valdez’s strength in NY-7 and Darializa Chevalier’s challenge to Adriano Espaillat in NY-13 show the same structure repeating across districts. These are part of a disciplined political strategy built around activist density, message repetition, volunteer infrastructure, social pressure, and a clear understanding that low-turnout elections reward intensity over breadth.

The establishment still behaves as if politics is a persuasion contest among normal voters. The new left understands that politics, especially in one-party cities, is often a turnout contest among abnormal voters.

I do not think the underlying issues are fake. Rent is crushing New York. Corporate concentration is real. Immigration enforcement can be arbitrary, cruel, and politically abused. The war in Gaza has exposed serious moral failures across governments, institutions, and media. A serious politics would treat all of these as hard problems. It would ask why housing supply remains structurally constrained in the richest city in America. It would ask how to enforce borders without destroying human dignity. It would ask why corporate power has become so concentrated and why both parties have enabled it. It would ask how America should defend allies, restrain allies, and maintain moral credibility in a world where every local conflict now becomes a domestic identity war. Obviously, that is not the politics being built here.

The politics being built here is much simpler. Rent is high because landlords and corporations are greedy. Immigration enforcement is illegitimate because borders are immoral. Corporate profit is exploitation. Israel is not a state making wartime decisions, it is a symbolic villain through which every other grievance can be organized. The moral framework is totalizing. Once someone adopts the wrong position, or even the wrong vocabulary, they are no longer mistaken. They are corrupt, bought, racist, fascist, complicit, colonial, genocidal, or some combination of the above. This is where the movement becomes dangerous.

The United States can obviously survive radical candidates. It has survived worse. What it cannot easily absorb is a generation of political actors trained to believe that governing is the process of identifying villains and redistributing moral blame. The next thirty years will be shaped by people now in their twenties, thirties, and forties moving into Congress, agencies, courts, universities, media companies, foundations, unions, city councils, and school boards. If their intellectual formation is grievance first, institution second, America will pay for it slowly and then suddenly.

The damage will not show up immediately. It rarely does. Cities decay through compounding. First the language changes. Then the incentives change. Then competent people stop wanting to serve. Then private capital hesitates. Then the tax base becomes more fragile. Then public services degrade. Then the same faction that helped create the institutional weakness argues that the failure proves the need for more control.

New York is especially vulnerable because it is rich enough to hide bad politics for a long time. The new left sees that inheritance as something to harvest. It did not build the machine, but it knows how to occupy the control room.

The establishment’s failure is equally important. This did not happen because the left is brilliant and everyone else is helpless. It happened because moderate Democrats became lazy inside their own monopolies. They assumed endorsements, money, incumbency, and institutional respectability would be enough. They outsourced voter contact to consultants. They spoke in the language of committees, seniority, and “getting things done,” while younger voters heard only a defense of a status quo that does not work for them.

You cannot beat an organized movement with vibes about competence. You cannot beat an affordability message by telling voters the other side is too extreme. You have to explain why rent is high, why the city does not build enough, why public-sector costs are out of control, why permitting is broken, why crime and disorder tax the poor first, why capital leaves when it is treated as an enemy, and why a city that punishes productivity eventually runs out of other people’s money. You have to fight on the actual terrain.

The right also misunderstands this. Republicans look at New York and see proof that blue cities are insane. That may be emotionally satisfying BUT it is analytically insufficient. The more important lesson is procedural. Any faction that can dominate a low-turnout primary can capture institutions disproportionate to its actual popularity. That is why turnout is the battlefield.

When only a small share of voters participate, politics stops reflecting the median citizen and starts reflecting the most organized citizen. The activists need to persuade, mobilize, or pressure enough voters in the relevant primary electorate. In a safe district, that is the whole game. Win June, govern November, claim democracy, repeat.

This is how fringe ideas become institutional common sense. The word “democracy” becomes cover for an organizational arbitrage.

I am not sure there is a quick fix. Open primaries would help at the margin. Better election timing would help. More competitive general elections would help. Civic education would help. Serious candidates who speak plainly about affordability, order, growth, and national cohesion would help. But none of this solves the deeper problem, which is that a large share of the electorate has withdrawn from the boring maintenance work of self-government, while the most ideological people have made politics their identity. That imbalance is what the new left is exploiting.

I don’t know - I don’t have the power to change these dynamics. I am just observing from distance fueled with dissappointment…

Thirty years from now, the people formed in this environment will be presidents, senators, agency heads, judges, university presidents, prosecutors, foundation officers, media executives, and members of Congress. Some will moderate. Some will learn. Some will govern better than they campaign. But many will carry forward the habits they are being rewarded for now: accusation over analysis, purity over prudence, mobilization over persuasion, and contempt for the country they want to control. That is the part I find hardest to watch.

In my deep-dive published on January 24, 2026 about Intel - I wrote the following: “I sold TSM -1.59%↓ early January 2026 and bought INTC 0.76%↑. If you are underwriting both companies in a model, and are purely focused on fundamentals - you’d regard this trade as extremely stupid. However, as I was thinking ways to hedge the current AI heavy equity portfolio I realized that the only true hedge to China/Taiwan escalation is owning Intel, now backstopped by Trump administration.”

What a paragraph!

As the first half of 2026 is ending, I am thinking about how to position the portfolio. Afterall, ADFM is a global macro fund with an equity sleeve - not a levered long AI fund. I’ll be honest with you, I am bearish on semiconductor names and it has been painful over the past 10 days or so. $KOSPI corrected ~25% to 50% in the first two weeks of June and then made a remarkable recovery.

I will share some charts to support my thesis, and build the theory as to why I am turning bearish semiconductor names as the article continues.

A snippet from the baskets we track at ADFM - the top baskets are literally all AI names, and AI alone. Earlier this year, there were some energy names, and commodity chemicals names but they are all gone now.

Relative to , on a 3 and 6 month basis, basically no sector outperformed except for (Technology sector).

Just absolute, once in a lifetime for momentum / high-beta factors that are usually within the broader technology sector.

Price, rate of change, and acceleration at 2000 (Dot-com) levels, right before the bust.

The true leader, , of the AI regime giving back outperformance since January 2024… I can keep going to add more supporting charts to my claims BUT I think this is enough.

Importantly, I am an AI bull and I caught the bull market - so my opinions are different from others who have been shouting for market participants to short this rally over the past 3 months to 2 years. You would’ve gone bankrupt, if you shorted this move. Reminds me of the period where Stanley Druckenmiller lost $3 billion on tech stocks during the 2000 bubble:
"I didn't learn anything. I already knew I wasn't supposed to do that". Full anecdote below:

The market needs a catalyst for the semiconductor sector to pullback.

The catalyst, in my view, is the financing of the trade. Everyone is looking for the wrong thing. They are looking for an guide-down, an AI CapEx warning from , a semiconductor inventory problem, or some fundamental crack that allows them to say, “Okay, that was the top.” That may eventually happen, but I do not think it needs to happen first. In 2024, when and / at large pulled back almost 50%, the catalyst was that the funding structure underneath global risk assets became unstable. That is the part I think the market is underpricing again.

I think July/August 2026 will follow the July/August 2024 template. The yen had been the cheapest funding currency in the world. Investors borrowed in yen (sold yen short), bought higher-yielding currencies (bought USD long), bought U.S. equities, bought Japan equities, bought basically anything with positive carry and low realized volatility. Then the yen stopped weakening because the Bank of Japan turned more hawkish. The Federal Reserve was less dovish than markets wanted. A U.S. labor report came in soft enough to trigger recession fears. None of these events, in isolation, should have caused a global liquidation. But, in 2024, together, they hit a market that had become one giant volatility-short, yen-funded, momentum-long expression.

What happened in practice was; the yen rallied and the carry trade lost money. Volatility spiked, and higher volatility forced deleveraging. Deleveraging forced selling of the most liquid winners. The most liquid winners were U.S. technology and semiconductors. Basically, a mechanical unwind due to modern market structure dynamics.

As of June 22, 2026 - USD/JPY is back near the zone where Japan can no longer pretend the currency is just a market variable. A weak yen supports exporters and nominal equity earnings, but at these levels it also imports inflation, pressures households, raises energy costs, and creates a political problem. Japan can tolerate a weak currency when inflation is dead. Japan cannot tolerate a disorderly weak currency when inflation expectations are rising and the BoJ is already trying to normalize policy. That leaves Japan with two bad choices: accept further yen weakness and import more inflation, or push back through intervention, rate hikes, or language strong enough to force speculators to cover. Both paths are bad for the carry trade.

If USD/JPY keeps rising, the intervention risk increases. If Japan intervenes, the yen rallies and carry books get hit directly. If the BoJ tightens again, the funding cost rises and the forward carry changes. If the Fed stays hawkish while the dollar rises against the rest of the world, the pressure shifts into high-beta FX, Korea, Taiwan, EM, commodities, and global cyclicals. In that scenario, the dollar can remain strong while the yen also squeezes higher against the dollar. That is the nightmare for global risk positioning, because the funding currency rallies and the investment currencies weaken at the same time.

And THAT I think is the catalyst. I am bearish $DXY (USD) AND I am bullish $JPY. Back to semiconductors: the semiconductor trade is no longer just a fundamental AI trade. It has become the cleanest expression of global liquidity, dollar liquidity, momentum, and the belief that the U.S. capex cycle is immune to macro. Nothing is immune to macro, I do not believe it.

When a trade becomes that successful, it stops being owned only by fundamental investors. It becomes owned by momentum funds, vol-control funds, pod shops, retail, systematic trend, factor books, index buyers, and macro funds using the sector as a proxy for global risk appetite. Success changes the liquidation profile of the thesis. A long-only investor can hold Nvidia through a drawdown because he believes in Blackwell, Rubin, data centers, sovereign AI, and the multi-year compute shortage. A levered momentum book cannot. A yen-funded risk book cannot. A vol-control strategy cannot. A pod that is down 300 basis points on the month cannot.

As I conclude the article, I want to be clear about what I am not saying. I am not bearish on AI. I still think AI is one of the biggest investment regimes of my lifetime, and the fund benefited from being early and aggressive. But macro funds are paid to recognize when a winning theme becomes a crowded expression of liquidity. This is that moment. The AI thesis can remain intact while the semiconductor stocks fall 20%, 30%, even 40% if the financing conditions change.

In 2024, the market learned this lesson in one week and then forgot it because prices recovered. My view is that the recovery made the lesson more important, not less. The market now believes the carry unwind was a contained accident. I think, and hope, it was a preview -

"Beware the Ides of March" is the infamous warning given to Julius Caesar by a soothsayer, forecasting his assassination on March 15, 44 BCE. I say “Beware the Ides of July”.

The story of Quantum is usually told through Soros:, at least I’ve been doing that since it is the story I know. A new revelation to me; Foglia gave Soros a trading account to manage before the world knew his name. The returns were staggering, reportedly 50 to 150 per cent a year, and Foglia became a seed investor in Soros’s early funds before Quantum launched in 1970.

Foglia had been born into finance in Milan in 1928. His father, Antonio, was a stockbroker who had made substantial profits by investing in the US market near its Depression lows after visiting America in the early 1930s, then used his home during the Second World War to host secret meetings for leaders of the Italian resistance. After the war, Antonio became president of the Milan Stockbrokers’ Executive Committee. When he died in the 1950s, Alberto and his brother Giambattista inherited the family brokerage firm, a set of financial relationships, and the problem of turning a domestic Milanese business into something international. They began writing research on Italian stocks in English for foreign investors, long before cross-border capital became the default language of markets.

Every great manager needs some seed capital before the rest joins to sustain the operations of the business. Before the consultants, before the databases, before the performance screens, there is often one person with enough taste and nerve to recognize a differentiated mind early. Foglia was that person for Soros. He was client, backer, partner, board chair, adviser, and a friend.

As Quantum grew, Foglia pushed for a more formal board structure, with serious investors around the table and a neutral chair strong enough to challenge and steady the machine. Soros’s edge was never bureaucracy, and based on my readings, it was Foglia that convinced him to establish such structure.

The 1992 sterling trade also carried Foglia’s fingerprints. Before Soros and Druckenmiller broke the Bank of England, he and Foglia had discussed a smaller trade against the Italian lira. It worked. The lira trade became confirmation that gave Quantum fund enough conviction and profits under its belt in currencies to size up the British pound trade.

Foglia did not seek celebrity. He backed others too, including Michael Steinhardt and Julian Robertson, and helped bring European capital into an industry that was still being invented. His influence sat upstream of the visible trade, in the judgment of who deserved capital, when trust should be extended, and how raw investment talent could be institutionalized without being diluted.

The hedge fund industry was built by managers who saw mispricings and exploited them for profits, but also by allocators who saw managers before the numbers made them obvious. Foglia’s is an interesting man. He saw Soros early, and stayed loyal through cycles. Wishing such limited partners for all hedge funds.

Thiel called this the return of the “question of human nature,” and that phrase remains the strongest bridge between the post-9/11 world and the age of artificial intelligence, because AI is now forcing the West to reopen the same buried questions under conditions that are more technologically powerful, more socially pervasive, and far harder to localize than terrorism. September 11 exposed the ability of small groups to project violence through modern systems; AI exposes the ability of small groups, large platforms, states, and eventually autonomous systems to project interpretation through modern society itself. The danger now runs through markets, media, labor, education, warfare, surveillance, identity, and politics, because intelligence is becoming infrastructure, persuasion is becoming software, and reality formation is becoming a contest between human judgment, institutional authority, model output, capital flows, and machine-generated narrative.

The better way to revisit Thiel’s essay is therefore through George Soros and Karl Popper, because the AI age is less a Straussian moment than a reflexive one, where the central issue is no longer only what liberalism must conceal to survive an enemy outside its assumptions, but how an open society can preserve correction, criticism, and reality-testing when machines begin producing the beliefs, images, forecasts, rankings, and simulations that shape the world they claim to describe.

Popper gives us the institutional and moral architecture of the open society, built on the premise that no one owns final truth and that every concentration of power must remain exposed to criticism.

Soros gives us the operating mechanism, because reflexivity explains how perceptions alter reality, how reality then alters perceptions, and how the loop can compound into booms, crashes, revolutions, panics, wars, and regime shifts. The age of AI is the age in which reflexivity moves from markets into the entire structure of civilization.

Thiel’s Break With Liberal Comfort

Thiel’s argument begins with the incompleteness of the liberal settlement. In his account, modernity was born from exhaustion, not innocence. Europe’s religious wars had made ultimate questions too dangerous for politics, so the modern state lowered the temperature by privatizing religion, moderating virtue, narrowing the aims of politics, and redirecting ambition into property, commerce, rights, self-preservation, and procedural order. John Locke becomes central because Locke offers the American compromise in philosophical form: a world where the state stops trying to answer the question of what human beings are for and instead protects the conditions under which individuals can pursue their own goods, accumulate property, trade, worship privately, speak publicly, and avoid being killed over metaphysical disagreement.

That compromise was one of the great civilizational achievements of the modern West, but Thiel’s insight is that it came with a blind spot that became harder to see precisely because the system worked so well. Liberal capitalism created immense wealth by removing the most explosive questions from public life. It treated religious conviction, sacrificial violence, heroic ambition, and metaphysical hatred as residues of older worlds that would recede under the pressure of development. It assumed that politics could be stabilized through incentives, that violence could be contained through state capacity, and that history would gradually discipline irrational actors into the language of interest. September 11 mattered because the attackers did not behave like poor material victims seeking redistribution, nor like ordinary criminals maximizing personal gain, nor like rational state actors seeking a bargain; they behaved like men whose hierarchy of values placed death, faith, revenge, and transcendence above the liberal goods of safety, comfort, and prosperity.

Thiel’s essay turns from Locke to Carl Schmitt because Schmitt forces the reader to confront the part of politics that liberal modernity tries to metabolize away. Schmitt’s friend-enemy distinction is ugly, but it captures the pressure that appears when a society organized around negotiation, rights, markets, and procedural neutrality faces an adversary that experiences politics as existential conflict. Thiel sees why Schmitt matters, because a liberal society that cannot name a real enemy may lose the capacity to defend itself; he also sees why Schmitt is poison, because a civilization that defines itself through enemies risks becoming a mirror image of what it fights. The West can defeat its adversary militarily and still lose the deeper struggle if the defense of liberal civilization requires abandoning the habits, restraints, and openness that made that civilization worth defending.

Leo Strauss enters Thiel’s essay as the figure of prudence, secrecy, hierarchy, and esoteric caution. Strauss offers a way to acknowledge that liberalism is incomplete without rushing into Schmittian absolutism, because he understands that some truths about founding, violence, human nature, and political order may be too dangerous to state plainly inside the public language of a liberal regime. Thiel is clearly drawn to this possibility because the visible machinery of democracy often looks too slow and too exposed when faced with existential threats, while intelligence networks, classified systems, elite coordination, and hidden supplements can appear more serious than parliamentary debate and public reassurance. The Straussian temptation is strongest whenever liberal societies feel endangered, because the argument for secrecy always arrives dressed as responsibility.

That temptation becomes much more dangerous in the age of AI. If the answer to technological risk becomes elite management through opaque systems, then artificial intelligence gives the Straussian impulse its most powerful tool. Classification can expand, surveillance can deepen, platform governance can merge with state security, and public debate can remain formally alive while the decisive decisions migrate into model weights, cloud contracts, export controls, risk committees, defense systems, and private-public coordination structures. The most dangerous AI regime for an open society may be one that preserves liberal vocabulary while moving judgment into systems the public cannot inspect, challenge, or understand.

Popper Against the Priesthood of the Model

Karl Popper begins from fallibility rather than hidden wisdom. His defense of the open society was formed against the great closed systems of the twentieth century, where leaders, parties, philosophers, and bureaucracies claimed to understand the laws of history and used that claimed knowledge to justify coercion. Popper’s enemy was the prophetic mentality, the belief that history has a discoverable direction and that a sufficiently enlightened elite can reorganize society around that knowledge. The danger did not disappear with communism or fascism; it has reappeared in technical form whenever institutions treat model output as neutral truth, bureaucratic ranking as objective judgment, and statistical prediction as moral or political authority.

AI creates a new temptation toward historicism because it gives prophecy the appearance of computation. A model predicts default risk, student potential, military escalation, criminal behavior, worker productivity, social instability, market direction, disease spread, political persuasion, or national security threat, and the output arrives with the cold authority of quantitative systems rather than the emotional language of ideology. The institution then acts on the prediction, the person or market or adversary responds to the action, and the changed world becomes evidence for the next iteration of the model. What begins as a forecast becomes part of the causal chain. What begins as a ranking becomes an allocation of opportunity. What begins as a risk score becomes a life path.

A Popperian AI doctrine would refuse to let any model occupy the position of priesthood. Every important AI system should be treated as a conjecture embedded inside an institution, with identifiable human responsibility, visible standards, adversarial testing, appeal rights, competing models, audit trails, and a clear distinction between prediction and judgment. The issue is not whether AI systems will make errors, because every system of knowledge makes errors; the issue is whether the errors can be discovered before they harden into power. A closed model used by a bank, court, university, employer, insurer, military agency, or intelligence service can become a private lawgiver if the affected person has no way to know why the decision happened, no way to contest the evidence, and no way to force the institution to admit uncertainty.

Closed societies will appear to have an early advantage because AI rewards scale, data access, surveillance capacity, and centralized deployment. A state that can fuse identity, payments, cameras, speech, travel, education, health records, employment, and policing into one political nervous system can deploy AI faster than a constitutional democracy that has courts, journalists, lawsuits, privacy rules, opposition parties, and public embarrassment. That speed will look formidable during the first phase of the AI race, especially in security, industrial policy, and social control. The weakness appears later, because closed systems punish inconvenient feedback, suppress bad news, and train their own models on politically filtered reality. The state that knows everything except what its people are afraid to say eventually becomes powerful and blind at the same time.

Open societies look inefficient because they argue, leak, litigate, overcorrect, and expose their failures in public. Yet the open society’s weakness is also its strategic reserve, because criticism is how reality reenters the system. Popper’s point was never that democratic societies are wiser by nature or morally immune to delusion; his point was that institutions built for criticism can remove errors without requiring the destruction of the whole order. In the AI age, that capacity becomes more valuable because the velocity of falsehood will rise, the plausibility of synthetic evidence will improve, and the institutional temptation to outsource judgment to models will intensify. The open society’s edge will come from faster invalidation, not from greater purity.

That distinction is quite important for policy. The answer to AI risk cannot be a soft libertarian refusal to govern, because the technology has obvious implications for fraud, cyberattacks, biosecurity, elections, weapons systems, labor displacement, market manipulation, and state capacity. The answer also cannot be an open-ended transfer of authority to security agencies and platform monopolies, because that would convert a technological challenge into a political settlement in favor of hidden rule. A serious open society must build AI governance that can act quickly while remaining challengeable, that can defend society without treating public oversight as childish, and that can use advanced systems without allowing technical experts to become a priestly class insulated from criticism.

Soros and Artificial Reflexivity

George Soros gives us the market structure of the problem. Reflexivity begins from the fact that participants in social systems are never clean observers. In markets, investors form views about reality, act on those views, move prices, change balance sheets, alter incentives, and then reinterpret the changed facts as confirmation or refutation. A boom can start with a valid insight, become self-reinforcing through credit and price action, and eventually detach from the conditions that made it true in the first place. The key Soros insight is that perception is not outside the world; perception is one of the forces acting inside the world.

AI is the most powerful reflexive technology yet introduced because it operates directly on perception while connecting perception to action at machine speed. It can generate the narrative, measure the response, refine the message, direct the capital, identify the target, personalize the persuasion, automate the transaction, and produce the next layer of evidence from the changed environment. In markets, this means AI can accelerate the movement from thesis to positioning to price confirmation to narrative validation. In politics, it means a synthetic narrative can move voters, polls, donors, media coverage, and elite expectations until the original artificial premise begins producing real political consequences. In labor, it means an algorithmic assessment of human potential can shape the opportunities that determine whether the assessment later appears accurate.

The financial market analogy is the cleanest because Soros’s framework was built in markets. AI is a real technological revolution with reflexive boom characteristics, and the refusal to hold both ideas at once is where most analysis becomes useless. The underlying capability is real, the capex is real, the energy demand is real, the chip bottleneck is real, the productivity potential is real, and the geopolitical stakes are real; the market structure can still become unstable if a true premise becomes a totalizing narrative that funds itself through rising prices, capital market access, venture financing, data center buildouts, and strategic fear of missing the next platform shift. Railroads were real, the internet was real, housing finance had real foundations, and China’s investment boom had real growth behind it. The strongest bubbles begin with truths powerful enough to seduce serious people.

AI has already created a Soros-style loop. Capital flows into compute because the future appears to belong to model scale. Compute improves capability, capability strengthens the narrative, the narrative raises the cost of underinvestment, underinvestment fear drives more capital into the supply chain, and the rising market capitalization of AI leaders lowers their cost of capital while pressuring every adjacent company to present itself as part of the same future. At some point the market stops merely discounting technological progress and begins financing the reality it expects. The risk emerges when the loop requires increasingly heroic assumptions about monetization, power availability, customer returns on investment, depreciation, model commoditization, and the durability of margins across the stack.

The same reflexive structure applies outside markets. In war, an AI threat model can identify hostile intent, push a state to change posture, cause the rival to respond defensively, and then validate the original model through the rival’s response. In domestic politics, AI-generated narratives can create the public mood they claim to measure, especially when synthetic media, bot distribution, influencer networks, polling, fundraising, and platform algorithms begin feeding on each other. In education and employment, model-driven sorting can determine who receives training, mentorship, capital, trust, and second chances, turning probabilistic classifications into self-fulfilling social facts. AI’s deepest political effect may come from its ability to make predictions behave like interventions.

This is where René Girard reenters the argument through Soros rather than through Thiel alone. Girard’s theory of mimetic desire held that human beings learn what to want by watching what others want, and modern social platforms already turned that mechanism into an industrial system of envy, rivalry, imitation, status anxiety, and scapegoating. AI extends that system by generating infinite objects of desire, infinite rivals, infinite humiliations, infinite enemies, and infinite synthetic proofs that other people are richer, happier, more desired, more oppressed, more dangerous, or more authentic than oneself. Thiel’s warning about “runaway mimesis” becomes more relevant in a world where the mimetic engine can generate content, characters, causes, markets, and enemies without waiting for organic human culture to produce them.

The Three AI Futures

The AI future is likely to move through three competing regimes rather than one clean destiny. The first is the Straussian-security future, where states, intelligence agencies, defense contractors, cloud platforms, and dominant model providers absorb AI into a new architecture of managed perception and managed risk. The case for this future will be strong because the threats are real: AI can scale cyberattacks, automate fraud, manipulate elections, accelerate weapons research, destabilize military command systems, and make truth verification harder during crises. The problem is that every real danger will justify another layer of secrecy, and over time the important decisions may move away from visible democratic institutions into classified systems, proprietary models, export-control bureaucracies, security partnerships, and infrastructure chokepoints.

This future does not require the abolition of elections, courts, newspapers, or markets. It can preserve the surface of liberal democracy while relocating the decisive layer of governance beneath public visibility. Citizens will still debate policy, but the boundaries of possible policy will increasingly be set by risk models, security classifications, platform rules, cloud access, payment rails, identity systems, and automated enforcement. The language will be resilience, safety, and national competitiveness. The structure will be control. A society can remain formally open while its operating system becomes closed.

The second future is the Girardian-chaos future, where AI breaks institutional authority faster than new trust systems can form. Deepfakes become normal, bot networks become indistinguishable from real publics, every scandal receives a counter-scandal, every election becomes a fight over the conditions of reality, and every market move arrives with a swarm of synthetic explanation, manipulation, and conspiracy. People do not need to believe every falsehood for this world to become unstable; they only need to lose confidence that verification is worth the effort. Once the cost of knowing becomes too high, citizens retreat into tribes, financial cults, ideological subcultures, private channels, machine-personalized media, and charismatic interpreters who offer certainty in exchange for loyalty.

This world would be loud, emotional, and politically saturated, but it would not be genuinely democratic because democracy requires some shared capacity to distinguish persuasion from fabrication and disagreement from manufactured unreality. The public square can remain crowded while losing authority. The system can produce constant engagement while destroying trust. AI would then function as a mimetic accelerant, pushing every group to see itself as threatened, every rival as illegitimate, every compromise as betrayal, and every institution as captured unless it confirms the tribe’s priors. Schmitt’s friend-enemy distinction would return through algorithmic personalization rather than formal ideology.

The third future is the Popper-Soros future, where open societies absorb AI by becoming better at criticism, verification, competition, and reversal. This path accepts the power of AI without surrendering institutional judgment to machines or allowing state-security logic to colonize every domain of life. It would require model plurality rather than model priesthood, auditability for high-stakes systems, appeal rights for automated decisions, independent technical capacity inside courts and regulators, competitive pressure against model monopolies, and a public culture that treats forecasts as provisional even when they arrive with the authority of advanced computation. It would also require investors, policymakers, and institutions to understand reflexive loops early enough to prevent false realities from compounding without challenge.

This future is difficult because it asks open societies to move faster without becoming closed, to govern technology without freezing it, and to defend themselves without treating public criticism as sabotage. It requires a type of institutional seriousness that the West has often lacked since the end of the Cold War. The reward is significant because AI can also strengthen openness if deployed correctly. It can improve scientific discovery, fraud detection, medical research, education, defense, legal access, software development, productivity, and public administration. It can help institutions test assumptions, find anomalies, model trade-offs, expose corruption, and widen access to expertise. The decisive question is whether AI compounds correction faster than it compounds delusion.

Markets, Power, and the New Stack

The AI regime will concentrate power where feedback loops already matter most: chips, compute, energy, cloud infrastructure, data, identity, cybersecurity, payments, model distribution, and defense integration. These are control points rather than ordinary sectors because they determine who can train models, who can deploy them, who can verify users, who can secure transactions, who can protect infrastructure, and who can scale applications across the economy. The companies and states that control the stack will shape the cost of intelligence, the distribution of capability, and the boundaries of permissible use.

The exposed institutions are those whose authority depended on information scarcity, credential scarcity, and slow verification. Universities, media companies, consulting firms, bureaucracies, legal processors, middle managers, and large parts of white-collar labor face pressure because AI attacks the production and distribution of knowledge-like outputs. Some of these institutions will become more valuable if they can transform themselves into trusted verification systems. Others will be repriced if they remain content factories, credential machines, or procedural bottlenecks. In a world of infinite synthetic output, scarce judgment becomes more valuable, but only if the institution can prove why its judgment deserves trust.

For investors, the Soros framework is cleaner than the standard bubble-versus-revolution debate. AI should be analyzed as a reflexive boom with real fundamentals. The right questions are where the capex produces durable cash flows, where the customer return on investment is measurable, where the supply chain has pricing power, where energy becomes the binding constraint, where margins compress through model commoditization, where demand is circular, and where the narrative depends on liquidity remaining abundant. A valid technological premise can still produce overvaluation, malinvestment, and violent drawdowns if the market begins extrapolating capability into monetization without enough evidence.

Geopolitically, AI is stack versus stack. The United States has deep advantages in capital markets, research universities, entrepreneurial speed, cloud infrastructure, frontier model companies, venture financing, and semiconductor design. It also has real weaknesses in permitting, energy buildout, political trust, regulatory coherence, and state execution. China has advantages in industrial policy, surveillance integration, manufacturing depth, and centralized deployment, while its weaknesses sit in feedback quality, censorship, capital allocation, demographic pressure, and the political difficulty of admitting failure. The AI race will reward scale and power, but it will punish systems that cannot process bad news.

That is where Popper and Soros converge. Popper shows why criticism is an institutional advantage. Soros shows why distorted perception eventually becomes a balance-sheet, political, or strategic vulnerability. A closed society can suppress negative feedback for a long time, especially when growth is strong and the state controls the narrative, but AI trained and deployed inside a politically filtered reality will eventually inherit the distortions of the regime. Open societies, if they preserve the courage to criticize themselves, retain a deeper adaptive advantage even when they appear less coordinated in the short run.

Against the Religion of Technicity

Thiel’s phrase “religion of technicity” is one of the most useful warnings in the essay because it captures the educated class’s temptation to believe that technical systems can dissolve political judgment. The machine will allocate credit, rank talent, detect threats, personalize education, optimize logistics, secure borders, identify misinformation, summarize law, diagnose disease, trade markets, and advise governments. Many of these uses will be beneficial. The danger begins when technical capability is mistaken for wisdom, when optimization is treated as legitimacy, and when the people who control the models inherit authority without passing through the tests normally required of political power.

AI can improve decisions, but it cannot remove the burden of choosing among values. It can identify patterns, but it cannot decide what a society owes its citizens. It can rank outcomes, but it cannot tell us which trade-offs are legitimate. It can simulate futures, but it cannot absolve human beings from responsibility for selecting one path over another. A society that worships AI will eventually be governed by the people who control AI, and that old political fact will not become less true because the interface looks clean and the output sounds neutral.

Popper would insist that every AI system remain open to criticism because knowledge advances through conjecture and refutation rather than obedience to authority. Soros would insist that every AI system be analyzed through its feedback effects because the model changes the world it measures, especially when institutions act on its outputs. Thiel would warn that liberal society cannot avoid the deeper questions forever, because technology raises the cost of every evasion. The synthesis is severe but useful: artificial intelligence must be powerful enough to strengthen open societies, constrained enough to remain contestable, and plural enough to prevent any single model or institution from monopolizing reality.

The Open Society After AI

The age of AI will test whether liberal civilization can preserve the habits of correction under conditions of machine-speed narrative formation, synthetic media, automated persuasion, concentrated infrastructure, and geopolitical competition. The old liberal order lowered the temperature of politics by moving ultimate questions into private life and directing public energy toward commerce, rights, procedure, and growth. AI raises the temperature again because it touches the machinery through which people decide what is true, who deserves authority, which risks matter, where capital should flow, and which futures appear inevitable.

The answer cannot be a retreat into nostalgia for a pre-AI world, because that world is gone. It cannot be a panic-driven embrace of security-state logic, because that would resolve the crisis of open society by quietly abandoning openness. It cannot be passive faith in market forces, because the market itself is now part of the reflexive system AI will accelerate. The only serious answer is disciplined openness: faster experimentation, faster invalidation, stronger verification, more pluralism, better institutional accountability, and a refusal to let any model become immune from challenge.

George Soros and Karl Popper are more useful than Leo Strauss for this phase because Strauss teaches caution around dangerous truths, while Popper and Soros explain how free societies can remain adaptive when truth itself becomes contested, probabilistic, reflexive, and technologically mediated. Popper tells us that no authority can be trusted with final knowledge. Soros tells us that false beliefs can move history before reality forces the correction. AI makes both insights urgent because it gives falsehood speed, scale, personalization, and institutional access.

The open society will survive only if it becomes more adaptive than the systems trying to manage it and more reality-sensitive than the machines trying to model it. It must see through its own narratives quickly, invalidate its own errors without humiliation becoming paralysis, and preserve enough public trust for criticism to function as a source of strength rather than a prelude to collapse. The central political question of the AI age is whether free societies can still correct themselves once machines begin shaping the perceptions, incentives, and feedback loops through which reality is made.

The trade is that SpaceX may be an extraordinary company coming public at the wrong reflexive moment, with the wrong shareholder base being created at the wrong price. The market is not just being asked to buy rockets, satellites, Starlink terminals, Mars, orbital data centers, defense infrastructure, and AI compute in space. What makes the IPO very interesting is, the buyer, being asked to buy all of those things at once, at a valuation that already assumes the dream remains frictionless.

That is where reflexivity cuts both ways. The reflexive long case is obvious. SpaceX’s stock goes up, the rising stock price validates the story, the validated story lowers the cost of capital, the cheaper capital funds more rockets, satellites, AI infrastructure, and Mars-adjacent ambition, and the execution then validates the original stock price. The price becomes part of the fundamentals and this will happen. That is the Soros framework in its cleanest form. Market perception changes economic reality, and economic reality then appears to confirm market perception.

But the inverse is just as important. If the stock breaks after the initial euphoria, the loop can reverse. A falling price weakens the story. A weaker story reduces sponsorship. Reduced sponsorship increases volatility. Higher volatility raises the cost of capital. A higher cost of capital makes the most ambitious parts of the plan look less financeable. The same valuation that once served as proof of inevitability becomes evidence that the market overreached.

That is the point of shorting SpaceX while staying long Tesla .

My idea is not to short Elon Musk. It is a relative trade between two different stages of the same reflexive ecosystem. Long is long the seasoned Musk belief network. Short SpaceX is short the newly issued version of that belief network at the moment it moves from private scarcity to public-market price discovery.

Tesla is not cheap and Tesla is not a normal stock. Tesla is a belief system with a ticker attached to it - just like SpaceX but the shareholder base is more crystallized.

Tesla’s shareholder base has already been through the fire. It has survived drawdowns, squeezes, index inclusion, dilution, production hell, demand scares, margin compression, regulatory fights, Twitter/X noise, robotaxi promises, AI re-ratings, and multiple cycles of institutional disbelief. The weak hands have been flushed out many times. What remains is a strange but powerful mix of retail believers, passive holders, index buyers, options traders, momentum allocators, and institutions that have learned the hard way that shorting Tesla purely on valuation can be fatal.

A seasoned shareholder base is an asset. It is invisible on the balance sheet, but it changes the way a stock trades. Tesla has a shareholder base trained to interpret volatility as opportunity. SpaceX does not, at least not yet.

SpaceX’s problem is that its IPO turns private-market scarcity into public-market supply. In the private market, owning SpaceX was a status symbol. Access itself created value. The valuation did not have to clear every day. There was no public tape, no liquid borrow, no daily analyst note, no options market, no hedge-fund factor model, no retail cohort trapped above the IPO price. The public market changes the physics.

Once SpaceX trades, the market gets a price every second. That price becomes a referendum. The shareholder base has to form in real time. Every buyer now has a mark. Every seller now has a reference point. Every dip creates the question that never existed in the private market: why is the future of humanity down 9% this week?

That sounds sarcastic, but it is the actual reflexive pressure point. The more cosmic the story, the more dangerous the first public drawdown becomes. If the stock is priced as inevitable, ordinary volatility can feel like theological failure.

The strongest argument for shorting SpaceX is not that the company is bad. It is that the IPO is being priced as if the private-market scarcity premium can survive public-market liquidity. That is a much higher bar.

Reuters reported that SpaceX priced the largest U.S. IPO ever, raising roughly $75 billion at a valuation around $1.77 trillion. The company is being placed immediately among the most valuable listed companies in America. It is also coming with an unusually large retail allocation, which tells you a lot about the intended demand source. That retail demand is powerful on the way up. It can overwhelm valuation discipline.

But retail-heavy shareholder formation is fragile when it happens at the peak of a narrative. Retail buyers are usually strongest when they are being invited to participate in something they already believe is historic. That belief can create the opening bid. It does not automatically create the long-term holder base. A holder base is not created by allocation. It is created by pain, time, and repeated tests of conviction.

The elegant version of the trade is therefore simple: stay long the old Musk duration asset and short the new one.

Moreover… the market loves category creators. It often overpays for their satellites. That is why, I believe, it makes sense to short the concentric-circle around SpaceX

If SpaceX borrow is unavailable or too expensive at first, the next-best expression is to short the public space complex around it: , , , -type names, and other equities that trade with the space narrative but lack SpaceX’s scale, capital access, technical advantage, and institutional importance. These are not all the same but they belong in the same perception field.

The key is to think in circles. At the center is SpaceX: the real company, the real asset, the primary reflexive node. One ring out is Tesla: the seasoned Musk public-market vehicle, indirectly linked, still carrying the broader Musk premium, but with a mature shareholder base and operating cash flow. Another ring out is the space equity complex: Virgin Galactic Holdings Inc, AST SpaceMobile Inc, Intuitive Machines Inc, Rocket Lab Corp, and the thematic basket that benefits when investors say “space is the next AI” without doing the hard work of separating category leader from category beneficiary.

The outer ring is usually where the best shorts live. It has the narrative beta without the dominant economics. It borrows the original company’s dream but lacks the original company’s power.

The historical analogy is Tesla and the EV/SPAC complex in 2020 and 2021.

Tesla was the category-defining reflexive winner. Its stock price lowered its cost of capital, validated the EV transition, attracted talent, allowed the company to raise money on favorable terms, and forced every incumbent and investor to take the category seriously. The rising price helped create the fundamentals that later justified parts of the rising price. That was reflexivity.

But then the market copied the Tesla template onto everything nearby. Rivian. Lucid. Nikola. Lordstown. Canoo. Fisker. QuantumScape. Battery stocks. Charging stocks. Pre-revenue SPACs with investor decks instead of businesses. The market looked at Tesla and concluded that every company with an EV slide deck deserved some piece of the same future, which was a MAJOR MAJOR mistake.

The better trade was not necessarily to short Tesla first. Tesla was the original. Tesla had the reflexive loop. Tesla had the capital markets machine, retail sponsorship, product reality, and an increasingly institutionalized shareholder base. The better short was often the concentric circle around Tesla: the copycats, suppliers, SPACs, and public-market wrappers that were using Tesla’s market cap as proof of their own destiny.

SpaceX may be the Tesla of space. The problem is that the rest of the space complex may be the EV SPAC complex of space. And even SpaceX itself may be coming public at a valuation that already assumes it will not merely become the Tesla of space, but Tesla, Nvidia, Lockheed Martin, Starlink, AWS, OpenAI, and NASA rolled into one security.

That may happen over enough time. But the stock market does not price “eventually” in a vacuum. It prices path, liquidity, holders, dilution, rates, competition, regulation, and the marginal buyer. The marginal buyer is the key.

To conclude, the best shorts are often not the most obvious frauds. They are the assets whose entire shareholder base was formed under a belief that stops being self-reinforcing, which is the unresolved question for SpaceX.

Can it create a public shareholder base as strong as Tesla’s? Or is it coming public at the point when the private-market scarcity premium is being handed to retail and growth funds at the top of the narrative arc? I do not know the answer yet. But I know that this is the right question.

If SpaceX trades well after borrow opens, absorbs supply, holds its IPO price, pulls in long-only ownership, and strengthens the rest of the space complex, the short thesis is wrong.

But if SpaceX fails to hold the first euphoric bid, the consequences can spread quickly. The center weakens, and the outer rings lose oxygen... The retail buyer discovers that dreams have daily prices.

The fund had a strong run following our 2025 returns, largely because we stayed nimble and caught the rally that began at the end of March. We also finished Q1 in the green, which I am proud of given the market environment.

This past week started very well for us. We entered June net short, and that positioning worked early. But on Friday, right before the real $NDX and $SPX drawdown happened, I lightened up on shorts. I also bought equities throughout the day. That decision hurt the fund, and we are now in the red for the week. Happens, won’t be the first - and won’t be the last time.

After the close on Friday, I said the selloff looked like the bottom for June, and most likely the bottom for the year, especially for the Nasdaq 100. That may sound aggressive given the violence of the move, but below is my reasoning…

As you can see above, the top performing baskets experienced real damage on Friday. We got RSI reset on 14D, but moving averages are still trending higher, and RVOL/SPY implies that the leading themes still have strong outperformance ratings versus the indices.

Put/Call ratio, which was very low, spiked to very significant levels for the next 20 trading days - the positioning got caught offsides, but investors were VERY QUICK to position for further downside.

People have been buying too much of these semiconductor and South Korea names, positioning was obviously skewed. When I wrote my long short / article - I got so many notifications from people saying I would get stomped, because the assets would keep trending higher. To give them credit, for ~2 weeks South Korea and did trend higher.

I am quite bullish, and was hoping for a 20-DMA retest, and you can see on the charts above that we had that 20-DMA retest for tech vs broad market, momentum vs market, high beta vs low vol, and semis vs tech.

Up until recently (technology) was the only game in town, and during this pullback we say capital that flowed outside of tech entering real laggard industries like (healthcare) and (financials).

All in all, the indices gave back the gains heading back to May 7, 2026 - so we got a reset following a quite strong month. I shared $SPX and $NDX charts above, but same can be said globally - the charts look the same across Israel, Europe, Asia, Latin America et al.

Finally - I will share some charts from the fund’s Rate-of-change regime explorer for identifying when price momentum is accelerating, fading, or transitioning. These track second-derivative acceleration to flag turning points in momentum.

The US indices don’t look frothy here.

Memory names do look frothy, but they got hit hard - I am expecting a bounce off of 20-DMA for most names, and 50-DMA for the KOSPI index.

The big boys - Magnificent-7 constituents are just fine. Equal weight basket YTD is up less than 3%, and I think these names will lead the indices out the way.

Seasonally, June is one of the worst months during mid-term years. On average, since 1994, the month is down -2.2%, but we have had -8% months - I am not expecting a devastating month.

Finally, the volume based tool Sam Smiley built for us implies that we have had a capitulatory sell-off, despite in short length, on Friday.

Of course this may be the beginning of a downtrend as evidenced by the October/November 2025 technology sell-off, we shall see.

I didn’t talk about fundamentals on this post, let me make a note on that in the end also. is considering raising tens of billions of dollars through a stock offering to fund AI infrastructure, per FT. The discussions intensified after $85B equity raise saw strong investor demand. Meta is preparing to raise AI-related CapEx to as much as $145B this year and even higher in 2027. Right now, people don’t want to acknowledge that the capital raised will be spent on semiconductor chips - give it some time, the analyst earnings estimates for semiconductor names (in my opinion) have not peaked yet. I won’t call this rally a bubble when is up less than Nasdaq 100 YTD, and we get real ROI on AI CapEx spend - as I said in the beginning, I bought some names and I will keep buying.

AD Fund Management LP is down slightly on the month, and we protected the gains from May - you size up when other market participants are distressed.

That month was one of those corrections that looked macro on the surface and mechanical underneath. The headline explanation was simple enough: a weak U.S. labor print, recession fears, a Bank of Japan that sounded less willing to underwrite the global carry trade, and a violent yen rally that forced people out of crowded risk.

The Bank for International Settlements later described the episode as a carry-trade unwind that followed a hawkish Bank of Japan rate hike, a more cautious Fed path, and a disappointing U.S. labor-market release. The yen, still the dominant funding currency for global carry, appreciated sharply. The VIX exploded above 60 in off-hours trading. Japanese equities were crushed, U.S. equities followed, and the losses were concentrated where positioning had been most crowded: tech, semis, momentum, and AI.

As of Tuesday afternoon, May 26, 2026, is trading around $599.89, up roughly 4.1% on the day, with an intraday range from $584.95 to $603.67. , the 3x semiconductor bull ETF, is around $223.66, up roughly 17.4%, with more than 43 million shares traded. , the 3x South Korea bull ETF, is around $1,005, up roughly 28.8%, after trading as low as $906.02 earlier in the session.

The explanation today is obvious enough. Semiconductors are leading again. Reuters reported that the Philadelphia Semiconductor Index hit an all-time high, with Micron surging nearly 17%, Qualcomm rising on a ByteDance chip deal, and Intel and Marvell also advancing as AI demand continues to dominate the tape. South Korea is part of the same story. Korean semis sit closer to the memory cycle, and memory has become one of the hottest expressions of the AI buildout.

Going back to August 2024.. You could say the following - which stands true today: the current tape is euphoric. Japan, yen funding, BOJ normalization, inflation, oil, and crowded semiconductor exposure - all rhyme big time.

Japan is again the key pressure point. Reuters reported today that the BOJ’s new trend inflation gauge rose to 2.8% in April, above the central bank’s 2% target, and that markets are pricing roughly an 80% chance of a June rate hike from 0.75% to 1%. Japan’s 10-year government bond yield has reached 2.8%, the highest level since 1996.

Japan story never really is isolated. The yen hits the global funding structure. For years, investors borrowed cheaply in yen and bought higher-return assets elsewhere. When the yen is weak, volatility is low, and global liquidity is forgiving, this trade looks painless. The danger is that stability itself encourages more leverage. The longer the carry trade works, the larger it gets. Then one day the funding currency moves, and what looked like yield pickup becomes forced liquidation. That is what August 2024 taught us.

Today, however, the market is still in the winning phase of the reflexive loop. is making new highs. is exploding higher. is acting like a call option on the Korean memory cycle. AI demand is still overpowering macro fear. Iran risk is being treated as a tradable headline rather than a regime break. Investors are still willing to pay for scarcity, and semis remain the cleanest scarcity asset in public markets.

But the same logic cuts both ways. When SMH is up 4% in a session, SOXL is up 17%, and KORU is up nearly 29%, the market is telling you where the leverage lives. These are the assets that scream when the liquidity impulse is positive. They are also the assets that will expose fragility first if the impulse turns. The higher the tape pushes without volatility, the more investors internalize the idea that AI demand can overwhelm every macro problem. That belief becomes its own fuel.

This is the Soros point. The fundamentals and the price action are now interacting. Higher semiconductor prices validate the AI capex story. The validated AI story pulls more capital into the trade. More capital pushes the stocks higher. Higher stock prices make the companies look stronger, make financing easier, attract more attention, and compress career risk for anyone buying the leaders. The loop keeps going until something interrupts it.

The interrupter does not need to be the same as August 2024. It could be a BOJ hike that strengthens the yen and pressures carry trades. It could be a Warsh Fed that cannot deliver the dovish pivot the market wants. It could be oil staying high enough to keep inflation sticky. It could be a semiconductor earnings print that is merely good when the market is priced for exceptional. It could be nothing more complicated than crowded positioning meeting a negative catalyst in thin liquidity.

For now, the tape says the AI boom is still in control. But I’d argue that August 2024 was the warning that semis had become the market’s liquidity instrument. And I think market’s liquidity will get tested over the coming weeks.