Amazon and Alphabet Are Already Among Europe’s Biggest Corporate Borrowers. With AI Infrastructure Heading Toward $5–7 Trillion by 2030, America’s Tech Giants Are Beginning to Tap European Pension Money to Finance a Race Europe Is Already Losing.

Subscribe now

For three years, we have been asking how much electricity artificial intelligence will consume.

Maybe we should start asking another question:

How much savings will it consume?

Because something extraordinary is beginning to happen inside one of the least glamorous corners of global finance.

Amazon. Alphabet. Microsoft. Meta. Oracle.

The giants building America’s AI infrastructure are no longer financing the boom primarily from the ocean of cash their businesses once generated.

The numbers have become too large. So they are borrowing. A lot.

First in dollars. Then through private credit. Then through increasingly elaborate infrastructure-financing structures. And now something particularly interesting is happening.

These giants are coming to Europe. Not to build Europe’s AI industry. To borrow Europe’s money to finance America’s.

According to the European Central Bank, US hyperscalers now have roughly €40 billion of euro-denominated bonds outstanding.

That sounds relatively modest. But look at the flow.

US Big Tech already represents just under 10% of gross new euro-denominated bond issuance by non-financial corporations.

Amazon and Alphabet have become the two largest non-financial corporate issuers in the euro-area bond market this year.

Amazon’s deal was the largest transaction of its kind ever recorded there. And European pension funds and insurers are buying. Aggressively.

Hyperscaler bonds accounted for roughly 15% of the increase in domestic holdings of euro-denominated corporate debt in the year to March 2026.

Think about what that means.

European savers are increasingly financing American AI infrastructure. While Europe struggles to build enough of its own.

That is one of the great ironies of the AI race. And we may only be seeing the beginning.

AI Has Found Its Next Raw Material

The AI boom began with a GPU shortage.

Then came the electricity shortage. Then transformers. Cooling equipment. Land. Grid connections. Permits. Gas turbines. Copper.

Now the shortage is moving into something even more fundamental: Capital.

It takes extraordinary amounts of money to convert AI from software into physical infrastructure.

• Data centers cost billions.

• GPU clusters cost billions.

• Power generation costs billions.

• Transmission costs billions.

• Networking costs billions.

• Cooling costs billions.

And the industry is no longer discussing a few isolated projects.

Reuters notes that estimates for AI-related infrastructure investment through 2030 increasingly fall between $5 trillion and $7 trillion.

AI capital expenditure could exceed $1 trillion in 2026 alone.

Trillions.

At that scale, even the world’s richest corporations begin looking capital constrained.

The Quantum Time Bomb: Why Satoshi’s $88 Billion Bitcoin Fortune is the Ultimate Target for Error-Corrected Supercomputers.

Shor’s Algorithm, exposed P2PK outputs, and the race to secure Bitcoin’s earliest coins.

Big Tech Was Supposed to Be Different

This is what makes the transition fascinating.

The largest technology companies historically had one enormous advantage over ordinary industrial companies. They generated absurd amounts of cash.

A software company did not need to continuously rebuild a steel mill.

Google could serve another search query at negligible marginal cost. Microsoft could distribute another software license without building another factory. Facebook could add another user without constructing another physical plant.

That was the magic.

High margins. Low capital intensity. Enormous free cash flow.

Then AI arrived.

Suddenly technology began looking suspiciously like heavy industry.

• More customers require more inference.

• More inference requires more GPUs.

• More GPUs require more data centers.

• More data centers require more electricity.

• More electricity requires power infrastructure.

The marginal cost structure changed.

Big Tech started becoming Big Industry. And Big Industry loves debt.

The Cash Machines Are Becoming Construction Companies

Look at what the hyperscalers are building.

Read more

Subscribe now

History is rarely kind to presidents who misunderstand the fundamental technological shifts of their eras.

The legacy of a chief executive is almost never defined by the daily political skirmishes, the passing legislative battles, or the ephemeral controversies that dominate the news cycle.

Instead, presidential legacies are carved into the bedrock of time by the structural economic choices they make when the world is standing on the precipice of a new epoch.

In the mid-19th century, it was the transcontinental railroad. In the mid-20th century, it was the interstate highway system and the space race.

Today, in the defining years of the 21st century, it is the mass proliferation of artificial intelligence, the energy grids that power it, and the physical fortresses where it is forged: data centers.

Donald Trump had to choose a side on the explosive, energy-intensive, and highly contested expansion of data centers across the American landscape. Amidst loud calls for degrowth, regulatory pauses on artificial intelligence, and environmental pushback against the massive energy requirements of modern compute, Donald Trump made a definitive choice.

Donald Trump took the side of building data centers, removing the friction for their construction, and unleashing the energy sector required to power them. Decades from now, long after the partisan noise of the 2020s has faded into the archives of political history, this singular decision will remain his most consequential legacy.

To understand why this choice is so pivotal, one must look beyond the screen of a smartphone and peer into the macroeconomic realities of the United States.

The American economy is currently carrying a burden of sovereign debt that is unparalleled in its history. Traditional levers of economic management are straining under the weight of unfunded liabilities, aging demographics, and a manufacturing base that spent decades migrating overseas.

In this precarious environment, there is only one mathematically viable path forward: explosive, unprecedented economic growth. And as we look at the technological landscape, artificial intelligence — which is essentially a direct product of data centers — is the only phenomenon capable of generating the sheer volume of economic value required to handle this massive debt and reindustrialize the United States.

People fundamentally misunderstand what compute actually is. It is not merely a service for storing photographs or hosting websites; compute is essentially a 21st-century money-printing machine. Once a hyper-scale data center is built and powered, it generates trillions upon trillions of intelligent tokens. Those tokens solve complex real-world problems in engineering, medicine, logistics, and software development, directly creating immense economic value. And in the modern world, economic value is money. Opposing data centers is, therefore, opposing growth, opposing wealth generation, and opposing the very survival of the American economic empire.

By aggressively backing the proliferation of data centers and the energy infrastructure required to sustain them, Trump has positioned the United States to dominate the next century. This comprehensive analysis will explore exactly why data centers are the most significant ingredients of the U.S. economy in the coming decades, how they function as engines of local and national prosperity, and why Trump’s pro-compute, pro-energy stance is undeniably on the right side of history.

Bitcoin Won the Money War. Now Wall Street Is Winning the Custody War.

BlackRock Has Already Processed More Than $5B of Direct Bitcoin-to-ETF Conversions. The Most Important Part Isn’t the Money — It’s That Bitcoin Holders Are Voluntarily Giving Wall Street Their Keys.

Chapter I: The Mathematics of Empire — Debt, Stagnation, and the Growth Imperative

To appreciate the absolute necessity of data centers and artificial intelligence, one must first confront the grim arithmetic of the United States balance sheet. The national debt has reached levels that historically precede the decline of global superpowers. When a nation’s debt-to-GDP ratio exceeds the total output of its economy, the structural integrity of its financial system begins to warp. Interest payments consume increasingly massive portions of the federal budget, crowding out infrastructure, defense, and social investments.

In classical economic theory, there are only three ways to escape a sovereign debt crisis of this magnitude. The first is default, which is unthinkable for the issuer of the world’s reserve currency; it would trigger a global depression and the immediate end of American financial hegemony. The second is hyperinflation, effectively inflating the debt away by destroying the purchasing power of the currency, which annihilates the middle class and destabilizes the republic.

The third option — and the only viable, prosperous option — is growth.

Debt inherently requires growth. Without growth, debt fails. The entire architecture of modern finance is predicated on the assumption that tomorrow’s economy will be larger, more productive, and more innovative than today’s. If an economy grows faster than the interest on its debt, the debt burden shrinks in relative terms. You can outgrow the problem.

But how does a mature, highly developed economy like the United States achieve the kind of explosive, 4-to-6 percent sustained annual GDP growth required to outrun tens of trillions of dollars in obligations?

Read more

Subscribe now

Everyone is looking for the company that will kill Nvidia.

AMD?

Broadcom?

Google?

Some mysterious Chinese chip startup?

Maybe we are asking the wrong question.

Because Nvidia’s most dangerous future competitor may already be one of its biggest customers.

Microsoft.

And what Microsoft is preparing could reveal the next phase of the AI infrastructure war.

Microsoft is expected to unveil its next-generation Maia 300 AI accelerator as early as September.

That alone would not be particularly surprising.

Big Tech has been designing custom silicon for years.

The number that matters is much bigger.

Microsoft has reportedly been discussing manufacturing capacity with TSMC for more than 300,000 Maia 300 chips for delivery in 2027.

Longer term, Microsoft reportedly wants capacity for more than one million chips.

That is no longer an experiment. That is industrial scale.

And if Microsoft succeeds, the implications extend far beyond one new semiconductor.

Microsoft doesn’t have to destroy Nvidia. It doesn’t even have to build a better chip. It only needs to make Nvidia slightly less indispensable.

That could be enough to change the economics of the entire AI boom.

Nvidia Has the Best Problem in Corporate America

Let’s establish something before going further.

Nvidia is not in trouble. Quite the opposite.

Its latest quarter was almost absurd.

Revenue reached $96.22 billion. Data-center revenue alone hit approximately $89 billion.

Nvidia is forecasting another enormous quarter and expects revenue growth of roughly 70% in the following fiscal year.

This is not what a collapsing monopoly looks like.

AI demand remains ferocious.

Hyperscalers are still spending extraordinary amounts of money building data centers. Models are getting larger. Inference workloads are exploding. Agents may multiply the number of tokens generated across the economy. Every additional AI workload needs compute.

Nvidia sits directly in front of that demand.

Its real problem isn’t selling GPUs. Its problem is producing enough of them. But extraordinary profitability creates its own enemy. It gives customers a reason to leave.

When one supplier captures enormous margins from something your company needs in gigantic quantities, eventually someone inside your organization asks a dangerous question:

Why are we paying them instead of building this ourselves?

Microsoft is beginning to answer it.

Meet Maia

Microsoft’s custom silicon ambitions are not new. The company introduced its first Maia accelerator in 2023.

Then came Maia 200.

And Microsoft says Maia 200 already delivers more than 30% better performance per dollar than the latest-generation hardware in parts of its existing fleet.

Even more interestingly, Microsoft says its own MAI models running on Maia 200 are producing 40% better performance per watt through hardware-software co-design.

Read those sentences again.

Not because Maia suddenly beats Nvidia everywhere. It doesn’t.

The important words are: performance per dollar.

And: performance per watt.

Those are the metrics that matter when you operate data centers at Microsoft scale.

Microsoft doesn’t buy 50 chips. It deploys entire gigawatts of computing infrastructure.

A 5% efficiency improvement can matter. A 10% improvement can matter enormously. A 30% improvement changes capital allocation.

That is why Maia 300 matters.

Microsoft doesn’t need the world’s fastest accelerator.

It needs an accelerator that is good enough for specific workloads and dramatically cheaper to operate at massive scale.

There is a huge difference.

The Market Keeps Looking for an Nvidia Killer

This may be the biggest analytical mistake investors are making.

They imagine semiconductor competition like a boxing match.

Nvidia versus AMD.

One chip versus another.

Benchmark versus benchmark.

Winner versus loser.

But hyperscalers don’t think that way.

Microsoft doesn’t need Maia to replace Nvidia completely. Google doesn’t need TPU to replace Nvidia completely. Amazon doesn’t need Trainium to replace Nvidia completely.

They can continue buying Nvidia’s most advanced systems while quietly moving enormous quantities of other workloads onto their own silicon.

That creates a far more dangerous dynamic.

Nvidia doesn’t lose the customer. It loses share of wallet.

Imagine Microsoft needs $50 billion of AI compute.

Under the old model, perhaps an enormous percentage goes toward Nvidia infrastructure.

Under the new model, Microsoft might still spend billions buying Nvidia’s most advanced systems for frontier training.

• But inference?

• Synthetic data generation?

• Internal Copilot workloads?

• Microsoft’s proprietary models?

• Enterprise AI?

• Lower-tier workloads?

Those can increasingly migrate to Maia.

Suddenly Nvidia is still growing. But Microsoft is capturing part of the economics Nvidia would otherwise have captured.

That distinction could become extremely important.

Pakistan Found a New Way to Build a Bitcoin Reserve: Mine It Instead of Buying It.

How an IMF-Constrained Country Could Turn Stranded Electricity Into Sovereign Bitcoin Without Spending Its Scarce Dollars.

The Real Battlefield Is Inference

Training gets the headlines. Inference may get the money.

Read more

Subscribe now

Everyone watching the US-China AI race is staring at the same things.

• Nvidia GPUs.

• DeepSeek.

• Huawei.

• Semiconductors.

• Data centers.

• Electricity.

• Kimi.

• OpenAI.

• Robotics.

• Export controls.

Those things obviously matter. But there may be another competitive advantage hiding in plain sight.

One that cannot be manufactured by TSMC. One that cannot be purchased from Nvidia. One that does not appear on an AI benchmark.

The willingness of millions of workers to actually use the technology.

And on that front, the United States may have a problem.

A new Pew Research Center survey found that 73% of Americans under 30 believe AI will lead to fewer jobs in the United States over the next twenty years.

Two years ago, the number was 61%.

Meanwhile, 55% of Americans aged 18 to 29 now say they are more concerned than excited about AI. Only 11% say they are more excited than concerned.

Those are extraordinary numbers.

America is building the world’s most powerful AI systems. And the generation expected to spend its career working alongside them increasingly thinks those systems are coming for its livelihood.

China looks very different.

A Chinese survey released by PCG Lab in August studied 1,000 people born between 1997 and 2009. Among respondents already working, 94.3% said they deeply use AI in their daily jobs.

Only 5.7% said they were “very worried” about technological substitution.

More than half — 54.9% — said AI represented both opportunity and risk.

That difference deserves far more attention. Because the next phase of the AI race may not be won by whoever invents the smartest model.

It may be won by whoever persuades millions of ordinary workers to use it.

And fear is terrible deployment infrastructure.

First, an Important Caveat

The 73% and 5.7% figures make a fantastic headline.

They are not directly comparable statistics.

Pew asked young Americans whether AI would result in fewer US jobs over the next twenty years.

The Chinese survey asked respondents how worried they personally felt about technological substitution, with 5.7% choosing the strongest “very worried” response.

Someone can believe AI will eliminate jobs without believing their own job is in immediate danger.

So this is not evidence that Americans are thirteen times more afraid of AI than Chinese workers.

That would be bad analysis.

But taken together with broader sentiment data, workplace adoption and labor policy, the surveys reveal something more interesting.

The emotional environments surrounding AI adoption appear radically different.

And that matters. A lot.

America Has an AI Trust Problem

For years, corporate leaders talked about AI adoption as though employees were simply waiting for better software.

Give them Copilot. Give them ChatGPT. Give them Claude. Give them agents.

Train them. Measure usage. Productivity appears.

Reality is proving messier.

McKinsey surveyed 750 AI-using workers globally earlier this year and found roughly one in five workers reported anxiety about AI-related workplace changes.

Among middle managers, the figure rose to one in four. Workers with low trust that their organization would support them through the AI transition were 1.5 times more likely to feel anxious.

That should surprise nobody.

Imagine your CEO tells you:

“We’re investing heavily in AI because we want everyone to become dramatically more productive.”

Then six months later the company announces layoffs.

What lesson do employees learn?

Not: AI will empower me.

They learn: My productivity improvements may eventually be used to eliminate my position.

Now management asks everyone to enthusiastically teach the AI how they work.

Good luck.

The $130 Million Cold Storage Illusion: What Hardware Wallet Breaches Teach Us About True Sovereign Bitcoin Custody.

Buying a hardware wallet doesn’t make you invulnerable — it just changes how you get attacked. Here is the brutal truth about blind signing, supply chain leaks, and why single-sig setups fail.

Every AI Productivity Pitch Contains an Uncomfortable Question

Suppose you are an analyst.

You spent ten years learning financial modeling. Then an AI agent learns to build a model in thirty seconds.

Management celebrates.

You improve the agent. You correct its mistakes. You document your workflow. You teach it the exceptions.

Eventually it performs 70% of what you previously did.

Your company now tells investors that AI will produce major productivity gains.

At some point, a thought enters your mind.

What exactly happens to me when the productivity gain is realized?

Read more

Banks say yield-bearing stablecoins could drain $1.3 trillion from deposits and erase $850 billion of lending capacity. The real war isn’t Crypto vs. Visa anymore. It’s Stablecoins vs. the Business Model of Banking itself.

Subscribe now

The banks finally said the quiet part out loud.

For years, stablecoins were presented as a payments story.

Faster settlement. Cheaper transfers. Digital dollars. 24/7 markets.

An easier way to move money around the Internet.

That was never the real threat. Because Visa does not fund your mortgage. Mastercard does not finance your local hardware store. PayPal does not use your checking-account balance to make agricultural loans. Banks do.

And now American community banks are warning that stablecoins could attack the raw material that makes that entire machine possible: your deposits.

The Independent Community Bankers of America estimates that allowing stablecoin holders to earn interest, yield, or economically similar rewards could ultimately pull as much as: $1.3 trillion out of community-bank deposits.

And according to the group’s modeling, that could translate into approximately: $850 billion less lending.

Those are enormous numbers. They’re also controversial.

The White House Council of Economic Advisers says the banking industry’s estimates dramatically exaggerate the likely damage and argues that prohibiting stablecoin yield could impose costs on consumers while doing very little to protect bank lending.

So this is not a settled economic fact. It is something much more interesting. It is the opening battle over who gets to hold your dollars in the digital economy.

And whoever wins may reshape not only crypto but banking itself.

Stablecoins Aren’t Competing With Visa

They’re competing with your savings account.

That is the conceptual shift.

Suppose you have $20,000 sitting in a bank.

Maybe the bank pays you:

0.5%.

1%.

2%.

Maybe almost nothing.

Behind the scenes, that deposit is economically valuable. It gives the bank relatively cheap funding.

The bank can combine deposits with capital and other funding sources and use its balance sheet to make loans.

• Mortgages.

• Small-business credit.

• Auto loans.

• Agricultural financing.

• Commercial real estate.

• Credit lines.

That is one of the core functions of banking.

Now imagine another product appears.

It looks like dollars. You can send it anywhere almost instantly. It settles 24/7. It can live inside crypto wallets. It can interact with financial applications. And — here is the important part — it also pays you a competitive return.

Suddenly the question becomes obvious.

Why leave your idle money earning almost nothing at a bank?

That is what scares bankers.

Not blockchain. Not cryptography.

Competition for deposits.

Yield Changes Everything

A stablecoin with no yield is mostly a transactional instrument.

You hold it because you intend to:

• trade,

• send money,

• settle something,

• use DeFi,

• move between exchanges,

• or store digital dollars temporarily.

But attach yield?

The economic category changes.

Now you’re not merely competing with payment apps.

You’re competing with:

• savings accounts,

• money-market funds,

• brokerage cash balances,

• and potentially bank deposits themselves.

That is a radically larger market.

This is why the debate over one seemingly boring word — yield — is becoming one of the most consequential fights in American crypto policy.

The GENIUS Act Already Saw the Problem

The United States already placed limits on this.

The GENIUS Act, enacted in July 2025, says that a permitted payment-stablecoin issuer cannot pay holders interest or yield solely because they hold, use, or retain the stablecoin.

That sounds clear.

Circle issues a payment stablecoin? Circle cannot simply turn it into an interest-bearing savings account under the statute.

But the financial industry sees a loophole.

• What if the issuer does not pay the reward?

• What if an exchange does?

• An affiliate?

• A trading platform?

• A wallet?

• A third-party intermediary?

Maybe the stablecoin technically pays zero. But the platform says:

Hold 10,000 USDC here and receive 4%.

Economically, the consumer may see little difference. Banks absolutely do.

That is why the ICBA is now lobbying to make the prohibition broader as Congress debates the CLARITY Act.

The group wants the restrictions to cover stablecoin issuers and exchanges, affiliates, and other digital-asset intermediaries.

For community bankers, there is apparently no acceptable compromise.

ICBA President and CEO Rebeca Romero Rainey said last week there was effectively no “middle ground” if legislation still allows stablecoins to undermine local deposit funding.

That is remarkably strong language. And it tells you exactly how seriously banks are taking this.

The Fed Can Fight Inflation — or Save the Treasury. Bitcoin Wins If It Can’t Do Both.

America’s debt machine is colliding with its inflation fight. If the Fed must choose between the dollar and the Treasury market, Bitcoin could win.

Why $1.3 Trillion?

This number needs context.

It is not a prediction that $1.3 trillion will disappear from community banks next Tuesday.

Read more

The political theater is nothing short of spectacular.

Following the dramatic geopolitical shifts of early 2026, President Donald Trump has announced what he calls “the biggest oil deal in world history”.

By establishing majority control over 17 fields in Venezuela’s Orinoco Belt and Lake Maracaibo, the United States effectively lays claim to a staggering 65 billion barrels of proven oil reserves.

The narrative being spun in Washington is one of absolute economic salvation.

Politicians and pundits are framing this unprecedented acquisition as a silver bullet—a multitrillion-dollar windfall that will simultaneously crush inflation, refill the U.S. Treasury, pay down the national debt, and bring cheap gasoline to the American consumer for the next century.

It is a beautiful, patriotic story. It is also mathematically impossible.

When you strip away the political grandstanding and subject this geopolitical coup to the cold, unforgiving laws of petroleum engineering, fiscal accounting, and macroeconomic reality, a very different picture emerges. The Venezuelan oil deal is not the financial savior of the American empire. It is a rounding error in the face of a sovereign debt crisis so vast that human intuition struggles to comprehend it.

To understand why this massive seizure of oil will not prevent the impending devaluation of the U.S. dollar, we must put aside the rhetoric and do the math.

Subscribe now

Chapter I: The Mathematics of the Black Gold Mirage

At first glance, the top-line numbers are intoxicating. Sixty-five billion barrels of oil is an ocean of energy. It is enough to theoretically supply the entire global economy for almost two years.

Let us run the most optimistic financial scenario imaginable. If the United States were able to sell every single drop of that oil at a sustained premium price of $100 per barrel, the gross revenue generated would be $6.5 trillion.

In the minds of the public, $6.5 trillion sounds like enough money to fix the world. But gross revenue is an illusion; net profit is the only metric that matters. Oil does not magically leap from the subterranean depths of the Orinoco Belt into the gas tanks of American cars. It must be surveyed, drilled, extracted, diluted, transported, and heavily refined.

This brings us to the brutal reality of Venezuelan crude. The vast majority of the reserves located in the Orinoco Belt do not resemble the light, sweet crude pumped out of Texas or Saudi Arabia. Orinoco crude is “extra-heavy”—a highly viscous, dense sludge that is rich in sulfur and heavy metals. It is closer to liquid asphalt than to conventional petroleum.

To turn this dense sludge into usable products like diesel, gasoline, and jet fuel, it requires specialized, multibillion-dollar refining units. Standard refineries simply cannot process it. The crude must be sent through massive Hydrotreaters (which use extreme heat and hydrogen gas to strip away the sulfur) and Delayed Cokers (which thermally crack the carbon-rich sludge into lighter fuels).

Because of the decrepit state of Venezuela’s oil infrastructure — hollowed out by decades of mismanagement and lack of capital investment — the capital expenditure required just to get this oil out of the ground and to a refinery is astronomical. While light crude might cost a few dollars a barrel to extract, the fully integrated cost of extracting, diluting, transporting, and upgrading Venezuelan heavy crude easily reaches $50 to $60 per barrel.

Therefore, assuming a 50% cost margin is not just realistic; it is highly conservative.

When we apply this 50% cost burden to our $100-per-barrel scenario, our $6.5 trillion gross revenue is immediately chopped in half.

The net financial gain to the United States is, at best, $3.25 trillion.

Is $3.25 trillion a massive sum of money? In the context of an individual, a corporation, or a small nation, yes. But in the context of the 2026 United States federal government, it is a fleeting drop in an infinite ocean of fiscal liabilities.

Chapter II: A Drop in the Ocean of American Debt

To understand why $3.25 trillion cannot save the United States, we must confront the horrifying scale of the American sovereign debt crisis.

As of late August 2026, the gross national debt of the United States officially breached $40.08 trillion. This is a number so large that it defies human visualization. It represents an accumulation of decades of structural deficits, unpaid wars, massive entitlement expansions, and endless economic bailouts. Since 2019 alone, the national debt has nearly doubled, expanding at a rate that mathematically outpaces the growth of the underlying economic output.

But the total principal is only half the problem. The true crisis lies in the annual cost of carrying that debt and the government’s inability to stop bleeding cash.

Let us look at the federal budget deficit. A deficit occurs when the government spends more than it collects in taxes, forcing it to borrow the difference by issuing Treasury bonds. In the first nine months of fiscal year 2026 alone, the federal deficit hit $1.4 trillion. When annualized, the United States is currently running a budget deficit of roughly $1.8 trillion to $2 trillion per year, during a period of relative peace and economic function.

Now, add the interest expense. For decades, the U.S. government was able to mask its borrowing binge because interest rates were artificially suppressed to near zero. That era is over. As the Federal Reserve fought inflation by hiking rates, the cost of rolling over the national debt skyrocketed. The U.S. government is now paying massive yields to bondholders just to maintain the existing debt load.

Interest payments on the national debt have now become the second-largest government expenditure, trailing only Social Security. The annual interest expense alone now hovers between $1.2 trillion and $1.5 trillion.

When you combine the structural operational deficit (~$1.8 trillion) with the annual interest expense (~$1.4 trillion), the U.S. government is facing an annual cash hole of roughly $3.2 trillion.

The math is absolute and terrifying:

The entire net profit of seizing 65 billion barrels of Venezuelan oil ($3.25 trillion) equals exactly one year of the U.S. government’s combined deficit and interest expense.

Read that again. If the United States could magically extract, refine, and sell 65 billion barrels of oil tomorrow—a physical impossibility—the resulting profit would only buy the Treasury exactly 12 months of solvency. By the end of next year, the government would be completely out of oil money, and the debt would resume its exponential upward trajectory.

Chapter III: The Tyranny of Time and the Supply Chain

The reality of the oil market makes the fiscal situation even bleaker, because the petroleum industry is governed by the rigid laws of physics, thermodynamics, and logistics.

Read more

Subscribe now

For most of the AI boom, everyone obsessed over the same shortage.

GPUs.

Nvidia couldn’t make them fast enough. Cloud providers fought for allocations. Startups raised billions largely to secure compute. Jensen Huang became one of the most important CEOs on Earth because possession of Nvidia accelerators increasingly looked like possession of the future.

But something strange is happening.

The bottleneck is moving.

Companies can order more GPUs. They can raise more money. They can design bigger clusters.

What they increasingly cannot do is find somewhere to turn those GPUs on.

Because an AI data center needs more than chips.

It needs land. Electricity. Transmission. Transformers. Cooling. Water. Permits. Construction crews. Grid connections. Political approval.

And increasingly, one more thing that Silicon Valley is not accustomed to negotiating for: permission from the people who live nearby.

The next stage of the AI race may therefore be decided by something far less glamorous than model intelligence.

Not who has the best algorithm. Not who buys the most Blackwell GPUs.

But who can actually build. And that changes almost everything.

The GPU Shortage Was the Easy Part

That sounds absurd considering how difficult the GPU shortage became.

For years, Nvidia accelerators were effectively rationed by demand. Companies signed enormous contracts. Cloud providers reserved capacity years in advance. AI startups measured their credibility partly by how many GPUs they controlled. Compute became status.

But semiconductor scarcity has one characteristic investors understand well:

• Scarcity creates an economic response.

• High prices create enormous incentives to produce more.

• Nvidia expands.

• TSMC expands.

• Memory manufacturers expand.

• Advanced packaging expands.

• AMD competes.

• Google builds TPUs.

• Amazon builds Trainium.

• Microsoft develops Maia.

• China pushes Huawei Ascend.

• Capital attacks shortages.

• Slowly, painfully, supply grows.

But you cannot apply exactly the same logic to electricity.

You cannot manufacture a gigawatt of grid capacity like another batch of GPUs. You cannot overnight create transmission infrastructure where none exists. You cannot accelerate a ten-year permitting process simply because Microsoft has another $20 billion. You cannot place a 1-gigawatt data center wherever a spreadsheet says the return is highest.

Physics has entered the AI business. And physics comes with zoning boards.

The Numbers Are Becoming Absurd

AI infrastructure is leaving the scale of ordinary technology projects.

We are talking about gigawatts now. Not megawatts. Gigawatts.

Nvidia recently agreed to provide guarantees supporting an enormous OpenAI-linked data-center project in Ohio expected eventually to reach 8 gigawatts of capacity. Nvidia is also investing directly in the developer, illustrating how far the company is willing to go to ensure that physical infrastructure exists for its chips.

Eight gigawatts is not a server room. It is an industrial energy system. And this is happening while hyperscaler capital expenditure continues exploding.

JLL estimates the four largest hyperscale cloud providers could spend roughly $725 billion in 2026, much of it connected to AI computing and data-center infrastructure.

At this scale, buying GPUs is only the beginning.

The question becomes:

Where do you put them?

The AI Industry Is Discovering Geography Again

Software trained us to believe location barely mattered.

A startup could be founded in a garage. An app could reach billions of people. Code could move instantly. Cloud computing made physical infrastructure feel almost invisible.

AI is reversing that illusion.

The physical location of compute suddenly matters enormously.

• Can the site access 500 megawatts?

• Can it eventually reach one gigawatt?

• How quickly?

• What does electricity cost?

• Is there fiber?

• Can new transmission lines be built?

• Is water available?

• What cooling technology is viable?

• Will local residents oppose the project?

• Does the region have enough construction workers?

• Can the transformers arrive before 2030?

This is why Europe’s AI data-center map is starting to change.

New hyperscale projects are increasingly moving away from traditional hubs such as London, Frankfurt and Amsterdam toward areas where power, land and grid access are easier to secure. JLL says hyperscale centers planned for 2026–2028 will on average be much farther from major urban centers than projects built earlier in the decade.

The new data-center rule is brutally simple:

Go where the electricity is.

Powered Land Is Becoming More Valuable Than Land

This creates a new kind of real estate.

Ordinary land is not enough. The valuable asset is powered land.

A field beside a highway may be cheap. A field beside a substation with a credible path to hundreds of megawatts can be worth dramatically more.

In Europe’s core data-center markets, JLL estimates powered land can cost millions of euros per megawatt of IT load, while secondary locations can be dramatically cheaper.

That tells us something important.

AI infrastructure is reorganizing around energy access. Not talent density. Not prestigious office addresses. Not proximity to venture capital.

Electricity.

The cloud is becoming geographically anchored to the grid.

Then Comes the Grid Queue

Suppose you find land.

You buy it. You obtain financing. You design the facility. You order the GPUs. You are still not done.

Now you need a grid connection.

And this may be where the AI infrastructure boom becomes genuinely difficult.

Electricity grids were not built for dozens of customers suddenly asking for power equivalent to small cities.

• Utilities have planning cycles.

• Transmission operators have reliability obligations.

• Power plants take years to develop.

• Interconnection studies take time.

• Equipment has to be procured.

• Local grids have to be reinforced.

• And data centers want power now.

That creates a collision between Silicon Valley time and utility time.

An AI company thinks in quarters. A transmission project can think in decades.

Those clocks do not match.

The Fed Can Fight Inflation — or Save the Treasury. Bitcoin Wins If It Can’t Do Both.

America’s debt machine is colliding with its inflation fight. If the Fed must choose between the dollar and the Treasury market, Bitcoin could win.

America’s Largest Grid Is Already Feeling the Pressure

Look at PJM.

PJM coordinates the electricity market serving roughly 67 million people across a huge part of the eastern United States.

It also contains some of America’s largest concentrations of data centers.

The tension has become so severe that PJM has proposed mechanisms that could force large data centers onto backup power during grid emergencies.

Why?

Because the system must prioritize reliability.

PJM recently fell short of its reliability requirement by 6.8 gigawatts even as data-center demand accelerates.

Pause on that number.

The AI industry is imagining individual projects measured in multiple gigawatts while one of America’s most important grids is already wrestling with gigawatt-scale reliability gaps.

This is no longer a theoretical bottleneck. It is an operational one.

AI Has Found Its Most Powerful Competitor: Your Electricity Bill

There is another problem.

Data centers do not exist in a political vacuum.

People pay electricity bills.

If data-center demand forces utilities to build additional generation, substations and transmission infrastructure, someone must pay.

Residents are beginning to ask whether that someone will be them.

This changes the political economics dramatically.

For years, communities welcomed technology investment because the trade looked obvious.

A company builds. Jobs arrive. Tax revenues rise. Politicians cut ribbons.

But enormous AI data centers create a more complicated equation.

They consume vast amounts of electricity. They can use significant quantities of water. They require transmission infrastructure. They may increase demand for gas generation. They can occupy enormous tracts of land.

And compared with a traditional factory employing thousands, a fully operational data center may employ relatively few permanent workers.

Residents are beginning to notice that trade-off.

The Most Important AI Poll May Not Be About ChatGPT

A Reuters/Ipsos poll this summer produced an extraordinary result.

Only about a third of Americans approved of the current pace of data-center construction.

And just 14% said they would support a data center being built in their own community.

Fourteen percent.

That number should terrify the AI infrastructure industry.

Because the AI boom has mostly been modeled as a capital-allocation problem.

Raise money. Buy chips. Build infrastructure. Generate intelligence.

But infrastructure requires political legitimacy.

If local opposition keeps rising, capital alone cannot solve the problem.

Suddenly the scarce resource is no longer compute. It is consent.

Pennsylvania Just Sent the Industry a Warning

This shift became especially visible in Pennsylvania.

The state looks almost perfect for AI infrastructure.

• Abundant natural gas.

• Existing power infrastructure.

• Large population centers nearby.

• Industrial land.

Amazon alone announced plans for tens of billions of dollars of investment there.

Yet on August 18, Pennsylvania Governor Josh Shapiro imposed new requirements on AI data-center development.

Projects were removed from a fast-track permitting program. Developers face stronger environmental and transparency requirements. Local community approval matters more. State agencies were barred from signing certain confidentiality agreements with developers.

That is a profound signal.

Governments spent the first phase of the AI boom asking:

How do we attract data centers?

The next phase may be:

How do we control them?

Someone Can Send You $0.01 of Sanctioned Bitcoin — and Freeze Your Exchange Account.

Bitcoin was designed so nobody can stop you from receiving money. Regulated exchanges are discovering the terrifying flip side: nobody can stop someone from sending it to you either.

Texas Is Learning the Same Lesson

Texas looked like the obvious American AI paradise.

• Cheap land.

• An enormous energy industry.

• A deregulated electricity market.

• Business-friendly politics.

• Rapid construction.

Then the scale of proposed demand became enormous.

Texas has already paused some approvals amid concern that data-center growth could threaten grid reliability, and official projections for future electricity demand have been revised as the state reassesses how quickly new load can safely connect.

That tells us something investors should remember.

Even politically friendly jurisdictions have physical limits.

You can deregulate planning. You cannot deregulate Kirchhoff’s laws.

New York Has Gone Even Further

New York imposed a temporary moratorium on some large new data-center developments.

Again, the direction matters more than any individual policy.

The AI industry assumed states would compete to attract every gigawatt developers offered. Instead, some governments are beginning to ration the buildout.

Why?

Because residents care about affordability.

• Grid reliability.

• Water.

• Land.

• Noise.

• Environmental impact.

AI companies are discovering that being strategically important does not make them politically frictionless.

Silicon Valley Has Met NIMBY

The technology industry has encountered regulation before.

• Privacy rules.

• Antitrust.

• Content moderation.

• Labor laws.

But data centers create a fundamentally different political problem.

They are physical.

You can see them. You can hear the generators. You can see the transmission lines. You can watch farmland disappear. You can notice construction traffic. You can worry about water pressure. You can connect your rising electricity bill to a giant facility down the road.

That creates localized opposition.

And localized opposition is extraordinarily powerful in infrastructure politics.

People may believe AI is strategically important.

They may believe America must beat China. They may love ChatGPT. They may own Nvidia stock.

And they can still say:

Not next to my house.

Malaysia Is Already Showing What Comes Next

This phenomenon is not uniquely American.

Malaysia became one of Asia’s fastest-growing data-center markets after Singapore constrained expansion.

Huge investments followed. Then the local backlash began.

Residents in Johor protested over environmental disruption and fears surrounding water availability as enormous data-center projects expanded. Authorities responded with stricter vetting requirements around energy, local content and environmental impact.

Again, the pattern is emerging.

Singapore restricts. Investment moves to Malaysia. Malaysia experiences pressure. Regulation increases.

Capital searches for the next location.

AI compute is becoming geographically nomadic.

Not because companies want to move. Because the easiest places eventually become constrained.

Water Is Becoming Part of the AI Stack

The industry prefers to talk about FLOPS.

Communities may care more about liters.

Cooling enormous computing clusters requires removing enormous quantities of heat.

There are different cooling architectures, and newer systems can dramatically reduce water consumption. But trade-offs remain.

A recent OpenAI-linked Australian data-center project abandoned a plan to use recycled water because the infrastructure and permissions needed for a new pipeline were not available.

The developers shifted toward a waterless cooling approach that requires more energy.

That small story captures the entire AI infrastructure challenge.

• Want to save water?

• You may consume more power.

• Want cleaner power?

• You may need transmission.

• Want transmission?

• You need permits.

• Want faster permits?

Local residents may object.

Solve one constraint and another appears. This is systems engineering at continental scale.

And Then There Are Transformers

Few people building AI models wake up excited about electrical transformers.

They should.

Transformers are becoming one of the most strategically important components of the AI economy.

Electricity generated at a power plant is useless to your GPU cluster unless the grid can transform and deliver it at the appropriate voltage.

Large transformers are difficult to manufacture.

Supply chains are concentrated. Factories take time to expand. Some conventional equipment can reportedly involve waits approaching several years.

Imagine the absurdity.

You can raise $10 billion. Order state-of-the-art AI accelerators. Secure a world-class model team. Buy land.

And then discover that your schedule depends on a giant piece of electrical equipment designed using technology far older than artificial intelligence.

Welcome to the physical AI economy.

The Next Nvidia May Sell Something Boring

This creates an interesting investment implication.

The greatest beneficiaries of the AI boom may increasingly come from industries that look nothing like AI.

• Generators.

• Transformers.

• Switchgear.

• Cooling systems.

• Power electronics.

• Steel.

• Copper.

• Construction equipment.

• Electrical contractors.

• Gas turbines.

• Nuclear engineering.

• Grid software.

• Transmission equipment.

Reuters recently documented how the American data-center boom is already transforming demand across industrial supply chains.

Generac is investing hundreds of millions to expand commercial-generator production. Siemens is expanding manufacturing.

Suppliers of cooling equipment, cables, bearings, prefabricated infrastructure and construction machinery are experiencing AI-driven demand.

AI is slowly turning old-economy companies into technology infrastructure companies.

That may be one of the most important rotations of the next decade.

Nvidia Understands This Better Than Anyone

Watch what Nvidia is doing.

The company could simply sell GPUs. Instead, it is increasingly getting involved in financing and infrastructure.

The Ohio project is especially revealing.

Nvidia is willing to support enormous financial guarantees around a data center because selling GPUs is useless if customers cannot deploy them.

This is the next logical evolution of Nvidia’s business.

First Nvidia solved compute. Then networking. Then software. Now it increasingly has to help solve infrastructure.

The deeper Nvidia moves into the stack, the clearer the message becomes:

The chip is no longer the limiting factor.

Deployment is.

Compute Is Becoming Project Finance

This may fundamentally change Silicon Valley’s financial architecture.

Software companies traditionally raised venture capital. Infrastructure companies use different tools.

• Debt.

• Project finance.

• Long-term leases.

• Power-purchase agreements.

• Infrastructure funds.

• Municipal negotiations.

• Real-estate vehicles.

• Utility contracts.

• Bonds.

• Guarantees.

AI is forcing technology companies into this world.

An 8-gigawatt compute campus looks less like a startup and more like an LNG terminal.

The economics are different. The timelines are different. The risks are different. And the investors funding them may be different.

This is why private credit, infrastructure funds, utilities and asset managers are becoming increasingly central to AI.

The technology boom is becoming a capital-intensive industrial boom.

The AI Race Is Becoming a Sovereign Race

Once electricity becomes the constraint, national differences matter enormously.

• Countries with abundant energy gain an advantage.

• Countries capable of approving transmission quickly gain an advantage.

• Countries with simplified permitting gain an advantage.

• Countries with available land gain an advantage.

• Countries willing to build nuclear, gas, renewables or all three gain an advantage.

• Countries that cannot build infrastructure quickly may produce brilliant AI research and still struggle to deploy it at scale.

That creates a strange possibility.

The next great AI hub may not emerge where the best AI researchers live. It may emerge where the electrons are easiest to obtain.

The Gulf Understands This

Saudi Arabia and the UAE possess something increasingly valuable.

• Energy.

• Capital.

• Land.

• Centralized decision-making.

They can approve enormous infrastructure projects relatively quickly.

That combination makes the Gulf strategically relevant to AI despite not historically dominating semiconductor design or software.

If compute becomes an energy transformation business, energy-rich nations move up the value chain. Instead of exporting hydrocarbons, they can export intelligence produced using domestic power.

That is a profound shift.

Europe Has the Opposite Problem

Europe has remarkable AI researchers.

Strong universities. Sophisticated industrial companies. Large markets.

But infrastructure can be difficult to build.

Permitting is complex. Power prices can be high. Urban land is expensive. Grid connections can be slow.

This is precisely why European hyperscale AI projects are moving toward secondary and rural locations where land and electricity are easier to secure.

Europe does not merely have an AI startup problem. It may have an infrastructure-speed problem.

And no subsidy for foundation models can fully compensate for an inability to build the power systems required to run them.

China Has a Different Advantage

China’s AI weaknesses are frequently discussed.

Access to Nvidia’s best chips is constrained. Leading-edge semiconductor manufacturing remains challenging. American export controls matter.

But China has another advantage Western investors sometimes underestimate:

it can build enormous amounts of infrastructure.

• Power generation.

• Transmission.

• Industrial parks.

• Factories.

• High-speed rail.

• Renewables.

China’s challenge is not that infrastructure constraints disappear entirely — they do not. Its electrical grid also experiences massive bottlenecks and renewable curtailment.

But China’s institutional capacity for large-scale construction remains strategically important.

The AI race may therefore increasingly become a contest between American leadership in frontier compute and China’s ability to deploy physical infrastructure rapidly.

That is far more complex than “who has the best model?”

America’s Greatest AI Risk May Be Self-Imposed

The United States has almost everything required to dominate AI.

• The world’s leading chip designer.

• The strongest cloud companies.

• Deep capital markets.

• Enormous natural-gas resources.

• World-class universities.

• Frontier AI laboratories.

• Entrepreneurial culture.

But America has repeatedly demonstrated difficulty building physical infrastructure quickly.

• Transmission lines.

• Nuclear plants.

• Housing.

• Rail.

• Factories.

• Large industrial projects.

• Permitting delays can stretch for years.

• Local opposition can stop enormous developments.

If AI becomes infrastructure constrained, America’s biggest rival may not be China. It may be its own inability to build.

That should worry Washington far more than another benchmark from DeepSeek.

The New AI Moat Is a Permit

For years, technology investors searched for software moats.

• Network effects.

• Data advantages.

• Developer ecosystems.

• Switching costs.

• Proprietary algorithms.

The next great AI moat may look incredibly mundane.

A site with: 500 megawatts already approved.

A transmission connection. Water rights. Transformer deliveries secured. Fiber access. Local political support. Environmental permits. Expansion rights.

Suddenly that piece of land becomes extraordinarily valuable.

Competitors may have identical GPUs. They may even have better models. But if your cluster turns on three years before theirs, you win.

Infrastructure time becomes competitive advantage.

This Changes Data-Center Valuations

For investors, not all announced gigawatts are equal. This distinction will become increasingly important.

One developer may announce a five-gigawatt pipeline. Another may announce two.

The first sounds larger.

But what if only 500 megawatts are actually permitted?

What if the grid connection is speculative? What if transformers have not been ordered? What if the community opposes construction? What if water availability is unresolved?

Meanwhile, the smaller project already has secured power, permits and equipment.

Which capacity is more valuable?

Clearly the second.

AI infrastructure investors will increasingly have to distinguish between:

announced megawatts and executable megawatts.

That difference could become enormous.

“Time to Power” May Become the Most Important Metric in AI

Forget GPU count for a moment. Ask something else.

How long from acquiring land to running compute?

That is time to power.

And it may become one of the most important metrics in the AI economy.

A company that can energize a data center in eighteen months may beat one requiring four years even if the second company has cheaper land.

The value of speed rises dramatically when model demand is expanding exponentially.

Every month of delay represents lost compute revenue.

Lost training capacity. Lost inference. Lost customers. Potentially lost technological leadership.

That means countries, utilities and developers capable of compressing time-to-power will command a premium.

The Industry Will Start Building Its Own Electricity

The predictable response is already beginning.

If the grid cannot provide power fast enough, AI companies will attempt to bring generation with them.

• Gas turbines.

• Solar.

• Battery storage.

• Small modular reactors.

• Large nuclear plants.

• Fuel cells.

• Microgrids.

• Whatever works.

This does not mean data centers will completely disconnect from public grids. But the boundary between technology company and utility will blur.

• Microsoft signs nuclear deals.

• Google signs power agreements.

• Amazon buys energy-linked infrastructure.

• Data-center developers build generation.

• Nvidia supports site financing.

The logical endpoint is fascinating.

Big Tech may become one of the largest investors in electricity infrastructure in modern history.

AI Could Rebuild America’s Grid

There is an optimistic version of this story.

The AI boom could force investments America needed anyway.

• New transmission.

• Modern substations.

• Domestic transformer manufacturing.

• New nuclear capacity.

• More gas generation.

• Battery storage.

• Advanced grid controls.

• Skilled construction jobs.

• Manufacturing expansion.

A grid built for twentieth-century electricity demand may finally be rebuilt for the twenty-first.

US electricity consumption is already expected to set new records in both 2026 and 2027, with AI and other data-center workloads contributing to the increase.

AI may become the economic catalyst that finally makes large-scale grid investment politically and financially unavoidable.

That would be a huge positive externality. But only if the costs and benefits are distributed intelligently.

Bitcoin Core Just Ran Inside a zkVM: What If New Nodes Could Verify 17 Years of Bitcoin in Milliseconds?

Bitcoin’s “Don’t Trust, Verify” Era Could Be Rewritten by Zero-Knowledge Proofs—Turning Years of Historical Computation Into a Cryptographic Receipt.

Otherwise the Backlash Will Be Brutal

There is also a darker scenario.

• Electricity prices rise.

• Residents see enormous data centers receiving preferential treatment.

• Water use becomes controversial.

• Tax incentives look excessive.

• Local employment disappoints.

• Grid upgrades are socialized.

• Corporate profits are privatized.

• Then voters revolt.

• Projects are delayed.

• Moratoriums spread.

• Politicians discover data-center opposition is popular.

The same infrastructure Wall Street values as essential becomes politically toxic.

That is why companies cannot treat communities as another permitting checkbox.

Political legitimacy is becoming part of the AI supply chain.

The Biggest AI Companies Will Need to Become Better Neighbors

This sounds almost quaint. It isn’t.

If a company wants to build a multi-gigawatt facility near a town, it increasingly needs a compelling answer to basic questions.

• Will my electricity bill rise?

• Will my water supply be affected?

• How many permanent jobs will you create?

• What happens to property values?

• Who pays for the transmission upgrade?

• Where does backup generation come from?

• How loud will the facility be?

• What tax revenue does the community receive?

• Can the grid still keep my lights on during a heat wave?

The companies that answer those questions well will build. The companies that dismiss them may not. That is the new reality.

AI Is Becoming an Infrastructure Civilization

This is perhaps the biggest conceptual shift.

Generative AI began as software.

A chat window. A model. An API.

Something almost weightless.

Then we looked underneath. And found giant GPU clusters.

Under the GPUs were data centers. Under the data centers were cooling systems. Under the cooling systems were electricity networks.

Under the networks were gas fields, nuclear reactors, solar farms, transmission lines, transformers, copper mines and permitting agencies.

The deeper you go, the more AI begins to resemble heavy industry.

The cloud was never really a cloud. It was steel, concrete and electricity. AI is simply forcing us to notice.

The Next Phase Will Reward Builders

The first winners of the AI boom were obvious.

• Nvidia.

• Microsoft.

• OpenAI.

• Anthropic.

• Broadcom.

• Memory companies.

• Cloud providers.

The next generation of winners may look different.

• Utilities that can add generation quickly.

• Electrical-equipment manufacturers.

• Transformer producers.

• Cooling companies.

• Power-electronics specialists.

• Nuclear developers.

• Gas-turbine manufacturers.

• Engineering firms.

• Data-center developers with secured power.

• Landowners controlling strategically located sites.

• Infrastructure funds.

• Grid software companies.

The AI portfolio is expanding from Silicon Valley into the industrial economy.

The smartest model may still capture enormous value. But the model cannot run without the machine.

And the machine cannot run without the grid.

The New Question

For the past three years, executives asked:

How many GPUs can we get?

That question is being replaced.

• How many megawatts can we secure?

• How quickly can we connect?

• How much will power cost?

• Can we obtain transformers?

• Can we cool the cluster?

• Will the community approve it?

• Can we get the permits?

• How quickly can construction start?

Those questions sound less exciting than model benchmarks. They may ultimately matter more. Because we are reaching the point where money is not enough.

GPUs are not enough. Models are not enough. Even electricity itself is not enough.

You need the legal, political and physical ability to assemble all of them in one place.

That ability is becoming scarce. And scarcity creates value.

AI’s Next Bottleneck Is Permission

Silicon Valley is accustomed to a world where the winning strategy is simple.

Move fast. Raise capital. Hire brilliant people. Build the product. Scale globally.

Physical infrastructure does not obey those rules.

You cannot deploy a transmission line with a software update. You cannot A/B test a nuclear reactor. You cannot prompt-engineer a transformer into existence. And you cannot ignore the people whose communities host the infrastructure.

The AI race is therefore entering a fundamentally different phase.

The first phase rewarded whoever could design the best chips. The second rewarded whoever could train the best models.

The next may reward whoever can build the physical systems required to run intelligence at industrial scale.

That is a very different competition. And it leads to an uncomfortable conclusion.

The world’s largest technology companies may soon have enough money to buy almost any GPU they want.

They may have enough capital to build almost any model they can imagine.

What they cannot buy instantly is a functioning electrical grid, a five-year permit compressed into five weeks, or community approval.

That is why AI’s next great scarcity will not necessarily be silicon.

It will be something much older.

Land. Power. Infrastructure. And permission to build.

Share Sylvain Saurel’s Newsletter

Subscribe now

Everyone was watching Nvidia’s revenue. They should have been watching the footnotes.

Nvidia just reported another absurd quarter.

Revenue: $96.2 billion. Up 106% from a year ago.

Data Center revenue: $89 billion. Up 117%.

Gross margin: 75%.

And Jensen Huang guided the current quarter toward approximately $108 billion in revenue.

Those numbers are spectacular.

But buried beneath the earnings fireworks was another number that may tell us much more about where the AI boom is heading.

$279 billion.

That is how much Nvidia now has in future supply and capacity commitments.

One quarter earlier? $119 billion.

In roughly three months, Nvidia increased those commitments by about $160 billion. More than doubled them. And according to Nvidia, the increase was driven primarily by one thing: Memory.

That changes the story.

For the past three years, investors have treated the AI boom as a race for GPUs.

• Who has the best accelerator?

• Who can challenge Nvidia?

• How many Blackwell systems can Nvidia ship?

• When will Rubin arrive?

Those questions still matter. But Nvidia’s latest filing reveals that the AI arms race is entering a very different phase.

The battle is no longer simply about designing the best chip. It is about securing the physical resources required to manufacture intelligence at scale before everybody else gets them.

• Memory.

• Foundry capacity.

• Advanced packaging.

• Networking.

• Power.

• Land.

• Cooling.

• Data centers.

And eventually, perhaps, anything that stands between a model and another trillion tokens.

AI is becoming a resource war. And Nvidia just placed one of the largest reservations in industrial history.

The Most Important Number Wasn’t $96 Billion

Wall Street understandably obsessed over Nvidia’s second-quarter results. They were extraordinary.

$96.2 billion in quarterly revenue.

$59.7 billion in GAAP net income.

$63.7 billion in operating income.

$89 billion of Data Center revenue.

The kind of figures that would have looked absurd for a semiconductor company only a few years ago. But revenue tells you what happened.

Commitments tell you what Nvidia thinks is coming. And this is where things become fascinating.

As of July 26, Nvidia disclosed:

$279 billion in supply and capacity commitments.

That compares with just $119 billion one quarter earlier.

The company says these commitments are designed to secure critical components for demand extending across several years and across both current and future product architectures.

This isn’t Nvidia cautiously preparing for AI demand. This is Nvidia behaving as though scarcity itself has become the enemy.

Nvidia Is Reserving the Future

Look at the schedule.

Nvidia’s $279 billion supply-and-capacity commitment breaks down approximately like this:

• $92 billion during the remainder of fiscal 2027.

• $87 billion in fiscal 2028.

• $88 billion in fiscal 2029.

Then much smaller amounts thereafter.

Pause on that.

Nvidia has effectively lined up nearly $270 billion of supply and capacity over roughly the next two and a half fiscal years.

This isn’t inventory sitting in a warehouse today. And it would be misleading to describe the whole $279 billion as a pile of debt Nvidia suddenly owes tomorrow.

Some supplier agreements can be cancelled, rescheduled or adjusted before firm orders are placed, potentially with additional costs. But the strategic message is impossible to miss.

Nvidia believes future demand is sufficiently large — and future supply sufficiently constrained — that it is willing to make enormous commitments today to ensure the components exist tomorrow.

That’s what companies do when they believe access matters more than price.

And the Biggest Problem Is Memory

For years, GPUs were the scarce thing. Now the bottleneck is moving.

Nvidia says the increase in supply commitments was primarily related to memory procurement.

Its filing describes the commitments as supporting data-center infrastructure systems and says they relate primarily to memory and manufacturing facilities needed across current and future architectures.

That matters because modern AI accelerators aren’t useful in isolation.

A GPU without enough high-bandwidth memory is like a Formula One engine without fuel.

AI models need to move enormous amounts of data between memory and compute extremely quickly.

As models get larger…

As context windows expand…

As inference increases…

As AI agents run persistently…

As multimodal models process video, audio, text and sensor data…

Memory requirements explode.

The GPU may remain the star. But the rest of the system increasingly determines how bright it can shine.

AI Has Found Its Next Oil

Think about what oil represented to the industrial economy.

Factories could exist. Cars could exist. Planes could exist. But without enough cheap energy, the entire machine slowed down.

HBM and advanced memory are starting to play a similar role inside the AI economy.

Nvidia can design increasingly powerful accelerators. TSMC can manufacture extraordinary silicon. But if there isn’t enough memory available at the right performance, yield and price?

The whole AI infrastructure buildout slows.

That gives memory suppliers something they haven’t traditionally enjoyed for long: pricing power.

Memory historically has been a brutal commodity business.

Boom. Shortage. Capacity expansion. Oversupply. Collapse. Repeat.

AI may be changing the amplitude of that cycle.

Maybe even its structure. Because this isn’t simply another PC replacement wave. We’re talking about hyperscalers planning infrastructure at almost unimaginable scale.

• Microsoft.

• Amazon.

• Google.

• Meta.

• Oracle.

• AI labs.

• Sovereign governments.

• Neo-clouds.

• Enterprises.

Everybody wants compute. Everybody needs memory. And everybody is trying to secure it at the same time.

Nvidia Is Now Competing With Its Own Customers for Resources

This creates a strange situation.

Nvidia sells systems to Microsoft. Google. Amazon. Meta. Oracle.

But those companies are themselves making enormous commitments to data centers, networking, memory, power and infrastructure.

They need many of the same industrial inputs.

So does Nvidia. So do AMD. Broadcom. AI startups. Cloud providers. Server manufacturers. Governments building sovereign AI.

Everybody is reaching into the same finite physical economy.

That is why AI is ceasing to look like a normal software boom. Software can be copied at nearly zero marginal cost.

Data centers cannot. HBM cannot. Transformers cannot. Transmission lines cannot. Leading-edge wafers cannot. Advanced packaging cannot. Power plants cannot.

This is where the digital revolution crashes into physical reality.

Bitcoin’s First Quantum-Proof Transaction Is Here. Why You Should Still Be Worried.

StarkWare just executed a historic mainnet transaction immune to quantum computers. But this massive, computationally brutal lifeboat proves the clock is still ticking for a mandatory network upgrade.

AI Is Becoming Industrial Policy

This transformation has enormous implications.

For the first phase of generative AI, the strategic advantage was intellectual.

Models. Algorithms. Researchers. Training methods. Data.

Then it became computational.

GPUs. CUDA. Clusters.

Now it is becoming industrial.

• Who has the fabs?

• Who has the packaging?

• Who has the memory?

• Who has the electricity?

• Who has the permits?

• Who controls the supply chains?

• Who can finance $100 billion data centers?

• Who can guarantee delivery three years from now?

The companies capable of answering those questions will increasingly control the pace of AI development.

This is why semiconductors have become national-security assets. And why the AI race is merging with geopolitics.

Nvidia Is Still “Fabless” — But Look at What That Now Means

Nvidia doesn’t own the advanced fabs producing its leading GPUs. That has historically been one of the beauties of its model.

Let TSMC spend tens of billions building fabrication plants. Let memory manufacturers build fabs. Let suppliers handle manufacturing.

Nvidia concentrates on architecture, software and system design.

Capital-light. Incredible margins. Elegant.

Except when demand becomes so enormous that suppliers themselves become the constraint.

Then being fabless creates another challenge:

You must convince someone else to build enough capacity for your future.

Read more

A new analysis puts a brutal number on America’s strangest financial experiment. But the bigger story isn’t who won and who lost — it’s what happens when the president, the regulator, the token issuer, the stablecoin promoter, and now a federally supervised crypto bank begin occupying the same financial ecosystem.

Subscribe now

Seven years ago, Donald Trump thought cryptocurrency was basically nonsense.

In 2019, he dismissed Bitcoin and other cryptocurrencies as highly volatile assets “based on thin air.”

Today?

Crypto is one of the biggest sources of income in the Trump business empire.

According to Donald Trump’s 2025 financial disclosure, reviewed by Reuters, his family’s crypto ventures generated more than $1.4 billion of reported income during the year.

Almost $800 million came through World Liberty Financial. Another roughly $635 million came from Trump memecoin sales.

That alone would be an extraordinary transformation. But now comes the other number.

Public Citizen, the Washington consumer-advocacy organization, published a new analysis estimating that investors across several Trump-linked crypto ventures are collectively sitting on at least:

$4.7 billion in losses.

The organization breaks that estimate down roughly like this:

$TRUMP memecoin: $3.2 billion

$WLFI: at least $1 billion

Trump Media’s digital-asset treasury: $450 million

Trump digital trading cards: at least $9.3 million

USD1 stablecoin: $0

The total: at least $4.7 billion underwater.

Put the two numbers beside each other, and you get one of the most explosive financial stories in American politics:

Trump reported more than $1.4 billion in crypto-related income.

Investors in Trump-linked crypto products are estimated to be $4.7 billion underwater.

But before anyone turns that into a simplistic equation — Trump gained $1.4B → investors lost $4.7B — we need to be precise.

That is not what the data prove. And the truth is actually more interesting.

Bitcoin Just Entered the American Mortgage System Without Being Sold.

Bitcoin Is Evolving From an Asset You Buy Into Monetary Property You Can Borrow Against — Without Giving Up the Upside.

First: $4.7 Billion Was Not Simply Handed to Trump

This distinction matters enormously.

Public Citizen itself acknowledges that much of its $4.7 billion estimate represents unrealized losses.

In other words: someone bought an asset. The asset fell. Their position is now worth less. They may not have sold.

That is very different from saying $4.7 billion moved directly from investor bank accounts into Donald Trump’s pocket.

Some investors also made money.

Early buyers of certain Trump-linked tokens obtained dramatically better entry prices than people who entered later.

Crypto markets work like that.

Someone buys $TRUMP at a low valuation. Someone else buys after the price explodes. The token collapses.

The first investor may still be profitable. The second may be down 80%.

Nothing mysterious there.

That happens every day in speculative markets. But there is something unusual here.

The person at the center of the brand powering these assets is not merely:

• a celebrity,

• a founder,

• an influencer,

• or a billionaire.

He is the President of the United States.

And his administration simultaneously controls enormous portions of the regulatory environment surrounding the industry in which those products operate.

That is where this becomes much bigger than another memecoin collapse.

The Real Story Is the Asymmetry

Forget Trump for a moment.

Imagine a hypothetical president.

President Smith launches:

• a memecoin,

• a governance token,

• a stablecoin,

• NFTs,

• a publicly traded digital-asset treasury,

• and a financial company applying for federal banking privileges.

Then President Smith’s administration begins rewriting the rules governing cryptocurrency.

You would immediately recognize the governance question.

Not because President Smith necessarily committed a crime. Not because every policy decision automatically benefits his personal holdings. But because the incentives now overlap.

Public power and private financial exposure have entered the same room.

That alone deserves scrutiny. And Trump has taken that theoretical problem into completely unexplored territory.

Bitcoin’s Rally Is Becoming a BlackRock Market: 70% of ETF Flows Came Through IBIT.

Bitcoin remains decentralized at the protocol layer — but as IBIT absorbs nearly 70% of ETF inflows, Wall Street’s gateway to Bitcoin is becoming strikingly concentrated.

Crypto Has Become One of Trump’s Biggest Businesses

The scale matters.

Reuters’ examination of Trump’s 2025 disclosure found that his companies reported nearly $800 million from World Liberty Financial.

That included more than $520 million from token sales and over $250 million from selling interests in the business.

He separately reported approximately $635 million from sales of the TRUMP memecoin.

Compare that with the traditional image of the Trump empire.

Hotels. Golf courses. Real estate. Licensing. Mar-a-Lago.

Those businesses still generate hundreds of millions. But crypto has become transformational.

Reuters calculated earlier this year that the Trump family had generated at least $2.3 billion from crypto-related ventures since Trump returned to the White House in 2025.

This isn’t a side hustle anymore. It is an economic pillar. And once you understand that, every crypto-policy discussion in Washington gains another dimension.

The Memecoin Is Where the Numbers Become Brutal

Let’s start with $TRUMP.

Memecoins are unusual assets because almost nobody pretends they have traditional fundamental valuation.

No discounted cash flow. No dividend. No claim on a factory. No mortgage portfolio. No bond coupon.

Value comes overwhelmingly from:

• attention,

• community,

• scarcity,

• speculation,

• and the power of the brand.

Donald Trump possesses arguably one of the most powerful brands on Earth.

So when a Trump-branded token launches, attention arrives instantly. And attention in crypto can become liquidity incredibly fast.

That creates opportunity.

It also creates an extraordinarily unequal playing field between early entrants and late buyers.

Public Citizen estimates that investors in $TRUMP are collectively down about:

$3.2 billion.

That number makes up the majority of the entire $4.7 billion estimate.

Again: these are not necessarily realized losses. And they are not automatically Trump’s profit. But the distributional pattern matters.

The people entering early capture the explosive first leg.

The people entering because the asset has become culturally unavoidable often provide the liquidity that allows early participants to monetize.

That is normal memecoin economics. The difference is the identity attached to the coin.

A Presidential Memecoin Changes the Game

Think about ordinary political merchandise.

Read more

Subscribe now

There is an alternate universe, not entirely disconnected from our own, where the most powerful banker in the world never becomes a banker at all.

In this parallel reality, he trades his tailored Wall Street suits for the Pacific Northwest’s fleece vests. He navigates supply chain logistics instead of interest rate yields. He builds fulfillment centers instead of fortress balance sheets.

In the late 1990s and early 2000s, as the dot-com boom was rewriting the rules of global commerce, Jeff Bezos was building what would become the “Everything Store.” But Bezos, a visionary founder, knew he needed battle-tested operational executives to help scale Amazon from a wildly ambitious bookstore into a global infrastructure titan. He needed a general who knew how to manage vast complexities, ruthless margins, and thousands of employees.

He looked toward Wall Street, and he found Jamie Dimon.

Bezos offered Dimon the role of CEO at Amazon. For almost any executive on the planet, an offer from Jeff Bezos to lead a rising tech behemoth would be the capstone of a career—an undeniable golden ticket. Silicon Valley was the future; traditional banking was the stodgy past. The financial upside alone was unfathomable. Yet, Dimon turned it down.

His explanation was as telling as it was grounded. He didn’t dismiss the genius of what Bezos was building. He didn’t scoff at the tech industry. Instead, he looked inward at his own competencies, his own passions, and the very nature of mastery.

“I loved what he did, it just was beyond - I spent my whole life in financial services - it’s a little bit like playing tennis your whole life and then going to play golf.”

This singular decision — and the astonishing series of events that followed it — serves as one of the most profound masterclasses in modern business history. Dimon’s refusal to switch sports, his subsequent firing from the empire he helped build, and his triumphant return to the arena by putting his own immense wealth on the line form a blueprint for what true corporate leadership looks like.

It is a story that transcends banking. It is a story about the vital difference between a “prestigious job” and a lifelong mission. It is about the courage of conviction, the extreme ownership required to truly lead, and the conscious choice to build something lasting rather than simply parachute into someone else’s success story.

Part 1: The Seduction of Silicon Valley and the Amazon Proposition

To fully grasp the magnitude of the Amazon offer, one must understand the macroeconomic and cultural zeitgeist of the era. At the turn of the millennium, a massive brain drain was occurring. Wall Street’s brightest minds, previously content with massive year-end bonuses and corner offices in Manhattan, were being lured to the West Coast. The siren song of stock options, disruptive technology, and a casual, world-changing ethos was intoxicating.

Jeff Bezos, ever the pragmatist, understood that a visionary founder needs a world-class operator. The tech industry has a long history of this dynamic. Bill Gates had Steve Ballmer; later, Larry Page and Sergey Brin would bring in Eric Schmidt as the “adult supervision” at Google, and Mark Zuckerberg would hire Sheryl Sandberg to monetize Facebook. Bezos was looking for his operational counterpart, a CEO who could bring disciplined execution, rigorous capital allocation, and relentless operational efficiency to Amazon’s sprawling ambitions.

Jamie Dimon fit the profile perfectly. He was notoriously rigorous, obsessed with details, a master of cutting bloat, and possessed an innate understanding of how money moved. He was a force of nature who had spent decades in the trenches of massive corporate mergers and integrations.

The pitch was undoubtedly compelling. Leave the highly regulated, cyclical, and often archaic world of commercial and investment banking. Step into the frontier of the internet. Become the chief executive of a company that was fundamentally altering human consumption.

The temptation to accept must have been immense. In the tech world, failure is often framed as “learning,” and success brings unprecedented wealth and cultural adulation. In banking, success is expected, and failure brings systemic economic collapse and congressional hearings.

Yet, Dimon recognized a fundamental truth about himself that many executives, blinded by ambition or money, fail to see. He understood the nature of his own edge.

Part 2: Tennis, Golf, and the Philosophy of Deep Mastery

When Dimon articulated his refusal by comparing the transition to a lifelong tennis player suddenly taking up golf, he was summarizing decades of psychological and economic theory regarding specialization.

The Illusion of the Generalist CEO

There is a pervasive myth in modern corporate culture: the idea of the “Generalist Executive.” This is the belief that management is a universal skill, disconnected from the underlying product or service. According to this theory, a great leader can seamlessly transition from selling consumer packaged goods to manufacturing automobiles, or from writing software to running a hospital network.

Dimon rejected this premise entirely. Management, in his view, is deeply intertwined with domain expertise.

• The Muscle Memory of Finance: In banking, risk isn’t an abstract concept; it is the core product. A master banker develops a visceral, intuitive feel for credit cycles, interest rate environments, regulatory shifts, and counterparty risk. This is the “tennis swing” Dimon had been perfecting since he first followed his mentor, Sandy Weill, into the industry in the early 1980s.

• The Nuance of the Game: In tennis, a player reacts to the spin of the ball, the surface of the court, and the momentum of the opponent in fractions of a second. This is muscle memory built over tens of thousands of hours. Dimon had that muscle memory for balance sheets. He could look at a bank’s financials and instantly spot the hidden rot or the unexploited value.

• The Danger of the Unknown: Golf, conversely, is a game of entirely different mechanics. It requires a different stance, a different grip, and a different mental approach. While a world-class tennis player might possess excellent hand-eye coordination and athletic discipline, putting them on the first tee at Augusta National against lifelong professional golfers would be a recipe for disaster.

Dimon knew that if he went to Amazon, he would be relying on his general intelligence rather than his deep, specialized wisdom. He would be playing a new game against people who had lived and breathed technology their entire lives. He respected Amazon’s arena enough to know it wasn’t his.

He chose to stay in the world he knew, where he wasn’t just an athlete with good reflexes, but a master of the specific game being played. He chose to remain on the tennis court.

Part 3: The Ascent and the Empire of Citigroup

To understand why Dimon’s commitment to banking was so deeply rooted, we must look at the empire he helped build. Dimon’s career was not a typical corporate ladder-climb. It was an entrepreneurial conquest disguised as corporate finance.

For 17 years, Jamie Dimon was the indispensable right-hand man to Sandy Weill, one of the most legendary and rapacious dealmakers in Wall Street history. Together, they operated like a father-son duo, executing a relentless string of acquisitions that fundamentally reshaped the American financial landscape.

The Roll-Up Strategy

Starting at a small consumer finance company called Commercial Credit in 1986, Weill and Dimon embarked on a breathtaking corporate shopping spree.

• They acquired Primerica.

• They bought Shearson Lehman.

• They took over Travelers Group.

• They acquired the venerable investment bank Salomon Brothers.

In each of these massive, complex mergers, Sandy Weill was the visionary dealmaker, the charismatic frontman who charmed boards and envisioned the synergies. But Jamie Dimon was the mechanic. Dimon was the one who went into the newly acquired, often bloated companies, rolled up his sleeves, fired redundant staff, integrated incompatible IT systems, slashed unnecessary expenses, and made the math work.

Their partnership culminated in the ultimate Wall Street megadeal in 1998: the merger of Travelers Group and Citicorp to form Citigroup. It was the largest corporate merger in history at the time, a $140 billion marriage that created the world’s first truly modern universal bank. It was the realization of a decades-long dream.

Dimon was named President of Citigroup. He was the heir apparent. He had spent his entire adult life building this specific, gargantuan institution. The tennis court he had built was now the largest in the world, and he was poised to be its undisputed champion.

Part 4: The Fall — “A Hell of a Run”

And then, abruptly, shockingly, the music stopped.

In November 1998, just months after the triumph of the Citigroup merger, Sandy Weill summoned Dimon into a weekend meeting at an executive retreat. The relationship between the two men, always volatile, passionate, and familial, had fractured.

Read more

Subscribe now

For decades, the United States possessed a weapon more powerful than almost anything in its military arsenal.

Not an aircraft carrier. Not a stealth bomber. Not a nuclear submarine.

The dollar.

If Washington wanted to isolate a country, it didn’t necessarily have to invade it. It could disconnect it.

Banks would stop dealing with it. Insurers would back away. Shipping companies would disappear. Multinationals would leave.

Foreign governments would calculate that access to the American financial system mattered more than trade with the target.

This worked because the choice was brutally simple:

Do business with the sanctioned country — or do business with America.

Most companies chose America.

Iran has lived under some version of that pressure for decades. But the latest confrontation with Tehran is testing something much bigger than Iran.

Because this time, Washington’s sanctions strategy increasingly runs into the world’s second-largest economy.

China.

And Beijing may be large enough to change the equation.

Washington Just Raised the Stakes

On August 24, the Trump administration expanded its sanctions campaign against Iran. Nearly 60 Iran-linked individuals, companies, vessels, and networks were targeted.

The sanctions architecture was widened beyond oil to encompass areas including shipping, technology, aviation, gold, cryptocurrency, cyber activity, and military procurement.

The message from Treasury Secretary Scott Bessent was even more important than the individual designations:

Countries continuing to do business with Iran could eventually face a choice between Tehran and access to the dollar-based financial system.

That is secondary sanctions in their most powerful form.

Washington isn’t merely saying: “American companies cannot trade with Iran.”

It is saying: “Foreign companies may also pay a price if they trade with Iran.”

That is how American financial power becomes global power. But there was an obvious omission.

Washington did not immediately hit the biggest Chinese financial institutions facilitating trade with Iran.

And that omission tells you almost everything.

Because China Is the Elephant in the Room

Iran needs buyers for its oil. China has been the buyer. A very big one.

The U.S. Treasury estimated earlier this year that China purchases approximately 90% of Iranian oil exports, with independent Chinese “teapot” refineries accounting for much of that trade.

Reuters puts China’s share of Iran’s shipped oil at more than 80%. Either way, the strategic reality is obvious.

If China stops buying Iranian oil, Tehran loses an enormous economic lifeline.

If China keeps buying it, Washington faces a much harder sanctions problem.

That makes this less a story about Iran. And more a confrontation between two different kinds of power.

The United States controls much of the financial architecture. China controls enormous amounts of physical demand.

Washington can threaten access to dollars. Beijing can say: We still want the oil.

Bitcoin Just Entered the American Mortgage System Without Being Sold.

Bitcoin Is Evolving From an Asset You Buy Into Monetary Property You Can Borrow Against — Without Giving Up the Upside.

This Is the Real Sanctions Test

Sanctions are often discussed as though they’re switches.

Sanction country. Economy collapses. Regime capitulates.

Reality is much messier.

Sanctions work through networks.

Banks. Currencies. Ports. Shipping companies. Insurers. Commodity traders. Technology suppliers. Buyers. Intermediaries. Front companies.

The more of that network Washington can control — or frighten — the more powerful sanctions become. But a sanctioned economy doesn’t need access to the whole world.

It needs enough of the world.

One major buyer.

A few financial channels. Alternative currencies. A shadow shipping network. Third-country intermediaries. Discounted prices. Enough revenue to survive.

China can provide a huge part of that ecosystem.

Which means the relevant question isn’t:

Can America sanction Iran?

Obviously it can.

The question is:

Can America economically isolate Iran if China refuses to participate?

That’s much harder.

We Already Know Washington Can Hurt Iran

This shouldn’t become an argument that sanctions are useless. They’re not. The current pressure is hurting.

Iranian oil shipments to China fell to around 534,000 barrels per day in August, down from 823,000 barrels per day in July and far below peaks earlier this year.

Iran had exported an average of roughly 1.4 million barrels per day in 2025.

A U.S. blockade and new enforcement pressure have significantly reduced the amount of Iranian crude reaching buyers. Iranian floating storage outside the blockade zone has also declined. Chinese refiners are being forced to search for alternatives from countries such as Brazil and Iraq.

So yes: American pressure works.

The United States can raise costs. Reduce volumes. Complicate financing. Destroy discounts. Force rerouting. Make trading with Iran riskier.

But reducing trade is not the same as eliminating it.

And damaging Iran’s economy is not the same as forcing Iran to capitulate.

That difference matters enormously.

China Doesn’t Have to “Save” Iran

This is where the geopolitical framing often becomes too dramatic.

China does not need to become Iran’s military ally. It does not need Chinese aircraft in Iranian skies. It does not need Chinese warships fighting the U.S. Navy. It doesn’t even need to publicly embrace Tehran.

Beijing’s most effective support can be far less theatrical.

Buy oil. Use yuan. Allow intermediaries. Protect selected companies.

Refuse to recognize unilateral American sanctions. Make Washington think twice before targeting major Chinese banks.

That’s enough.

The genius of this strategy is that it forces the United States to choose how far it is willing to escalate economically.

Because sanctioning a small Iranian shipping company is easy.

Sanctioning a minor Chinese refinery is harder.

Sanctioning one of China’s major financial institutions? That’s something else entirely.

Sanctions Become Dangerous When the Target Is Too Big

Imagine Washington sanctions a small bank in a developing economy.

That bank loses access to dollar clearing. Clients flee. The institution becomes economically radioactive. The message spreads. Powerful deterrent.

Now imagine Washington threatens a giant Chinese financial institution.

The calculus changes. China can retaliate.

American companies have major Chinese exposure.

Supply chains cross the Pacific. Treasury markets matter. Technology restrictions matter. Trade flows matter.

China can target American companies.

Read more

Subscribe now

The year 2026 will not be remembered for the dizzying height of the United States national debt, though the $40 trillion mark stands as a monument to fiscal hubris.

It will not be remembered primarily for inflation metrics or the esoteric debates of central bankers in Jackson Hole.

Instead, history will record this era as the moment the world finally understood a brutal, unspoken truth of global economics: fiat currency is a political instrument, and its value is inextricably backed by the barrel of a gun.

When the United States faced geopolitical humiliation in Ukraine, followed by a devastating strategic defeat in Iran, the foundations of the global financial system shuddered.

These failures did more catastrophic damage to the hegemony of the US dollar than a $40 trillion debt load ever could.

Debt can be restructured, monetized, or inflated away. But the loss of global deterrence — the sudden realization that the hegemon is no longer the undisputed master of the geopolitical chessboard — is an irreversible shattering of trust.

To understand the crisis of the US dollar today, we must strip away the sanitized jargon of modern macroeconomics. We must look past yield curves, quantitative easing, and inflation targets. The reality of money is far more primal. You will never find a currency that serves as the global reserve while its issuing nation wields a tiny military or weak geopolitical power. It is an economic and historical impossibility.

The nation whose currency becomes the bedrock of global trade is, without exception, the nation with the greatest military and geopolitical supremacy. The bond market, ultimately, is a marketplace of promises. And human nature dictates a simple rule: people only trust the promises of the most powerful person in the room. When that power wanes, the currency collapses. We are standing at the edge of that collapse, driven to this precipice by a fatal decision made over half a century ago, a decision that traded America’s industrial soul for a few fleeting decades of unbridled consumption.

This is the anatomy of a monetary empire in decline, and a guide to surviving the dust it will leave behind.

Part I: The Sword and the Coin — A History of Hegemony

To grasp the magnitude of the current dollar crisis, we must view money through the lens of history. Economists often treat money as a neutral medium of exchange, a convenient alternative to bartering. This is a fairy tale. International money is, and always has been, a projection of state power.

Look at the lineage of global reserve currencies. The pattern is unbroken, undeniable, and written in blood and iron.

The Spanish Empire and the Silver Real

In the 16th century, the global reserve currency was the Spanish real de a ocho (the piece of eight).

Why did the world trade in Spanish silver?

It was not simply because Spain had access to the massive silver veins of Potosí in the New World. It was because the Spanish galleons, backed by the fearsome Spanish Armada, dominated the Atlantic and Pacific trade routes.

The Spanish infantry — the tercios — were the most feared military units in Europe. The currency was a derivative of the Crown’s ability to extract, transport, and fiercely defend its wealth against any challenger.

When the Spanish military began to falter, culminating in the destruction of the Armada and grueling, unwinnable wars in the Low Countries, the dominance of the Spanish Reale began its terminal decline.

The Dutch Guilder and the Mastery of the Seas

As Spain retreated, the Dutch Republic rose. By the 17th century, the Dutch guilder became the undisputed currency of global trade.

The Netherlands was a tiny geographic nation, but it possessed a sprawling, hyper-efficient naval force and the world’s first multinational corporation: the Dutch East India Company (VOC).

The Dutch built an armada of merchant ships protected by heavily armed warships that could project power from Amsterdam to Batavia (modern-day Jakarta).

The world trusted the guilder because the Bank of Amsterdam was backed by a geopolitical juggernaut that monopolized the spice trade and crushed naval rivals. When the Dutch military was eventually ground down by successive wars with France and England, their currency lost its luster, fleeing to the new center of power.

The British Pound and the Empire on Which the Sun Never Sets

The 18th and 19th centuries belonged to Great Britain and the Pound Sterling. The British did not achieve financial supremacy purely through the ingenuity of the Bank of England, though financial innovation played a role.

The Pound was the global reserve because it was underwritten by the Royal Navy. The British doctrine dictated that their navy must be larger and more powerful than the next two largest navies combined. They controlled the chokepoints of global trade: Gibraltar, Suez, Singapore, and the Cape of Good Hope.

A merchant in Qing Dynasty China, a cotton trader in Egypt, and a railroad baron in Argentina all transacted in Pound Sterling because the British military guaranteed the security of the global commons.

Contracts settled in London were enforced by the implicit threat of a gunboat appearing on your horizon if you defaulted. But two World Wars destroyed Britain’s ability to maintain its massive military footprint.

As the British Empire militarily contracted, the Pound Sterling was unceremoniously dumped from its global reserve pedestal.

The Ascendancy of the US Dollar

The United States did not politely ask the world to adopt the dollar in 1944 at Bretton Woods. The dollar became the global standard because the United States emerged from World War II with half of the world’s manufacturing capacity, the sole possession of the atomic bomb, and a navy that had just annihilated the Japanese fleet and secured every major ocean.

The dollar was pegged to gold, yes, but more importantly, it was pegged to the United States military. When the petrodollar system was established in the 1970s, the agreement with Saudi Arabia was explicit: the Saudis would price their oil strictly in US dollars, and in exchange, the US military would guarantee the security of the House of Saud.

Fiat is not “backed by nothing.” It is backed by aircraft carriers, nuclear submarines, stealth bombers, and the political will to use them.

The $40K Bitcoin Crowd Has a Problem: What Price Finally Makes You Admit the Bottom Is Gone?

Bitcoin is above $80K, yet thousands are still waiting for a price that may never return. The real danger isn’t missing the bottom — it’s refusing to change your mind when the market already has.

Part II: The Oracle of Wall Street

You do not need to be a fringe economic theorist or a cynical geopolitical analyst to understand this dynamic. You only need to listen to the people who operate the plumbing of the global financial system.

Consider Jamie Dimon, the CEO of JPMorgan Chase, arguably the most powerful and well-connected banker on the planet. Dimon sits at the nexus of global capital, government policy, and multinational corporate strategy. He understands the mechanics of money better than 99.9% of the human population.

Dimon recently made a statement that cut through the usual corporate platitudes with terrifying clarity:

“If the US military does not remain the most powerful military in the world, the US dollar will lose its global reserve currency status within 25 years.”

Why would a Wall Street banker care about the lethality of the US Armed Forces? Because Dimon knows that the trillions of dollars his bank moves every day are ultimately secured by American hegemony.

When the US Treasury issues a bond, it is making a promise to pay the bearer back with interest over 10, 20, or 30 years. But why does a foreign central bank, a sovereign wealth fund, or a multinational corporation believe that promise?

They believe it because the United States has the power to tax the most productive economy on earth, and more importantly, the power to enforce its will globally. The US can sanction adversaries, seize assets, freeze central bank reserves, and ensure that the global trade of vital commodities (like oil and copper) happens on its terms. The bond market, the bedrock of the fiat system, is entirely reliant on the perception of unshakeable American power.

If that power is challenged and defeated, the promises become suspect. If a local bully is suddenly beaten up in the town square by a rival, the townspeople no longer pay him protection money. In the global geopolitical arena, the same psychology applies. The defeats in Ukraine and Iran were not just tactical military losses; they were a catastrophic piercing of the veil of American invincibility. And when the invincibility vanishes, the currency that relies on it begins to burn.

Part III: The Original Sin of 1971

To understand why the US military failed to assert its dominance in 2026, we must go back to the fatal pivot point in American economic history: the year 1971.

Read more

Subscribe now

Nvidia reports earnings tomorrow.

And Wall Street has a strange problem.

Almost nobody seems worried that Nvidia will have a bad quarter. The problem is almost the opposite.

Everyone expects Nvidia to crush it. Again.

Analysts are looking for quarterly revenue of roughly $92 billion, compared with Nvidia’s own guidance of $91 billion, plus or minus 2%.

Just three months ago, the company reported $81.6 billion in revenue — up an extraordinary 85% year over year — with Data Center revenue exploding 92% to $75.2 billion.

Those numbers would be transformational for almost any company on Earth. For Nvidia? They are becoming normal. And that may be the company’s biggest problem.

Not competition. Not China. Not AMD. Not custom chips. Not even the growing questions around the hundreds of billions being poured into AI infrastructure.

Expectations.

Nvidia has become so successful that investors no longer ask whether it will beat Wall Street estimates. They ask: By how much? And increasingly: Will even that be enough?

Bitcoin’s Rally Is Becoming a BlackRock Market: 70% of ETF Flows Came Through IBIT.

Bitcoin remains decentralized at the protocol layer — but as IBIT absorbs nearly 70% of ETF inflows, Wall Street’s gateway to Bitcoin is becoming strikingly concentrated.

The Curse of Spectacular Execution

This is what happens when a company repeatedly does the extraordinary. The extraordinary becomes expected. Nvidia’s first-quarter fiscal 2027 numbers were ridiculous.

• Revenue: $81.6 billion. Up 85% year over year.

• Data Center: $75.2 billion. Up 92%.

• Data Center networking: $14.8 billion. Up 199%.

• Gross margin: About 75%.

Nvidia then guided toward approximately $91 billion in second-quarter revenue without assuming any Data Center compute sales from China.

Think about the psychology this creates.

A company grows revenue 20% sequentially from an already enormous base. The market doesn’t say: “Incredible.” It says: “What’s next?”

That is the curse of spectacular execution. Every beat raises the hurdle for the next quarter. Every guidance increase raises the baseline. Every supply shortage creates expectations for future pricing power. Every hyperscaler capex increase becomes embedded in tomorrow’s revenue estimates. Success compounds. So do expectations.

The Hedonic Treadmill of Financial Markets

To understand what is happening to Nvidia stock, you have to understand the hedonic treadmill of financial markets. Human beings rapidly adapt to positive changes in their environment, returning to a baseline level of happiness. Financial markets do the same thing with corporate earnings. When a company repeatedly shatters records, those records stop being pleasant surprises; they become the new required baseline just to keep the stock price flat.

We are watching a real-time psychological experiment on Wall Street. The options market tells the story. Ahead of Nvidia’s August 26 earnings report, options imply roughly a 5.4% move in the stock following the results. That sounds large. For Nvidia, it isn’t. Before the May report, options implied about 6.5%. Nvidia’s historical average post-earnings move has been closer to 7.4%.

In other words, the market appears to be pricing Nvidia’s results as becoming more predictable, not less.

That’s fascinating. Because Nvidia is sitting at the center of what may be the largest infrastructure buildout in modern technology. AI spending is exploding. Data centers are consuming unprecedented capital. Memory costs are surging. Financing structures are becoming increasingly complex. Governments are treating compute like strategic infrastructure.

Yet investors are increasingly saying: “Yeah. Nvidia will probably beat.”

That confidence is itself a risk. When everyone expects surprise, surprise becomes almost impossible. Reddit understands the problem better than some institutional desks. In the current WallStreetBets earnings discussion, one question keeps appearing in different forms:

Should you buy Nvidia puts even if you expect an earnings beat because the stock could sell off anyway?

One commenter summarized the mood perfectly: Nvidia may beat and still fall.

That isn’t sophisticated discounted-cash-flow modeling. But it captures the problem perfectly. The market has stopped asking whether Nvidia’s business is good. That debate ended a long time ago. The question is whether Nvidia’s business can be better than the future already embedded in Nvidia’s stock price. Those are completely different questions.

Great Company. Difficult Stock.

Investors often confuse the two. A great company is a business question. A great stock is a price question.

Read more

Subscribe now

America’s debt problem has become too large to understand in trillions.

$30 trillion. $35 trillion. $40 trillion.

After a while, the numbers stop meaning anything. They become political wallpaper.

So here is a much simpler number.

21 cents.

Through the first ten months of fiscal year 2026, the U.S. federal government recorded roughly $931.4 billion in net interest costs against $4.485 trillion in federal receipts.

That’s approximately 20.8 cents of interest for every dollar Washington collected.

For comparison, the ratio was just 9.7 cents in fiscal year 2022.

In four years, the burden has roughly doubled.

Technically, “tax dollar” is shorthand here — the denominator is total federal receipts, which include taxes plus some other government revenue.

But the economic message doesn’t change.

Before Washington pays a soldier...

Before it builds a highway...

Before it funds scientific research...

Before it sends Medicare reimbursement...

Before Congress debates a new program...

A rapidly growing share of federal revenue is already spoken for. Not for something America is building today.

For something America already bought yesterday. Debt.

And unlike almost every other major line in the federal budget, interest does not care what voters want.

The Most Expensive Federal Program Nobody Voted For

Americans argue constantly about government spending.

• Defense.

• Medicare.

• Social Security.

• Infrastructure.

• Education.

• Border security.

• Industrial subsidies.

• Foreign aid.

• Climate programs.

• Tax credits.

Every dollar becomes ideological territory.

Interest is different. There is no constituency demanding better interest payments.

No ribbon-cutting ceremony. No new bridge. No aircraft carrier. No hospital. No research laboratory. No pension check.

Nothing tangible appears.

Washington simply transfers money to the holders of Treasury securities because the bills have come due.

And the bill is becoming enormous.

The Congressional Budget Office expects net federal interest costs to exceed $1 trillion in fiscal year 2026.

That’s not some distant 2050 projection. That’s now!

And CBO expects net interest spending to roughly double again to $2.1 trillion by 2036 under current-law projections.

By then, interest would consume about 4.6% of the entire U.S. economy.

CBO projects it would nearly equal all federal discretionary spending combined.

Think about that. One line item. Created largely by decisions already made.

Potentially costing almost as much as everything Congress annually appropriates for defense and domestic discretionary programs combined.

That is not just a debt problem.

It is a freedom-of-action problem.

Interest Is Different From Every Other Expense

If the defense budget becomes too large, Congress can theoretically cut weapons programs.

If a subsidy becomes too expensive, lawmakers can eliminate it. If an agency grows too large, budgets can be reduced. Politically difficult? Absolutely. Mathematically possible? Yes.

Interest is not like that.

Congress cannot hold a hearing and announce:

“We’ve decided Treasury bondholders will receive 25% less this year.”

Not without triggering a financial event that would make the savings rather counterproductive.

Interest expense is the accumulated consequence of two things: How much debt exists.

And: What interest rate the government pays on that debt.

Neither changes quickly.

That makes interest one of the stickiest parts of the budget. Once it grows, reversing it takes time.

How America Got Here

The story isn’t simply: “Politicians spent too much.”

That is true in the same way saying “planes fly because of engines” is true. Correct. Incomplete.

The deeper story involves decades of deficits interacting with one enormous monetary shift.

For years, debt was cheap. Extremely cheap.

Following the global financial crisis, interest rates fell toward historic lows. Then the pandemic pushed rates even lower.

The U.S. could borrow trillions of dollars at extraordinarily attractive financing costs.

This created a seductive illusion. Debt was rising rapidly. But debt service remained manageable.

The national credit card balance kept growing. The interest rate stayed close to zero. That made aggressive borrowing feel almost free. Until it wasn’t.

The Era of Free Money Ended

Inflation changed everything.

The Federal Reserve raised interest rates dramatically after the inflation surge of the early 2020s.

Treasury yields followed.

Suddenly Washington wasn’t refinancing debt at pandemic-era rates. It was refinancing at much higher ones.

This is important because federal debt doesn’t reprice all at once.

Treasury securities mature gradually. A bond issued years ago might still carry a low coupon. But eventually it matures. Then the Treasury must repay it.

If the government doesn’t have the cash—and it usually doesn’t because the budget remains deeply in deficit—it issues another security.

The old low-rate debt disappears. New higher-rate debt takes its place.

This process happens quietly.

Month after month. Auction after auction. Billions at a time.

The budget experiences it with a delay.

That’s why today’s interest burden isn’t simply the result of today’s interest rates.

It is the accumulated effect of yesterday’s debt gradually colliding with today’s rates.

Washington Has a Refinancing Problem

Imagine someone bought a house with a gigantic adjustable mortgage.

For years, the rate was 2%. Life was comfortable.

Then the rate begins resetting toward 4%. Then 5%. The mortgage balance didn’t necessarily increase. The financing cost did.

Now apply that logic to tens of trillions of dollars.

This is essentially what the federal government is experiencing.

Except Washington has another problem.

It keeps borrowing more at the same time.

So the process looks like this:

1. Old debt matures.

2. It gets refinanced at higher rates.

3. New deficits require additional borrowing.

4. That borrowing also carries higher rates.

5. Interest expense rises.

The higher interest expense increases the deficit. The larger deficit requires more borrowing. That additional borrowing creates more interest expense.

This doesn’t automatically become a crisis. But it does create a feedback loop.

Debt creates interest → Interest creates deficits → Deficits create debt.

That is the arithmetic Washington increasingly has to fight.

The 21-Cent Number Is Rising Faster Than Revenue

The most uncomfortable part of the latest data isn’t simply the absolute level. It’s the direction.

Through ten months of FY2026, federal receipts were up about 3.2% compared with the same period a year earlier. Net interest costs were up roughly 10.8%.

Revenue is growing. That’s good.

But the interest bill is growing more than three times faster. That’s the problem.

Because fiscal sustainability isn’t determined simply by whether government revenue rises. It depends on whether revenue can keep pace with the promises attached to it.

Right now, interest is winning that race.

In 2022, This Looked Very Different

The speed of the transformation is remarkable.

The interest-to-receipts ratio was approximately 9.7% in FY2022.

Then:

• 14.9% in FY2023.

• 17.9% in FY2024.

• 18.5% in FY2025.

And through the first ten months of FY2026: 20.8%.

That progression matters more than another screenshot of the national debt clock. The debt clock tells you the balance.

The interest ratio tells you how much fiscal oxygen that balance is consuming. And fiscal oxygen is finite.

Payjoin Just Hit 1.0: Bitcoin Is Quietly Attacking the Surveillance Assumptions Chain Analysis Depends On.

Payjoin Doesn’t Hide Bitcoin’s Blockchain. It Makes the Surveillance Models Built on Top of It Less Trustworthy.

Every Interest Dollar Has an Opportunity Cost

Here’s where the abstract numbers become real.

Suppose the federal government spends an additional $100 billion on interest. That $100 billion has to come from somewhere.

Washington can:

• Collect more taxes.

• Cut spending somewhere else.

• Borrow another $100 billion.

• Or some combination.

Borrowing is politically easiest. But borrowing the interest expense simply pushes part of the problem forward.

The government doesn’t escape the cost. It capitalizes it. Debt gets added to debt. That is why high interest costs gradually reduce policy flexibility.

Every future president inherits less room to maneuver. Every future Congress begins negotiations with more revenue already committed.

Conservatives Should Care About This

Because interest creates pressure for higher taxes.

A government cannot permanently spend far more than it collects without expanding debt. Eventually political pressure builds to raise revenue.

The larger the interest burden becomes, the harder it becomes to balance the budget using spending restraint alone.

That should worry anyone whose priority is keeping taxes low.

Progressives Should Care About This

Because interest competes with public programs.

Money transferred to bondholders cannot simultaneously fund healthcare, education, housing, infrastructure, or income support.

As debt service grows, every new social spending proposal enters a more constrained fiscal environment.

That should worry anyone who wants a larger public investment role.

Bitcoin Was Too Scary at $64K. Now Everyone Wants It Near $78K.

Bitcoin Didn’t Become 20% Safer in Four Days — But Once the Price Jumped From $64K to Nearly $78K, Fear Turned Into FOMO and the Crowd Came Running.

Defense Hawks Should Care Too

Because military power requires fiscal capacity.

Wars are expensive. Ships are expensive. Missiles are expensive. Nuclear modernization is expensive. Artificial intelligence infrastructure is expensive.

A country whose interest bill consumes an ever-larger fraction of revenue has less flexibility when the next geopolitical emergency arrives.

Financial capacity is national-security capacity. History has repeatedly demonstrated this.

Great powers rarely fail because they literally run out of weapons.

Sometimes they run out of affordable options.

And Investors Should Definitely Care

Treasury securities sit at the center of global finance.

They’re used as reserves.

• Collateral.

• Safe assets.

• Benchmarks for virtually every other interest rate.

• Mortgages price off Treasury markets.

• Corporate bonds price off Treasury markets.

• Equity valuations depend partly on Treasury yields.

• Banks hold Treasuries.

• Governments hold Treasuries.

• Pension funds hold Treasuries.

• Insurance companies hold Treasuries.

The entire financial system assumes the U.S. government can issue vast quantities of debt smoothly.

That assumption remains extraordinarily strong.

But debt sustainability doesn’t have to produce a default before markets care.

It can show up as:

• Higher term premiums.

• Higher long-term yields.

• Greater auction volatility.

• Pressure on the dollar.

• Reduced foreign demand.

• Higher private-sector borrowing costs.

• Or simply investors demanding more compensation to finance Washington.

Any of those outcomes makes the interest problem worse.

The Cruel Irony of Higher Treasury Yields

Investors often celebrate higher Treasury yields.

More income. More attractive bonds. Great. But every extra percentage point eventually filters into Washington’s financing costs.

At current debt levels, even small changes matter enormously.

The federal government has more than $32 trillion of debt held by the public, according to the latest figures underlying the current interest-ratio analysis.

Not all of it refinances immediately. But over time, rates compound through the debt stock.

A 1 percentage-point change in the average financing rate across roughly $32 trillion would represent more than $300 billion per year in additional annual interest expense once fully reflected.

That’s approximately the scale of entire federal departments. Generated by a single percentage point.

This is why the bond market matters.

Bitcoin’s New Macro Driver Isn’t the Fed — It’s the U.S. Treasury’s Debt Machine.

Everyone watches Kevin Warsh. But the institution increasingly capable of moving the liquidity conditions that matter to Bitcoin may be sitting a few blocks away—the U.S. Treasury.

The Fed Now Has an Awkward Passenger

The Federal Reserve’s mandate is not to make Treasury financing cheap. It is supposed to pursue price stability and maximum employment.

That separation is essential.

But fiscal arithmetic doesn’t disappear just because central-bank independence exists.

When inflation is too high, the Fed may need to keep rates high.

Higher rates increase Treasury financing costs. Higher financing costs increase federal deficits.

Larger deficits require more Treasury issuance. More issuance can put upward pressure on yields.

That creates a difficult interaction.

The Fed fights inflation. The Treasury absorbs the consequences.

This doesn’t mean the Fed should cut rates to save the budget.

Doing that recklessly could create an even larger problem.

If investors conclude monetary policy is being subordinated to government financing needs, inflation expectations and bond-market credibility could deteriorate.

That is how a fiscal problem can become a monetary problem.

“Just Cut Rates” Is Not a Solution

This is one of the most tempting arguments.

If interest rates are causing the problem, lower interest rates. Simple. Except rates can’t sustainably be ordered lower by political decree.

Markets set Treasury yields based on inflation expectations, economic growth, supply and demand for bonds, Federal Reserve policy, and perceived fiscal risk.

The Fed can influence the short end dramatically. It can influence longer rates indirectly. But forcing yields artificially low while inflation remains elevated carries a price.

Usually somewhere else.

• Currency weakness.

• Inflation.

• Asset bubbles.

• Financial repression.

• Loss of credibility.

There is no free interest-rate button.

“Just Inflate the Debt Away” Isn’t Free Either

Inflation can reduce the real value of nominal debt.

Historically, governments have sometimes benefited from exactly that.

If prices and nominal incomes rise rapidly while previously issued debt remains fixed in dollars, the debt burden can decline relative to the economy.

Sounds convenient. Bondholders understand this.

If they expect persistent inflation, they demand higher yields. New debt then becomes more expensive.

Inflation also increases living costs, distorts investment, and can force the central bank into tighter policy.

So inflation can reduce yesterday’s debt burden while increasing tomorrow’s financing cost.

Again: No magic button.

“Just Tax the Rich” Isn’t Enough by Itself

Higher taxes on high earners and corporations can certainly raise revenue.

How much depends on the policy design, economic responses, and enforcement.

But the fiscal gap is becoming large enough that serious solutions eventually confront arithmetic rather than slogans.

CBO projects a $1.9 trillion federal deficit in 2026 despite an economy that is not experiencing a classic deep recession.

That matters.

Large deficits are easier to explain during wars, recessions, or pandemics. They’re much harder to stabilize when they become structural.

A permanent gap of that magnitude probably cannot be solved painlessly by targeting one narrow group or one narrow spending category.

“Just Cut Waste” Isn’t Enough Either

Every government contains inefficiency.

• Fraud exists.

• Waste exists.

• Bad programs exist.

Finding and eliminating them is worthwhile. But trillions aren’t hiding behind office coffee machines.

The biggest federal expenditures are large because they represent large political commitments.

• Social Security.

• Medicare.

• Defense.

• Healthcare.

• Debt service.

When deficits approach $2 trillion, eventually the conversation reaches things voters actually care about.

That’s why fiscal reform is difficult. The arithmetic is simple. The politics are brutal.

Growth Is the Best Solution — and the Hardest to Guarantee

There is one relatively painless escape route.

Grow faster.

If the U.S. economy expands rapidly, incomes rise.

• Tax receipts increase.

• Debt becomes smaller relative to GDP.

• Businesses earn more.

• Households earn more.

• Fiscal pressure becomes easier to manage.

This is why AI productivity matters far beyond Nvidia’s stock price.

If artificial intelligence produces a genuine productivity boom—raising output, wages, corporate profits and taxable income—it could materially improve America’s debt arithmetic.

This may be one of the strongest macroeconomic bull cases for AI.

Not another chatbot. Not another agent demo.

Growth.

But hoping productivity arrives is not fiscal policy.

The gains must be real. Large. Broad. And sustained.

Building trillion-dollar data centers doesn’t automatically solve a trillion-dollar interest problem. Productivity has to reach the rest of the economy.

CBO’s Baseline Is Already Uncomfortable

The Congressional Budget Office does not project an imminent American default.

That distinction is important.

The United States remains the issuer of the world’s dominant reserve currency.

Treasury markets remain the deepest government bond markets in the world.

The country has enormous productive capacity. It controls its own currency. This is not a household about to have its credit card declined. But CBO’s long-term trajectory is still striking.

Debt held by the public is projected to equal about 101% of GDP in 2026.

By 2036: 120%.

The previous historical record came immediately after World War II.

Today’s trajectory occurs without a world war requiring comparable national mobilization.

And rising interest costs are one of the main reasons deficits keep expanding.

Interest Is Becoming the Budget’s Fastest-Growing Monster

CBO projects net interest spending rising from around $1 trillion in 2026 to $2.1 trillion in 2036.

That’s roughly doubling within a decade.

To understand how unusual that is, compare interest with programs people actually recognize. CBO says net interest in 2026 already exceeds mandatory spending on every individual federal program except Social Security and Medicare.

America effectively has another trillion-dollar program. Except nobody receives healthcare from it.

Nobody retires on it. Nobody gets a job from it directly.

Its product is: Past borrowing.

This Is What Fiscal Crowding Out Really Looks Like

“Crowding out” sounds like something economists invented specifically to make normal people stop reading.

The idea is simple.

Resources are finite.

If the government absorbs increasingly large amounts of savings by issuing debt, those resources cannot simultaneously finance everything else at the same price.

The effects can appear through higher interest rates.

Reduced private investment. Different portfolio allocations. Or government budgets increasingly dominated by debt service.

There is also political crowding out.

Congress may want:

• More defense.

• More infrastructure.

• More AI research.

• More healthcare.

• More tax cuts.

• More climate investment.

• More border spending.

Then Treasury sends the interest bill. Suddenly there isn’t as much room. That is fiscal crowding out before a single private investment is displaced.

The Dangerous Thing About Interest Is That Nobody Feels It Immediately

This helps explain why governments tolerate rising debt.

• If Congress raises taxes by $100 billion, people notice.

• If Congress cuts benefits by $100 billion, people notice.

• If Congress borrows $100 billion, almost nobody notices.

The pain gets delayed.

Borrowing separates political benefit from political cost.

Spend today. Pay later. And later has arrived.

Today’s trillion-dollar interest bill is the invoice for years when borrowing felt painless.

Today’s voters are paying for yesterday’s political choices.

Tomorrow’s voters will pay for today’s.

That’s how compounding works in public finance.

America Has Something Most Countries Don’t

There is an important reason not to become hysterical.

The dollar.

The United States issues debt in its own currency. Global central banks hold Treasuries. Financial institutions need Treasury collateral. International trade still relies heavily on dollars.

In moments of crisis, investors often run toward U.S. government debt rather than away from it.

That is an extraordinary advantage.

It means America has far more fiscal room than a typical country borrowing in a currency it cannot create.

But reserve-currency status is an advantage. Not immunity from arithmetic.

The dollar can make the runway longer. It cannot make interest disappear.

The Real Risk Is Constraint, Not Bankruptcy

This is where many debt discussions go wrong.

One side says: “America is going bankrupt tomorrow.”

Highly misleading.

The other says: “The U.S. prints dollars, so debt doesn’t matter.”

Also misleading.

The more realistic risk is boring. And therefore more dangerous. America gradually loses options.

• Want a tax cut? Harder.

• Want a major new social program? Harder.

• Want to fight a war? More expensive.

• Want a stimulus package during a recession? Less fiscal room.

• Want to invest massively in infrastructure? Competing with interest.

• Want to stabilize Social Security? Another difficult tradeoff.

Nothing collapses tomorrow.

The state simply becomes less flexible. Year after year. That’s what 21 cents means.

Bitcoin’s Attention Economy Is Broken.

Bitcoin solved the double-spend problem. Social media may be creating a different one: the same idea, recycled through ten identities, can look like ten independent people agreeing with each other.

And 21 Cents Isn’t the Destination

That’s the crucial point.

Today’s interest burden is not necessarily the peak.

CBO projects deficits remaining large throughout the coming decade.

Debt keeps accumulating. Higher-cost debt continues replacing lower-cost securities as they mature. Interest costs continue climbing.

The mathematics don’t require a crisis for the situation to deteriorate. They only require the status quo.

That may be the most unsettling part.

The Fiscal Debate Is Going to Change

For years, American politics could promise contradictory things.

• Lower taxes.

• Higher spending.

• Better benefits.

• More defense.

• No entitlement reform.

• No serious sacrifice.

• And markets kept financing it.

That doesn’t mean markets are about to stop. But the price of financing has changed.

The interest bill is forcing all those promises onto the same spreadsheet. Eventually politicians will have to choose.

• Higher revenue.

• Lower spending growth.

• Faster economic growth.

• Some inflation.

• Benefit reforms.

• Tax reforms.

• Or continued large borrowing.

Probably some combination.

There is no painless ideology hiding behind the curtain. Just arithmetic.

That’s Why 21 Cents Matters More Than $40 Trillion

The national debt number is enormous. But it doesn’t tell you when the debt becomes restrictive.

Interest does.

A wealthy household can comfortably carry a $2 million mortgage.

A poorer household can struggle with $300,000.

The balance itself isn’t the complete story.

You need to know income and financing costs.

For a government, the equivalent question is:

How much revenue does debt service consume?

In FY2022, the answer was under 10 cents per dollar of receipts.

Today, through ten months of FY2026, it is nearly 21 cents.

That is a much more informative alarm bell than another trillion being added to the debt clock.

Then Comes the Recession Nobody Can Schedule

The current numbers also arrive before the next major recession.

Recessions create an ugly fiscal combination.

Tax revenue weakens. Unemployment-related spending rises. Automatic stabilizers activate. Congress often approves stimulus. Deficits expand. Debt issuance rises.

Historically, interest rates often fall during recessions, partially offsetting the damage.

That’s useful.

But start from a much larger debt stock and the dynamics become more sensitive.

The government enters the next downturn carrying far more debt than it carried into earlier cycles.

That reduces room for error.

What If Inflation Stays Sticky?

Then the problem gets even more complicated. A weakening economy normally gives the Fed space to cut rates.

High inflation limits that ability.

That creates the scenario fiscal policymakers least want:

• Slower growth.

• Weakening revenues.

• Large deficits.

• And interest rates that cannot fall quickly.

Again, this does not guarantee crisis. But it removes one of the automatic stabilizers governments have relied on for decades.

The larger the existing debt stock, the more important that becomes.

Fiscal Capacity Is a Strategic Asset

Countries accumulate weapons. Oil reserves. Gold. Foreign exchange reserves. Strategic minerals. Industrial capacity.

They should think about fiscal capacity similarly.

A low debt burden gives a government the ability to respond when something extraordinary happens.

• Pandemic? Borrow.

• War? Borrow.

• Financial collapse? Borrow.

• Natural disaster? Borrow.

• Recession? Borrow.

The United States used that capacity aggressively during the pandemic. It worked in the narrow sense that Washington could mobilize trillions extremely quickly.

But every emergency drawdown leaves less unused capacity unless the fiscal position improves afterward.

You cannot know what the next crisis will be.

That’s exactly why flexibility has value.

Nobody Knows Where the Breaking Point Is

There is no magical debt-to-GDP number at which a siren sounds.

Japan has demonstrated that a rich country can sustain extraordinarily high government debt for a very long time under particular financial conditions.

Other countries have experienced crises at much lower levels.

Institutions matter. Currency regimes matter. Domestic savings matter. Inflation matters. Economic growth matters. Investor confidence matters. Maturity structure matters.

The United States has exceptional structural advantages.

Which means anyone telling you:

“America collapses when debt reaches exactly X”

is selling certainty that economics cannot provide.

But uncertainty about the threshold is not a reason to accelerate toward it.

The Bond Market Doesn’t Need to Panic

That’s another misconception.

A fiscal crisis doesn’t have to arrive as a single dramatic day. No Black Monday required. Instead, investors might simply demand 20 basis points more. Then another 20. Then foreign central banks diversify a little more.

Pension funds shorten duration. Treasury auctions become slightly softer. The term premium increases. Mortgage rates remain slightly higher. Corporate borrowing remains slightly higher.

Washington refinances slightly more expensively. Every change looks manageable. Until the accumulation itself becomes the story.

Compounding is powerful precisely because nothing dramatic has to happen.

The U.S. Still Has Time

This is the constructive part.

The current numbers are serious. They are not destiny.

America remains enormously wealthy.

Its private sector remains innovative. Its capital markets remain unmatched in scale. Its currency remains dominant.

Productivity growth could improve. AI could increase output.

• Immigration policy can affect labor-force growth.

• Tax policy can change revenue.

• Spending growth can change.

• Entitlement formulas can change.

• Healthcare costs can change.

• Interest rates can change.

None of this is fixed.

The challenge is political. Most fiscal reforms impose visible pain now in exchange for less visible benefits later.

Democracy naturally struggles with that trade.

Borrowing does the opposite.

Visible benefit now. Invisible cost later.

Guess which one politicians prefer.

The Clock Isn’t the Debt Clock Anymore

Watch the interest ratio.

That’s the number that tells you how much room remains.

$931 billion of net interest in ten months. $4.485 trillion of receipts. 20.8%.

Nearly 21 cents of every federal revenue dollar consumed before Washington decides what it wants to accomplish next.

And CBO’s baseline suggests interest costs keep rising from here. That should change how we think about the American debt debate.

The problem isn’t that the United States is suddenly going to wake up one morning and discover it is bankrupt.

The problem is more subtle. More gradual. And arguably more important.

America may wake up one year and realize the future has become expensive because too much of today’s revenue is paying for the past.

The Bill for Free Money Has Arrived

For more than a decade, Washington lived in an extraordinary financial environment.

Rates were low. Capital was abundant. Borrowing was easy.

Every crisis could be answered with more debt. Every political compromise could be financed by pushing costs forward.

That era left behind an enormous balance. Now the balance is being repriced.

The United States is discovering something households and companies already know:

The size of your debt matters less when money is cheap.

And much more when it isn’t.

That’s why the most important fiscal number in Washington today may not be the national debt. Not the deficit. Not even the debt-to-GDP ratio.

It may be this: 21 cents.

The amount of every federal revenue dollar currently disappearing into net interest.

In 2022, it was less than 10.

If the trajectory doesn’t change, the next milestone won’t simply be another trillion dollars of national debt.

It will be another piece of America’s future budget that has already been spent.

Before the debate even begins.

Share Sylvain Saurel’s Newsletter

Subscribe now

For decades, the Strait of Hormuz operated on an assumption so powerful that most investors barely noticed it.

Oil would pass. Tankers would pass. LNG carriers would pass.

And if someone seriously threatened that flow, the United States Navy would make sure the threat didn't last long.

That assumption was never written into the price of a barrel of oil. It didn’t appear on corporate balance sheets.

No line item was called “American naval supremacy.” Yet it quietly supported enormous parts of the global economy.

Today, that assumption is being tested. And the result may matter far beyond Iran.

On August 21, only seven commodity vessels were recorded transiting the Strait of Hormuz, down from 14 the day before. No very large crude carriers and no LNG tankers were among them. Before the U.S.-Iran conflict erupted earlier this year, the strait was one of the world’s busiest energy corridors.

A few days earlier, the disruption was even more dramatic.

Only five commodity vessels crossed on August 15. None crossed on August 16, according to Kpler data cited by Reuters.

Before the war, more than 130 vessels a day could transit the waterway.

The United States has overwhelming naval power in the region.

It has aircraft carriers. Destroyers. Submarines. Long-range aircraft. Surveillance systems. Bases. Missiles. Allies. And the U.S. Fifth Fleet has spent decades describing freedom of navigation through Hormuz as one of its central missions.

Yet commercial traffic has still collapsed.

That is the paradox.

America can dominate the battlefield around Hormuz without necessarily controlling the economics of Hormuz.

And that distinction may define the next era of global power.

The Most Important 33 Kilometers in the Global Economy

Hormuz looks tiny on a map.

At its narrowest point, the strait separating Iran from Oman is roughly 33 kilometers wide.

Economically, it is gigantic.

In the first half of 2025, approximately 20.9 million barrels of oil per day moved through Hormuz.

That represented roughly 20% of global petroleum-liquids consumption and about a quarter of all internationally traded seaborne oil.

More than 20% of global LNG trade also passed through the strait.

This is the physical reality behind the geopolitical drama.

The global economy may be digital. Finance may move at the speed of light. AI models may exist in data centers thousands of kilometers away. But civilization still runs on ships passing through narrow pieces of water.

• Hormuz.

• Malacca.

• Bab el-Mandeb.

• Suez.

• Panama.

These are the arteries of globalization.

And arteries have a strange property.

You do not need to destroy the entire body. You only need to squeeze the right point.

Iran Understands This Perfectly

Iran cannot match the United States ship for ship.

It doesn’t need to.

That is the mistake people make when looking at asymmetric warfare. The weaker power does not need to defeat the stronger military symmetrically.

It needs to make the stronger power’s objective prohibitively expensive.

Washington’s objective isn’t merely to destroy Iranian military assets. It is to keep international commerce flowing safely through Hormuz.

Those are two very different missions.

Iran only needs enough capability to maintain uncertainty.

• Mines.

• Missiles.

• Drones.

• Fast boats.

• Coastal batteries.

• Attacks on tankers.

• Threats against ports.

• Ambiguous incidents.

A tanker does not need a 50% probability of being hit before its operator changes behavior. The probability can be tiny.

Because the potential loss is enormous. And because shipowners are not military commanders.

They are businesses.

Bitcoin Was Too Scary at $64K. Now Everyone Wants It Near $78K.

Bitcoin Didn’t Become 20% Safer in Four Days — But Once the Price Jumped From $64K to Nearly $78K, Fear Turned Into FOMO and the Crowd Came Running.

A Navy Can Say the Strait Is Open. An Insurer Can Say Otherwise.

This may be the most important point in the entire story.

What does “open” actually mean?

If an American destroyer can safely transit Hormuz, the strait may be militarily open. If commercial tankers refuse to transit, it is economically closed.

These are not the same thing.

The distinction becomes critical because shipping decisions are made by a network of private actors.

• Shipowners.

• Charterers.

• Insurers.

• Reinsurers.

• Banks.

• Commodity traders.

• Port operators.

• Crews.

• Governments.

One naval commander can declare a route secure. A hundred corporate risk departments can disagree.

That matters.

War-risk insurance premiums rise. Charter rates rise. Crews demand compensation. Banks become nervous about financing voyages. Companies reroute cargo. Tankers wait offshore. Some vessels turn off their transponders.

Others simply stay away.

The result can resemble a blockade even if no physical barrier stretches across the strait.

This is exactly why current vessel numbers matter.

The military balance overwhelmingly favors Washington.

The commercial behavior of shipping still says:

Risk remains unacceptable.

That Is a Much Bigger Problem for America

The United States has spent decades acting as the principal guarantor of maritime globalization.

This arrangement was one of the foundations of the postwar international economy.

Not because America owned the oceans. It didn’t. But because American naval power made one principle unusually credible:

International sea lanes should remain open.

The U.S. Navy still explicitly describes its presence around Hormuz as supporting maritime security, freedom of navigation, and the free flow of commerce.

That created an extraordinary global public good.

Japan could import energy. Europe could trade with Asia. China could become the world’s factory. Saudi Arabia could export oil.

Container ships could cross oceans without every country needing its own globally dominant navy.

The system wasn’t perfectly neutral. American power always served American interests. But open maritime commerce became one of those interests.

Washington effectively helped police the plumbing of globalization. And much of the world benefited.

Bitcoin’s New Security Crisis: Your Keys Are Safe. Your Identity Isn’t.

Bitcoin solved the problem of securing digital money without a bank. It did not solve the problem of everyone around you knowing that you own it.

The “Global Commons” Was Never Really Free

There is something almost ironic about this system.

Freedom of navigation sounds free. It isn’t.

Someone pays.

Navies cost money. Bases cost money. Carrier groups cost money. Surveillance costs money. Diplomacy costs money. Deterrence costs money.

For decades, the United States accepted an enormous share of that bill.

Why?

Read more

Subscribe now

For most of crypto’s history, the enemy was obvious.

Banks.

Banks were slow. Banks closed on weekends. Banks charged fees to move your own money. Banks froze accounts. Banks decided who could participate. Banks stood between people and their assets. Bitcoin was supposed to route around them.

Crypto exchanges emerged as the gateway to that alternative world.

You deposited dollars. You bought Bitcoin. You withdrew the Bitcoin.

Ideally, the exchange disappeared from the relationship. That was the theory.

Now something strange is happening.

Kraken doesn’t simply want to compete with banks anymore. It is exploring becoming one.

Payward and Kraken co-CEO Dave Ripley recently said the company is looking at becoming a “full bank” in some jurisdictions outside the United States, while expanding across trading, banking, and asset management.

And his explanation was even more revealing.

What does a bank actually do?

• Payments.

• Money movement.

• Lending.

• Yield.

• Custody.

Kraken already does versions of all four.

That should make everyone in finance pay attention.

Because the crypto revolution may be entering a phase nobody expected.

Crypto may not kill the banks.

It may turn crypto exchanges into banks.

The Exchange Was Supposed to Be Temporary

Bitcoin’s original architecture did not require Kraken. Or Coinbase. Or Binance. Bitcoin works peer-to-peer.

One person signs a transaction. Another person receives it. The network validates it. Done.

The exchange became necessary because most people live in the fiat world.

They earn dollars. Euros. Pounds. Yen. And they need somewhere to convert those currencies into Bitcoin.

So exchanges became bridges.

Fiat on one side. Crypto on the other.

But bridges have a tendency to become cities.

Once millions of people were already keeping money on exchanges, the obvious question became:

What else can we sell them?

• Trading.

• Staking.

• Stablecoins.

• Margin.

• Futures.

• Custody.

• Cards.

• Yield.

• Institutional services.

• Tokenized assets.

• Traditional stocks.

Crypto exchanges gradually stopped looking like simple marketplaces. They started looking like financial conglomerates.

Kraken’s banking ambitions are simply the logical conclusion.

Bitcoin’s New Macro Driver Isn’t the Fed — It’s the U.S. Treasury’s Debt Machine.

Everyone watches Kevin Warsh. But the institution increasingly capable of moving the liquidity conditions that matter to Bitcoin may be sitting a few blocks away—the U.S. Treasury.

Kraken Already Looks Surprisingly Bank-Like

The funniest part of the story is that Kraken does not have to imagine banking from scratch.

It already has one foot inside the system.

Kraken Financial holds a Wyoming Special Purpose Depository Institution charter, a banking framework requiring fiat deposits to be backed by at least 100% in unencumbered liquid assets. It is not the same as a conventional fractional-reserve commercial bank, and its deposits are not FDIC-insured.

That distinction matters.

Kraken Financial cannot simply operate like JPMorgan. But the direction is obvious.

Kraken also says it now holds more than 100 active regulatory licenses across more than 30 countries, spanning multiple categories of financial activity.

The company has built derivatives infrastructure. Broker-dealer capabilities. Investment-advisory infrastructure. European licenses. Payments infrastructure. Institutional custody.

And now management is openly discussing full banking licenses elsewhere.

At what point does the label crypto exchange stop describing the company?

Probably sooner than most people think.

The Most Important Sentence Wasn’t “We Want to Be a Bank”

It was Ripley’s definition of banking.

• Payments.

• Money movement.

• Lending.

• Yield.

• Custody.

Strip away the marble buildings and suits, and that is essentially what a modern bank is.

It stores value. Moves value. Lends value. Pays something on stored value. And provides infrastructure around all of it.

Crypto companies have spent years rebuilding these functions with different technology.

• Stablecoins instead of checking balances.

• Wallets instead of bank accounts.

• On-chain transfers instead of correspondent banking.

• Crypto collateral instead of traditional collateral.

• Staking or market-based yield instead of deposit interest.

What Kraken is effectively saying is:

We already built much of the product. Now we may acquire the legal wrapper.

That is a very different story from a crypto company begging to enter traditional finance.

It is a crypto company realizing that traditional finance already resembles the business it has become.

Bitcoin at $1 Million Is “Mathematically Impossible” — Here’s Why That Argument May Be Completely Wrong.

A $20 Trillion Bitcoin Market Cap Does Not Require $20 Trillion of New Money — And That Changes the Entire $1 Million Debate.

Crypto Wanted to Unbundle the Bank

This was one of the great promises of fintech.

Banks were bundles.

Read more

Subscribe now

Everyone is watching AI capital expenditure.

That may be the wrong number.

For months, investors have obsessed over how much Microsoft, Amazon, Meta, Google and the rest of Big Tech are spending on GPUs, servers, data centers and power infrastructure.

The numbers already look absurd.

Hundreds of billions of dollars. Then hundreds of billions more next year.

But those headline capex figures may only show the visible part of the AI spending machine.

Because behind them sits something much larger:

Long-term leases. Purchase commitments. Cloud-capacity contracts. Energy agreements. Infrastructure guarantees.

Recent analysis of major technology companies estimates that these commitments collectively run into roughly $3 trillion — much of it associated with the infrastructure needed to build the AI economy.

And that’s where the AI story becomes much more interesting.

Because Big Tech isn’t merely spending enormous amounts of money today.

It is increasingly promising enormous amounts of tomorrow’s money too.

The $3 Trillion Number Changes the Conversation

Let’s start with an important distinction.

This is not $3 trillion of conventional bank debt suddenly sitting on Big Tech balance sheets.

It isn’t a hidden fraud. It isn’t money that has already been spent. And it isn’t all payable tomorrow.

A large portion consists of future contractual commitments: leases that haven’t started yet, agreements to purchase infrastructure or energy, cloud-capacity arrangements and other obligations extending years—sometimes decades—into the future.

Accounting rules often mean those commitments don’t appear as conventional liabilities until the asset becomes available or the lease begins.

They are disclosed. You just have to look deeper. Sometimes much deeper.

That distinction matters.

But so does the number.

The Financial Times recently reported that major hyperscalers had accumulated close to $1.5 trillion in purchase commitments, alongside roughly another $1.5 trillion of lease obligations identified in industry analysis.

Reuters separately calculated that Microsoft, Meta, Oracle, Amazon and Alphabet alone had committed approximately $1.09 trillion to leases that had not yet commenced, largely for AI data centers.

Suddenly, the familiar discussion about annual capex looks incomplete.

The AI boom isn’t just consuming cash. It is locking companies into future cash requirements.

And those two things are not the same.

Bitcoin’s New Macro Driver Isn’t the Fed — It’s the U.S. Treasury’s Debt Machine.

Everyone watches Kevin Warsh. But the institution increasingly capable of moving the liquidity conditions that matter to Bitcoin may be sitting a few blocks away—the U.S. Treasury.

Capex Tells You What Was Spent

This is the number everyone knows.

A company builds a data center. Buys GPUs. Installs networking equipment. Constructs electrical infrastructure.

That spending appears as capital expenditure.

Investors can see it. Analysts model it. Earnings calls debate it. Wall Street publishes charts showing the explosion.

Simple.

But imagine a company signs a 20-year lease today for a data center that won’t open until 2028.

No rent is being paid yet. The building isn’t operational. The lease may therefore not yet appear in the same way as a current lease liability.

But economically, something important has happened.

The company has made a promise.

Once that facility becomes available, cash must start flowing.

This is how the AI infrastructure boom can be simultaneously visible and partially invisible.

The spending appears gradually. The commitment happens immediately.

That is the crucial difference.

Meta Is a Perfect Example

Meta’s filings show just how quickly the numbers can become enormous.

As of June 30, 2026, Meta disclosed approximately $279 billion in leases that had not yet commenced, largely involving data centers, colocation facilities and network infrastructure.

Those leases are scheduled to begin between 2026 and 2036 and can run for as long as 30 years.

Then Meta added approximately another $68 billion of data-center lease commitments in July, with terms of roughly 18 to 20 years.

That’s before we even reach another category.

Meta also disclosed approximately $349 billion of non-cancelable contractual commitments at the end of June, much of it related to cloud capacity, servers, networking infrastructure and data centers.

Read those numbers again.

Meta isn’t simply spending aggressively on AI this year. It has constructed a financial pipeline stretching well into the next decade.

And in some cases beyond it.

That may turn out to be brilliant.

If AI demand expands as expected, Meta will have secured infrastructure that competitors desperately need.

But it also changes the nature of the bet.

This isn’t experimenting with AI anymore.

This is committing the future corporation to AI.

Google Is Crossing the Same Bridge

Alphabet offers an equally fascinating example.

For the first six months of 2026, Google spent approximately $80.6 billion in capital expenditures, more than double the comparable figure from the previous year.

Its full-year capex expectations have since climbed to roughly $195 billion to $205 billion as AI infrastructure spending accelerates.

That number alone is breathtaking.

But again, it’s not the whole story.

Read more

Subscribe now

We said, almost everywhere on trading floors and in macroeconomic circles, that the market would start testing Kevin Warsh on long-term US rates. It was the general expectation: a showdown between the “bond vigilantes” and monetary authorities. But ultimately, it was Scott Bessent, from the US Treasury, who decided to step into the arena and come to the rescue.

After an intervention that can already be described as historic alongside Japan to defend the yen, the US Treasury made a landmark decision: intervening directly and surgically in the bond market. Bessent seems increasingly unwilling to remain a mere spectator in the face of certain financial movements that he clearly deems problematic for the country’s economic and strategic stability.

But before the financial ecosystem gets overly excited and headlines scream about the return of the money printer, let me clarify one thing, which is fundamental to understanding what comes next: the United States has not just launched a disguised Quantitative Easing (QE).

What we are witnessing is much more subtle, much more tactical, and in a way, much more revealing of the deep fears currently driving Washington. Bessent announced that he would double the size of certain buyback operations on 10- to 30-year debt. Specifically, this goes from 2 to at least 4 billion dollars per operation between September 9 and November 4. The end date strangely coincides with the midterm election period... How cheeky, some might say. But in macroeconomics, there are no coincidences, only well-managed calendars.

In this extensive analysis, I will dissect the mechanics of this intervention, the “Soros/Druckenmiller” investment philosophy behind it, the systemic dangers of such yield curve manipulation, and above all, why all of this is intimately linked to the global race for Artificial Intelligence and the inevitable devaluation of fiat currency.

Bitcoin’s Capitulation Dashboard Is Flashing 8 Out of 12 Red — Is This What a Bottom Looks Like?

VanEck Says Eight of Twelve Capitulation Signals Are Flashing as Long-Term Holders Dump 356,000 BTC — Exactly the Kind of Pain That Has Historically Preceded Bitcoin’s Best Opportunities.

Part 1: Anatomy of an Intervention – It’s Not QE, It’s Plumbing

To understand the magnitude of the signal sent by Bessent, we must first understand the technical mechanics of what was announced. The Treasury is absolutely not creating money. It does not have the power to do so. It cannot print reserves out of thin air to buy bonds, as the Federal Reserve (Fed) would during a traditional QE program.

Here, the dynamic is purely balance-sheet-related on the Treasury’s side. They are buying back older securities, known in the jargon as off-the-run Treasuries. Why is this important? In the bond market, recently issued securities (on-the-run) are extremely liquid. Everyone trades them. However, securities issued months or years ago (off-the-run) end up piling up on balance sheets and lose liquidity. When the market starts to panic, or rates rise violently, this illiquidity turns into a risk premium, driving yields up in a disorderly fashion.

Bessent is intervening exactly where the plumbing is clogged. By doubling buyback operations on these illiquid 10- to 30-year debts, while continuing to finance massive American deficits through new issuances on the other side, the Treasury executes a neutral operation in terms of overall debt. American debt doesn’t magically disappear. They are simply exchanging one form of debt for another, smoothing out the market by improving liquidity on specific portions of the yield curve.

It’s a plumber’s intervention, not a central banker’s. But the psychological impact is devastating for short sellers.

The Psychological Shock and the Market’s Reaction

The market’s reaction was immediate. The 30-year Treasury yield, which threatened to skyrocket, fell back to around 5.19%, a spectacular drop of nearly 15 basis points below its recent peak. Better yet (for the Treasury), this drop in long-term rates spread like wildfire to other maturities, easing the entire curve, while equity markets, relieved to see the pressure on rates relax, rebounded violently.

Let’s be clear: the global bond market is worth tens of trillions of dollars. It wasn’t the paltry 14 billion dollars (in total) of announced buybacks that mathematically bent the curve. The market didn’t buy the volume; it bought the signal.

And this signal is clear, powerful, and interventionist: when rising long-term rates become a systemic problem, the US Treasury is ready and willing to alter marginal market flows to prevent the bond sell-off from becoming self-sustaining. This is the very essence of what can now be called the “Bessent Put.”

Financial markets, addicted to state interventionism since 2008, love safety nets. Knowing that an actor with the firepower of the US government is ready to step in to “smooth” the yield curve reassures institutional investors. However, even if we can speak of a “Bessent put,” it is highly likely that such attempts will ultimately fail in the long run given the underlying macroeconomic fundamentals. You only need to look at the USD/JPY (Dollar/Yen) pair’s evolution since last month’s intervention to be convinced. Market forces always end up overriding short-term manipulations.

Part 2: The Timeline of a “Sniper” Strategy

To appreciate the full finesse—and cynicism—of the current situation, we must zoom out and look at the timeline of recent events. Nothing is isolated. Everything forms a relentless logic that reveals the US Treasury’s new doctrine.

Read more

Subscribe now

The history of money is a chronicle of abstractions. We began with physical commodities—cattle, grain, shells, and precious metals—where the value of the money was intrinsic to the object itself. Then came paper receipts for those metals, an abstraction that introduced the concept of trust in a central depository.

In 1971, the world severed the tie to physical commodities entirely, entering the era of pure fiat currency, where the money was backed by nothing more than the economic output and military might of the issuing nation.

Today, we are standing on the precipice of the next great abstraction.

The United States dollar, the undisputed hegemon of the global financial system, is undergoing a metamorphosis. It is migrating from the closed, proprietary ledgers of commercial banks and the Federal Reserve onto open, public cryptographic networks. On the surface, this appears to be a victory for American financial dominance—a technological upgrade that will entrench the dollar’s supremacy for another century.

However, beneath the surface of this transition lies a profound paradox. By tokenizing the dollar and placing it on public blockchains, the architects of the modern financial system are inadvertently stripping the dollar of its foundational power: its monopoly on the rails of settlement.

The dollar’s final form may be brutally simple: a system whose primary purpose is maintaining demand for itself. But in achieving this final, self-perpetuating form, the dollar is quietly setting the stage for its own obsolescence. Eventually, the dollar will cease to be the road upon which global commerce travels. It will simply be the mile marker.

To understand how this transformation will unfold, we must dissect the mechanics of the new digital dollar loop, the technological shifts rendering traditional financial plumbing obsolete, and the ultimate destiny of a currency that dominates accounting but has lost control of settlement.

Bitcoin Is About to Fork Again: The eCash Experiment Begins at Block 964,000.

At Block 964,000, Paul Sztorc’s eCash Fork Will Copy Bitcoin’s History, Activate Drivechains, and Put a Decade-Old Idea to Its First Real Market Test.

Part I: The Brutal Simplicity of the Modern Dollar Loop

To grasp the future of the dollar, one must first recognize what it has become in the present. The United States dollar is no longer just a medium of exchange; it is a sprawling, self-feeding organism. As the US national debt spirals to unprecedented heights, the primary existential imperative of the American financial system is to ensure a constant, insatiable, and global demand for US Treasuries (government debt).

This is where the genius—and the brutal simplicity—of the tokenized dollar comes into play.

Over the past decade, a new financial instrument has emerged: the stablecoin. Stablecoins are cryptographic tokens issued on public blockchains (like Ethereum, Solana, or Tron) that are pegged 1:1 to the value of a fiat currency, overwhelmingly the US dollar. Companies like Tether (USDT) and Circle (USDC) have created hundreds of billions of dollars worth of these tokens, which circulate globally at the speed of the internet.

But the real magic happens behind the scenes.

The Infinite Financial Ouroboros

Stablecoins are not created out of thin air. To maintain their peg to the dollar, they must be backed by reserve assets.

Stablecoins need dollar reserves. When a user anywhere in the world—a merchant in Nigeria, a trader in Singapore, or a corporate treasurer in London—purchases a dollar stablecoin, they are exchanging their local capital for a digital representation of a dollar. The stablecoin issuer takes that capital and places it in a reserve to back the newly minted token.

Dollar reserves buy Treasuries. To generate a yield on these massive reserves and ensure liquidity, stablecoin issuers do not simply leave billions of physical dollar bills in a vault. They purchase US government debt. Tether and Circle are currently among the largest holders of short-term US Treasuries in the world, rivaling the holdings of sovereign nations. They are highly motivated, price-insensitive buyers of American debt.

Treasuries finance the system. By purchasing these Treasuries, stablecoin issuers provide critical funding to the US government, helping to finance its deficits and keeping the machinery of the state running. They suppress yields by maintaining high demand for the debt, which in turn allows the US to continue borrowing at favorable rates.

The system expands dollar liquidity. As the US government spends the money raised from selling debt, that liquidity washes through the global economy. Concurrently, the stablecoins circulating on blockchains provide frictionless, 24/7 dollar liquidity to the global crypto and digital asset markets.

That liquidity creates more demand for dollar-denominated assets. With more digital dollars flowing through the global financial plumbing, market participants use them to trade, lend, and borrow against other assets. This expands the use cases for the digital dollar, bringing more participants into the ecosystem, which in turn requires the minting of more stablecoins.

The loop feeds itself in an endless cycle:

Debt → collateral → stablecoins → dollar demand → more debt.

From the perspective of Washington, D.C., this system appears to be the holy grail. At a time when traditional foreign buyers of US debt (like China and Japan) are slowing their purchases, the crypto economy has stepped in as an insatiable buyer of last resort. The dollar’s dominance seems more entrenched than ever, exported digitally to every smartphone on the planet.

But this is an illusion.

Part II: The Quiet Removal of the Settlement Rails

While the world watches the explosive growth of stablecoins and marvels at the digital proliferation of the dollar, a critical component is quietly being removed from the equation: the dollar’s own settlement rails.

For the past seventy years, the power of the US dollar has not just been in its role as a unit of account. Its true power lay in the physical and digital infrastructure required to move it.

The Old World Plumbing

Read more

Subscribe now

For most of the AI boom, the formula was almost embarrassingly simple.

• Beat expectations.

• Raise guidance.

• Mention artificial intelligence.

• Watch the stock rise.

• Revenue growth was rewarded.

• Data-center demand was rewarded.

• Higher AI spending was rewarded.

• Bigger GPU orders were rewarded.

Even gigantic capital expenditure plans could send shares higher because investors interpreted every new billion spent on AI as evidence that the revolution was accelerating.

That relationship is beginning to break.

And that may be one of the most important signals in the stock market right now.

Because AI companies have developed a new problem:

Good news is no longer enough.

Companies can report strong revenue.

They can announce enormous AI demand. They can expand data-center capacity. They can beat earnings expectations.

And investors can still sell the stock.

Not because the AI revolution suddenly disappeared. But because expectations may have finally outrun reality.

That distinction matters enormously.

The greatest danger for AI stocks was never that artificial intelligence would fail.

It was that AI would succeed — and investors would already have priced in even more success than reality could deliver.

We may be approaching that moment.

Bitcoin Is About to Fork Again: The eCash Experiment Begins at Block 964,000.

At Block 964,000, Paul Sztorc’s eCash Fork Will Copy Bitcoin’s History, Activate Drivechains, and Put a Decade-Old Idea to Its First Real Market Test.

Markets Don’t Trade on Good or Bad

This is something many investors learn the hard way.

Stocks don’t rise because results are good. Stocks rise because results are better than what the market already expected.

There is a huge difference.

Imagine a company grows revenue 20%.

That sounds excellent.

But if investors expected 30%, the stock can collapse.

Now imagine revenue falls 5%.

Terrible.

But if investors expected a 15% decline, shares can rally.

Markets aren’t grading companies against absolute reality. They’re grading reality against expectations.

And after several years of extraordinary enthusiasm around artificial intelligence, expectations have become extraordinary too.

AI companies are no longer competing against ordinary growth assumptions. They are competing against a future Wall Street has already imagined.

A future of enormous productivity.

Exploding inference demand. Autonomous AI agents. Mass enterprise adoption. Trillions of dollars of infrastructure. Dominant margins. Massive free cash flow. And potentially entirely new industries.

When investors have already priced part of that future into today’s share price, delivering merely “excellent” numbers may no longer be enough.

You have to deliver spectacular numbers.

Again.

And again.

And again.

That becomes increasingly difficult.

AI Is Outbidding Bitcoin for Electricity: What Happens When Mining No Longer Pays Enough to Secure the Network?

The AI boom is turning electricity into the world’s hottest strategic asset—and forcing Bitcoin miners to prove just how much the network is really willing to pay for security.

The Nvidia Problem Is Becoming Everyone’s Problem

Nvidia taught investors to expect miracles.

For multiple stages of the AI boom, demand for accelerated computing expanded so quickly that forecasts repeatedly looked conservative.

Companies wanted more GPUs than Nvidia could deliver.

Cloud providers raced to secure capacity. AI laboratories wanted larger training clusters. Governments wanted sovereign AI infrastructure. Enterprises wanted inference capacity. Everyone seemed to want more compute.

Nvidia became the financial embodiment of that scarcity.

But extraordinary performance creates an uncomfortable consequence.

It resets the baseline.

Once 50% growth becomes normal, 30% can look disappointing. Once enormous earnings beats become routine, smaller beats feel weak. Once investors expect capacity to be sold out, merely strong utilization stops being exciting.

The company’s success raises the standard by which future success is judged.

And the same phenomenon is now spreading across the AI trade.

• AMD.

• Broadcom.

• Micron.

• Cloud providers.

• Data-center operators.

• Networking companies.

• Power infrastructure companies.

Anyone attached to AI spending has increasingly been valued against an exceptional future.

That means the question is changing from:

“Is AI demand strong?”

to:

“Is AI demand strong enough to justify what we are already paying?”

Those are very different questions.

When Great Earnings Produce a Falling Stock

This is one of the oldest signals in markets. And one of the most useful.

When bad news stops pushing an asset lower, sellers may be exhausted. When good news stops pushing an asset higher, buyers may be exhausted.

Neither guarantees an immediate reversal.

But both deserve attention.

Suppose an AI semiconductor company announces:

• Revenue up sharply.

• Data-center sales exploding.

• Management optimistic.

• Demand remains strong.

• New products progressing well.

And the stock falls.

The instinctive reaction is:

“That makes no sense.”

Actually, it can make perfect sense.

Perhaps everyone who wanted to buy already bought. Perhaps the good news was expected. Perhaps positioning is crowded. Perhaps valuation already assumes years of exceptional execution. Perhaps investors were hoping for something even better. Perhaps the market is beginning to question the returns generated by all the AI spending.

The earnings report may not have changed.

The price investors are willing to pay for those earnings has.

And that’s where things get dangerous.

AI Has Moved From Discovery to Expectations

The first phase of the AI trade was discovery.

Read more