Postcards from the Edge of the World: The Bailout

Editor’s Note: I want to stress up front that what I’m writing today is a scenario…

It’s grounded in what - based on my background in monetary, fiscal, and government policy - a major financial bailout (not called a bailout) would look like in the next few years if AI doesn’t pan out. I’m not saying this will happen, but I am saying that if this does happen, it will likely be the biggest buying opportunity for investors in the next few years because of the relationship between monetary and fiscal policy support and the performance of equity markets. I appreciate you keeping an open mind, because I don’t want this to be considered “doom editorial.” I propose it as one thing… how the world works.

Dear Fellow Traveler:

It started on a Saturday.

And that’s the worst day for all of this to happen.

It happened on a Saturday because that’s what the American government does.

They lock a bunch of people in a room over the weekend and try to come up with a plan for something we’d been warning about for years.

They didn’t do it on a Friday because that would be too much work.

They didn’t do it on a Thursday because the repo spike didn’t happen until the next day.

And they didn’t do it on the previous Monday because that was the day they were going to be doing the opposite of the new plan.

It started on a Saturday because they knew most of us would be at a bar drowning our sorrows, and that the day before, CNBC’s Jim Cramer had been on cable warning about the next Great Depression.

The Saturday I’m talking about was February 5, 2028.

Scott Bessent was leading the Treasury Department in one room.

Kevin Warsh was leading the Federal Reserve in the other.

There were people in another room from Congress wondering what all of this was about… And most of them were just looking at their phones.

Lawyers had gathered in a fourth room. They were eating sandwiches.

The fifth room was what really mattered.

That is where a team of public relations experts gathered to make sure that anyone in the other rooms wasn’t using the term “Bailout.”

On Sunday night, a press release went out. It was about eleven pages, and the word bailout didn’t appear in any of them. And they had checked this at least 12 times and even screwed up a previous draft and changed words to make it clear.

THIS WAS NOT A BAILOUT.

The press release went on and on about what happened in those rooms, and they explained that the different programs they used to stabilize the American economy were not new.

Every single program that was used to inject capital into the system was borrowed from a previous financial program and historical precedent.

They had used these tools in other industries like defense and energy… and did so on different Saturdays… under different Presidents… which was the point.

But in the end… there was a truth to it all.

Narrator’s Voice:

It was a bailout…

I am writing this from the future, and want to be clear that this is a scenario analysis. We have to consider the possibility of a future bailout…

Given the fact that the United States is so heavily invested in the AI trade and that our markets are now centralized around this effort to make it a major reality… let’s just suggest that it’s only a 5% chance…

In any scenario analysis, that figure is meaningful.

Keep an open mind, because it’s not necessarily a story about predicting the future so far as it is about showing you exactly how the machine works and what would happen if we experienced another liquidity crisis that was comparable to 2008 or 2020…

And let me be clear.

We will have another major financial crisis.

I don’t know if you’ve noticed, but we’ve been teetering on edge for six years, with multiple-sigma events happening… things that should only be happening every 63 years are happening annually.

So, if our bond markets crack because of AI, I want you to understand the terminology and the inside baseball.

This is how I think a bailout, or stabilization, or whatever term they choose, would work.

So… we need to go back to the fourth quarter of 2027.

Interest rates had been pressing higher for the better part of a year, and the Fed raised interest rates two times, creating a wall of pressure on foreign nations dealing with their own dollar-denominated debt.

People will point to an earnings report…

A large U.S. hyperscaler suddenly missed AI revenue expectations for the second straight quarter.

At the same time, a Southeastern data-center joint venture failed to refinance a $2.3 billion tranche… There was a brutal headline on a Friday afternoon about it, but it went ignored because people were too busy talking about the 2028 election.

The story I return to was the fact that a research lab out of China published a paper explaining that their frontier-scale model trained on domestically fabricated chips did so at a meaningfully lower compute cost…

That paper didn’t just hit AI stocks…

This fundamentally shifted the economics and the expected economics of the physical AI buildout.

This changed model pricing expectations, because if comparable models could be trained and operated with materially less compute, then data-center utilization assumptions fell…

And from there, it all moved downstream.

Power contracts now looked oversized, while projected lease income slumped.

The collateral that was now supporting hundreds of billions of dollars in private loans now looked a hell of a lot less secure.

This was the latest DeepSeek moment that we’d seen, and this was devastating.

It was a moment that we all collectively admitted that the financial assumptions of this buildout had been wrong. There were winners in the AI space, but not every hyperscaler was going to survive…

This news blew out credit spreads on investment-grade tech paper, as they widened by 40 basis points across five trading sessions.

It had been a while since we all panicked about private credit funds, but another four of them gated the fund level, and it was a big deal because we hadn’t seen that level of coordination on withdrawals since late 2025.

Ten days later, a senior official at the Federal Reserve Bank of New York used a phrase that made my skin crawl. He used the term “orderly market functioning in long-duration corporate financing markets” at a bond dealer conference.

Most people in the media didn’t pick up on this, but the private equity crowd certainly did, especially the leaders at Apollo, Ares, Blackstone and Blue Owl.

What came next was another four months of stress in the private credit space, and by the last week of December 2027, private credit funds gated a second time.

The New York Times wrote an article about Doug Ostrover, the co-head of Blue Owl and one of the owners of the Tampa Bay Lightning, sitting alone by himself in his box as his team got blown out 7-1 by the New York Rangers at home.

The investment-grade primary market went quiet again, and two commercial paper programs failed to roll their positions on a Thursday morning.

The Fed wasn’t at the time able to just cut interest rates into this stress.

It was considered, as the U.S. thought about going the route of the Bank of England at the height of the GILT Crisis. But inflation was still printing after yet another wave mid-2027 at nearly 3.5%, and it had sucked all the oxygen out of the room with the elections approaching.

With the government now over $44 trillion in debt, fiscal repression had accelerated, and with it inflation was the tail wagging the dog.

The dual mandate didn’t allow a March-2020 style monetary response, and the central bank didn’t have the authority to go out and start buying equities like Oracle, which was trading in the $40s by January and seemed to be spending money like they were betting on the government to rescue them...

Under the Fed’s Section 13(3) - which was amended by the Dodd-Frank Act, emergency Fed programs must be broad-based and can’t be designed to just rescue a single company.

It also requires Treasury approval, reliable collateral, and must be terminated in a timely and orderly manner.

It was these constraints that made it harder to address the real problems that had emerged that fall.

The Fed’s market intervention had to travel through funding pipes that weren’t linked to the Fed funds rate, and it had to move through more than one funding pipe, because one individual pile wasn’t really going to fix the stress in this market…

The weekend required coordination across four different institutions, and each of them did what they had the authority to do. And each of them deferred power to the other three institutions if they lacked the mandate.

The design came from four different programs that were stapled together so that everyone could check their boxes and never exceed their legal authority.

Again… they didn’t create anything that didn’t have a precedent.

And that is what is so troubling… and incredibly boring.

The Federal Reserve was first up to the plate and they’d been talking about this for weeks.

This wasn’t something that could be solved through the Standing Repo Facility, which had already seen its cap lifted on perpetual support to the financial system in late 2025.

But they did create a new facility to bailout the plumbing. It was called something very bland, very ordinary, and not at all controversial.

They had changed the font and font size a few times when they named it on that white paper, and it was simply known as the Strategic Infrastructure Credit Facility.

The media joked that it was called “SICK…”

This program bailed out commercial paper, asset-backed securities, banks, primary dealers, and largely private-credit financing… and provided heavy injections into any and all of the places where AI-related projects were funded.

The collateral that could be used for funding was largely investment-grade corporate paper, senior secured private-credit loan participations, mortgaged-backed securities, and a few new innovations that were linked to the growing Stablecoin ecosystem.

The Fed extended the loans at the discount window rate, and added a small spread that came down to the type of collateral.

They set haircuts on a schedule that was made public during the announcement, which was in the press release and on the Federal Reserve site.

It was confusing to everyone…

Only Reuters and Bloomberg took the time to explain all of this to the American public, but the sell-side research over at Goldman and JPMorgan was gloating about what this really meant.

That the S&P 500 was probably heading to the moon.

Insider buying activity compared to insider selling in real dollar amounts surged to levels that we hadn’t seen since the European Central Bank abandoned austerity measures in 2011. That had been the cleanest signal for markets on policy accommodation going back to the first round of QE in 2008.

There was a lot of controversy on CNBC on Monday, and Rick Santelli looked like he was in dire need of a Marlboro Red. “I guess we’re just going to bailout everyone,” was the statement that led the S&P 500 to rally about 3% that morning premarket, and we finished the day up 5.5%. And then another 4% the next day.

Some younger journalists were wondering what was the precedent for the SICF.

People thought it might have been linked to the 2008 alphabet soup of AIG and Bear Sterns activities.

But it was actually the Bank Term Funding Program that the Fed created in March 2023 at the height of the Silicon Valley Banking crisis.

The BTFP had been a Section 13(3) facility that used a very quiet statutory power to lend money against bonds at par value instead of at mark-to-market value.

They had built the BTFP over a weekend, as they always do.

They likely did it on a Saturday and announced it on a Sunday, and one never took the time to see if everyone’s hands at the Fed were okay after they broke so much glass in a hurry.

It was also supported by a Treasury equity contribution from the Exchange Stabilization Fund.

And, it was terminated a year after they opened it up. That’s important.

The SICF carried all of this architecture, and 97% of Americans had no idea what in the hell any of this even meant. A lot of traders tried to raise cash… and margin calls happened on that Thursday.

Some people were talking about the next Great Depression on a Friday.

By Monday, we were all talking about the Last American Bull Market… as insider buying spiked, and the S&P 500 started yet another rally that put the COVID rebound to shame. Democrats helped build the bailout package while railing against it. Republicans and moderates stood around trying to make sense of the last 20 years, ignoring the fact that this permanent bailout machine was feeding so much class resentment and would continue to do so in the years ahead.

All that said, the only thing that was different about this program compared to the 2023 SVB Crisis bailout… was that the collateral eligibility expanded and the Treasury backstop had increased by a large margin.

Again… they wouldn’t use the term bailout.

The Fed wasn’t allowed under Section 13(3) to take equity in companies whose paper it was accepting as collateral. And this was all happening around the hyperscaler space.

For that to happen, Congress would need to authorize a separate program for this. It was now campaign season.

Most pundits believed that there wouldn’t be any real tolerance for yet another bailout, but this was backstopping a major bet by the United States.

And it all went very quickly…

The thing that was interesting was that the U.S. was still ahead of China in the AI race. But China never had to win.

They simply needed to be close, and as their costs dropped, the data center spending went into a frenzy until it suddenly halted… and everyone, everywhere suddenly started to become an expert on the free cash flow of hyperscalers.

The revenue projected in this arms race never reached its expected goals, and now people started to openly ask if some of the biggest players would abandon AI the way that Meta had abandoned the Metaverse after ramping up unsustainable spending levels.

Congress did have to act though… and it was a circus and was so loud and noisy that the debates were covered by CNN… and people who didn’t understand finance trying to explain finance… and asking stupid questions like “How did we get here?” and blaming greed and capitalism instead of say… central planning.

Congress, after two failed votes, authorized the Strategic Compute Stabilization Program. Bernie Sanders voted against it, and went on an epic rant on 60 Minutes in late April that broke a 24-hour record on YouTube.

People seemed to miss the fact that capitalism had nothing to do with this. The government was at the heart of this, and the Fed and Treasury’s actions dating back to 2008 were the opposite of “voluntary transactions among willing participants.”

This program was structured much like TARP in 2008, and allowed the government to buy preferred stock and warrants in qualifying AI-stack firms…

These companies received capital support and specific behavioral covenants.

Interesting enough, this worked out much like the bailout of AIG.

The Fed stabilized the financing markets.

And the U.S. Treasury department bought the preferred stock and warrants.

The preferred stock not only bounced back over the next 18 months, but it paid a rich dividend.

The warrants attached to the preferred gave Treasury an upside participation if the firm’s equity recovered. Not every company did recover, but on a net level, the government bragged about the deal and ended up making tens of billions of dollars that it then turned around and spent in the next 24 hours.

These covenants also did include executive-compensation caps, dividend restrictions, and mandatory CapEx commitment in various locations around the nation. It was the government telling private businesses what to do (dirigisme).

The government also demanded employment-maintenance quotas that were linked to construction jobs and U.S. manufacturing. They started using the term “shovel ready” to describe jobs again, even though a lot of the jobs weren’t shovel ready.

The SCSP was the part that gave the Treasury Department direct equity holdings and made the second half of this decade legal...

The TARP parallel and the increasing role of government in U.S. industry wasn’t an accident. Between 2008 and 2013, the Treasury bought direct stakes in banks and in GM and Chrysler.

TARP had its own covenants and warrant grants. It had repayment schedules. And that 2008 period had established the legal template for how Treasury investment can be structured, priced, and ultimately exited.

But the exit was the tricky part.

Given that the U.S. had already taken stakes in previous companies like Intel and IBM, this also became an interesting period because Democrats started to talk about strategic industries as well.

In early 2028, while on stage for the Democratic debates, someone would pitch that the U.S. should buy a public healthcare company, and use that ownership as a tool to press nationalized healthcare and the public option.

The hosts of MSNBC’s Morning Joe would say the following week, “Why didn’t we think of that?”

The thing about this bailout was that Congress would do more than just this SCSP.

They were, by default, subsidizing the physical AI buildout through the fiscal pathway.

And they were doing it on a schedule that ran in parallel to the Fed’s efforts.

This also boosted CHIPS-style tax credits on qualifying advanced-manufacturing investments.

It also funded even more power generation and transmission support that aimed to expand grid capacity in specific regions of the nation.

We’d already seen data-center demand outrun available electricity base load and there was really only one thing that could get everyone in one room around the grid.

Money.

Congress would also shift the game a bit and allow for accelerated depreciation on qualifying AI infrastructure investment and federal loan guarantees on strategic projects at a level that straight up dwarfed the existing DOE Loan Programs Office.

None of this required an emergency meeting on a Sunday.

But a lot of this was already drafted before a January 2028 liquidity stress event.

They had this printed… ready to break the glass.

When the weekend of that stress arrived, the emergency vehicle was parked in the driveway of the Rayburn building.

There were four parts to this, and the last part involves the Pentagon and the U.S. Department of Energy. Because no one really thinks about the demand side as part of a bailout. But it is actually the single most important function.

The procurement channel is the one source of demand that goes ignored, but it’s incredibly important in helping to create a backlog across almost any industry.

So, what did they do to backstop it even more?

Both groups signed agreements that committed the U.S. government to guaranteed purchases of specific amounts of compute capacity at pre-determined prices over multiple years.

It’s not just about printing money.

It’s about creating demand - good old fashioned economic demand from a buyer of last resort.

In the spring of 2028, we witnessed domestic-content requirements linked to all those procurements, and they were enforcing the fiscal channel incentives.

The Fed printed the money, the Treasury managed it, and the government’s defense department and energy leaders created the demand.

Government spending as a share of GDP surged.

The Defense Production Act was also back in play. Because the administration decided that supply-chain assets and compute were now strategic infrastructure. The government provided priority-rating and allocation authority through the DPA’s Title I framework.

This gave the government the right to jump ahead in line on anything.

We’re talking about anything from specialty concrete to substation transformers.

And that was it…

That was the bailout.

It wasn’t Kevin Warsh showing up at an Nvidia board meeting with a bunch of money in a Subaru and a leather jacket. It was all incredibly boring… and most Americans didn’t even realize that it was happening.

It carried a bunch of acronyms, and dry procurement schedules, and Federal Register notices that went onto websites but were never read.

There wasn’t a helicopter dropping money from a sky, and there were people everywhere claiming that it wasn’t a bailout…

It wasn’t “Quantitative Easing” it was selective maintenance of a massive financial and psychological bet that the nation made, a Cold War around AI, where no one really won, but China didn’t lose.

Because it was so dull, every official continued to insist that it wasn’t a bailout, even though it was…

“For the love of God…” I’d eventually scream in the pages of Me and the Money Printer in the summer of 2028… “the American AI industry was physically carried out of the burning building on a Federal Register stretcher, wrapped in an appropriations rider, escorted by three lawyers and a Congressional Budget Office estimate marked down by two decimal places to fit inside the fiscal window.”

People would be confused by what I was saying… and they’d turn to the media to explain it. And when someone at Reuters finally got around to asking whether this was a financial rescue, Kevin Warsh’s answer on 60 Minutes was that each individual instrument and program preexisted and had they had the authority to deploy it.

And that’s all true…

And also extremely disturbing…

On Our Precedent

Once again, nothing was invented from scratch. As the Bank for International Settlements pointed out, this has been the game for the last 40 years, and there isn’t political will to make it stop.

The Federal Government and its central bank was building the toolkit that it used in 2028… since the 1970s… and every instrument used in February 2028 had a precedent that was specific and precise. Sure, all of these things were announced as one-time exceptions.

It’s always interesting to listen to people act like this all started in 2008 and that there was a one-time exception around TARP.

But the first modern precedent of the government guaranteeing the debt of a large American firm was Lockheed Martin in 1971.

Because Lockheed was central to U.S. defense, Nixon signed the Emergency Loan Guarantee Act in August 1971.

That act authorized up to $250 million in federal loan guarantees, and the initial and only user was Lockheed Aircraft,

The company was carrying cost overruns on the C-5 Galaxy program and had been hit by the collapse of Rolls-Royce.

Milton Friedman opposed the guarantees…

The Nixon administration argued that Lockheed was the country’s largest defense contractor and that allowing its bankruptcy would disrupt ongoing military production, and destroy part of a specialized manufacturing business that wasn’t easy to replace. Congress agreed.

Eight years later, the Chrysler Corporation Loan Guarantee Act authorized $1.5 billion in federal loan guarantees for Chrysler… They ran out of cash under Lee Iacocca in the second oil shock.

The legal architecture was almost identical to Lockheed. This was federal guarantees rather than direct loans. If a firm’s failure took out enough American manufacturing jobs, it qualified. That extension is the exact one Congress relied on again in 2008 and again in 2028.

Eleven days after the September 11 attacks, Congress passed the Air Transportation Safety and System Stabilization Act. It authorized $5 billion in direct compensation payments to airlines for their losses during the September 11 to 14 grounding. The U.S. created the Air Transportation Stabilization Board to handle the loan-guarantees.

The U.S. took direct equity positions through TARP in General Motors and Chrysler between 2008 and 2013. GM and Chrysler received combined support of roughly $80 billion.

The CHIPS and Science Act signed in August 2022 authorized roughly $52.7 billion in semiconductor-manufacturing support. The bill was pitched as national security.

The Department of Defense’s investment in MP Materials in 2025 became the largest shareholder of a publicly traded rare-earth miner through preferred stock and warrants.

And that deal went pretty far.

It combined direct federal equity with a physical-commodity price floor. It routed the transaction through the Defense Department procurement authority rather than TARP-style statute. And it was announced without any financial-stress trigger. For the first time, we had a real political precedent for direct federal equity in a strategic firm… no crisis required.

So, in 2028, when Congress was drafting SCSP, the MP Materials structure was the immediate legal precedent that made the direct-equity provisions politically comfortable.

And the Bank Term Funding Program was the structural precedent for the SICF.

BTFP was announced Sunday March 12, 2023, two days after Silicon Valley Bank failed. It was a Section 13(3) emergency facility that lent to eligible depository institutions for up to one year against collateral valued at par rather than market.

It was hilarious in its ability to bend the rules of modern finance and how bond markets work. But it was deemed necessary… or else.

This is How It Works

Trace those precedents and the shape becomes obvious.

• Each intervention was supposed to be a one-time exception.

• Each expanded the view of what qualified as “essential” enough for federal intervention.

• Each borrowed instruments from earlier interventions and added new ones.

• Each shortened the timeline over which the next intervention could be completed.

• Each normalized a piece of legal and institutional architecture that the next crisis could reach for without a philosophical debate.

Once an industry gets designated essential to defense, industrial capacity, or systemic employment, the philosophical objection to intervention evaporates.

That’s what makes this nation so absurd.

Nothing is ever… EVER a one-time exception. The toolkit only grows… and the money printer only hums.

But keep something in mind…

You’re not reading this in Summer 2028.

It is really July 2026.

The Saturday in February 2028 hasn’t happened… yet.

The 11-page press release has not been drafted. The SICF has not been announced, and the SCSP isn’t authorized yet…

The S&P is actually still near an all-time high.

Everything above is a scenario.

Even if this is just a 5% probability, I want you to be prepared and I want you to be able to see how it works.

The Fed prints the money… the Treasury manages the money… and the government creates the demand through strategic language at the Pentagon, Department of Energy, and other agencies.

But there’s one last thing that I want to point out, and that’s the irony in all of it.

The fact that when (not if) the next bailout comes… Americans will likely be extremely bearish… people will be talking about the next Depression… and the world may look extremely dangerous.

At that point… when the printer hums, it will be the ideal time to buy…

Historically, some of the greatest buying opportunities have arrived when monetary support and fiscal demand turned in the same direction.

You can love that… or you can hate it.

But you can’t fight it.

I appreciate you taking the time this evening…

Now… it is time for me to give you a stock recommendation for the long-term.

But I’m going to tell you that it’s going to come in two parts…