Five American technology companies report about $1.35 trillion of debt. A Nikkei study puts another $1.65 trillion below that line — legal, disclosed, and larger than the number almost everyone is watching. The financing can break even if the technology does not: that is exactly what happened after March 2000, when the index fell seventy eight percent, took fifteen years to recover, and the internet kept compounding the whole time.

This page walks the evidence — the waterline, a Louisiana pastry holding twenty seven billion dollars, the circular trade, the dot com running order — then the strongest case that this is not a bubble, and finally a plan built on three things you control: runway, independence, and position. Six moves, in order. Every one of them leaves you better off in the world where nothing ever goes wrong.

The bubble is not the reason to do any of it. It is only the reason to start now instead of later.

Above the line is what investors read. Below it is the rest.

In a rural stretch of Richland Parish, Louisiana, Meta is building a data centre called Hyperion. It is enormous. It is also, in a specific accounting sense, not entirely Meta's.

Blue Owl Capital, a private credit firm, formed a legal entity to hold it. The entity is called Beignet Investor LLC, named after the fried dough you eat in New Orleans. Beignet sold roughly twenty seven billion dollars of bonds to Wall Street. Blue Owl owns eighty percent of the venture. Meta owns twenty, leases the facility, and pays rent. That rent flows to Blue Owl, which uses it to pay the bondholders.

The naming convention is not accidental. Beignet is beige. Bland. Forgettable. Enron hid debt in entities named after Star Wars characters, and the market laughed until it didn't. Meta names theirs after pastries. The mechanism is identical.

Read that again with the accounting in mind. Meta gets the data centre. Somebody else gets the debt. The rent obligation is real, enormous, and multi decade, and it does not appear as debt on Meta's balance sheet because it is not, technically, Meta's borrowing.

This is not a loophole somebody stumbled into. Morgan Stanley designed the structure and ran a competitive auction to pick the partners. It carries an investment grade rating precisely because the rating agencies looked through the structure and priced Meta's creditworthiness anyway. Everybody involved knows exactly what it is.

Here is the thing almost nobody says out loud, and it is the reason this page exists.

The financing story and the technology story are two different stories, and they can end differently.

Every time this gets discussed, the two collapse into one. Either AI is a fraud and the whole thing is about to evaporate, or AI is inevitable and the debt is a footnote for people who do not understand growth. Both readings are lazy, and both will cost you money if you plan around them.

Financing can break spectacularly while the underlying technology keeps compounding. That is not a hypothetical. It is the single most reliable pattern in the history of infrastructure booms, and it happened in living memory, to a generation of people who had exactly your job description.

Money that keeps arriving back where it started

A chip maker invests in a model company. The model company spends the investment on chips. A cloud provider signs a capacity deal with the model company and books it as revenue. The same dollar can be counted three times, in three sets of accounts, as three different kinds of good news. Filter the diagram by relationship type to see how tight the loop is.

This is not an economy. This is a carousel. And carousels stop spinning.

March 2000, and the fifteen years that followed

The dot com crash is the closest thing we have to a map. Not because the technologies rhyme, but because the sequence of who got hurt, and in what order, is almost certainly going to repeat. Read this as a running order, not a prophecy.

The index peaked in March 2000 and bottomed in October 2002, down about seventy eight percent over thirty months. The recession itself was mild and lasted eight months. The job market was the slow part, and it was slow for years.

The index did not reclaim its 2000 high until 2015. Fifteen years. Not because the technology failed, but because the price had run so far ahead of the earnings that the earnings needed fifteen years to catch up.

And here is the part that should shape your planning more than any of the above. The recovery did not come from the old companies returning. It came from new companies buying the wreckage cheaply. All that overbuilt dark fibre sat unused, got sold for a fraction of its cost, and became the physical substrate for streaming video and cloud computing a few years later. Amazon fell about ninety five percent and survived because it held cash and ran a real business. Google went public in 2004, into the aftermath, and won partly because advertising was cheap and talented people were available.

The infrastructure was real even though the valuations were fake.

The strongest case that this is not a bubble

If you only read the alarming version, you will make expensive decisions. So here is the opposing case, made as well as its proponents make it, because a plan built without it is a plan built on half the evidence.

The honest synthesis, and the position this page takes: both readings can be substantially right at once. The technology can be genuinely transformative and the financing can still be overextended, because those are separate claims about separate things. Railways were transformative and the railway financiers still went bankrupt in waves. Electrification was transformative and the utility holding companies of the 1920s still collapsed. Being right about the technology has never protected anyone from being wrong about the price or the leverage.

Which is why nothing in the plan below requires you to pick a side. Every step in it is defensible whether the optimists or the pessimists turn out to be closer to correct, and that is the specific property that made each one worth including.

Three things you control

Not three predictions. Three variables that sit entirely inside your own life, that behave well under uncertainty, and that get stronger the longer nothing happens. Everything in the plan hangs off one of these.

Step one is arithmetic, and you can finish it this afternoon. Write down your bare bones monthly cost of living. Housing, food, utilities, school fees, transport, minimum debt payments, insurance. Nothing aspirational, nothing optional, nothing you would keep paying out of habit if the income stopped. Multiply by six. That number is your target. Most people have never calculated it and are carrying a vague sense of dread instead, which is both more stressful and less useful than a figure on a page.

The bare bones version matters more than the comfortable version. In a genuine income interruption you do not spend at your current rate, you spend at your floor, and the floor is usually thirty to forty percent below what people guess. Calculating the floor honestly is what turns six months of savings into eight months of actual survival.

Step two is debt, and it comes before saving past one month. If you are carrying a balance above roughly ten percent interest, paying it down is a guaranteed, tax free return that beats anything else available to you. There is no investment on offer with that risk profile. Hold one month of expenses as a working buffer so a broken laptop does not put you back on the card, then send everything spare at the highest rate balance first. Do not move to step three with credit card debt alive. It is the one place where the order genuinely matters.

Step three is the cushion itself, and currency is part of the design. If you earn in one currency and spend in another, that split is a feature and you should use it deliberately. Holding a meaningful portion of the cushion in the currency you earn in means that when a risk off move strengthens that currency, your buffer grows in local terms at exactly the moment your income is threatened. Roughly sixty percent in the earning currency and forty in the spending currency is a defensible starting split. Adjust for where your obligations actually are.

Keep it boring. Instant access or short term deposits under ninety days. Not crypto, not equities, not anything with a lock up longer than the emergency you are insuring against. This money has one job and that job is not growth. The cushion earning a poor return is the price of the cushion being there in the week you need it, and it is a cheap price.

A reasonable objection: six months feels like a lot when the money could be invested. It is a lot. It is also why the people who survived 2001 were not the smartest ones, they were the ones who could wait thirty months without being forced to take a bad job at a bad rate. Optionality is the return on cash, and it does not show up in a spreadsheet.

Count your single points of failure honestly. One employer is one. One sector is another. One client geography is a third. One currency is a fourth. If the same event knocks out all four simultaneously, you do not have four exposures, you have one exposure wearing four hats. This is the most common misdiagnosis in personal financial planning and it is almost always optimistic in the same direction.

The test is uncomfortable and takes about a minute. Name the single event that would most damage your income. Now go down your list of assets, income sources and plans, and mark every one that gets worse in that same scenario. Whatever is left unmarked is your actual diversification. For a lot of people working in or adjacent to technology right now, the honest answer is close to nothing.

Add one income stream that does not share the failure mode. Not a second job in the same sector for the same kind of buyer. Something where the customer, the currency, or the reason they are paying is genuinely different. Twenty to thirty percent of total income from a source that survives the primary scenario is enough to change the character of a bad year completely. It is the difference between a crisis and a pay cut.

The realistic version of this is not a new career. It is usually the same underlying skill sold to a different kind of buyer, or productised into something that does not require your hours. The test to apply is simple: would this customer still be paying in a market where the category I currently work in has become unfashionable? If the answer is yes, it is real diversification. If the answer is that they would probably pause and see, it is the same exposure with a different invoice on top.

Understand what your index fund actually holds. A broad market fund today is substantially a position in a small number of very large technology companies. That is not an argument against holding it. It is an argument against believing that holding it, plus a job in technology, plus savings in the same currency, constitutes a hedge. It does not. It is one bet expressed three ways, and the danger is not the bet, it is the false sense of safety.

Worth saying plainly: none of this is a recommendation to sell anything or to time a market. Timing is the thing that actually destroys returns for individuals, and people who correctly identified the last bubble frequently sat out the four best years of their investing lives waiting to be proved right. Participate. Just participate in a form that survives being wrong.

Sell outcomes, not categories. This is the highest leverage change available to most people reading this, and it costs nothing. Identical work, described two ways: "AI powered automation for your business" versus "this removes twelve hours a week from your front desk and cuts missed calls to near zero." The first is priced on enthusiasm for a category and dies when the category becomes unfashionable. The second is priced on a measurable saving and survives, because the twelve hours are still twelve hours regardless of what the market thinks of the word in the headline.

Go through everything customer facing and check which version you are running. Your profile, your proposals, your pitch, the first line of your outreach. If the noun doing the persuading is the technology rather than the result, you are exposed to sentiment you do not control, and sentiment is precisely the thing that moves first and fastest in a correction.

Compound what does not commoditise. Generic execution is getting cheaper every quarter, and that trend is not reversing. What is getting more valuable is the judgment about what is worth executing: understanding a specific domain deeply enough to know which problems are actually expensive, holding relationships with people who trust your read, and having an audience that arrives without being bought. These are slow to build and cannot be downloaded, which is exactly why they hold value.

Practical version: pick one domain where you understand the operational reality better than a generalist ever will, and get deliberately deeper there rather than broader everywhere. Depth in one vertical plus competent use of modern tools beats broad tool knowledge with no domain, and the gap between those two is widening fast.

Write the contingency once, while you are calm. One page. If income stops on Monday: what gets cancelled in week one, who gets contacted in week one, what the cushion covers and for how long, what rate you will accept as a floor and what you will not. Writing it now is the entire point, because you will not think clearly in the actual week. People in that week take the first thing offered, at the first price offered, and regret it for two years.

Include the fallback rate specifically, and decide it in advance. It is the number people abandon fastest under pressure and the one that takes longest to recover once conceded, because your next client's price anchors to your last one.

If nothing breaks, this pillar still pays. Outcome based positioning wins work in good markets too, it just wins it at better margins. That is the test every step on this page had to pass before it was included: does this leave you better off in the world where the alarm turns out to be nothing?

Six moves, in order, starting this week

Print this. Put it somewhere physical. The order is deliberate and the sequence is where most of the value sits, because doing step five before step two is how people end up with an investment account and a credit card balance at the same time.

Why this works in both worlds. Run the plan and nothing breaks: you end up with no expensive debt, six months of liquid savings, a second income stream, and a sharper commercial position. That is not disaster preparation, that is just being in better financial shape than you were, and it would be worth doing in a year with no headlines in it at all.

Run the plan and something does break: you are one of the people who can wait. You do not take the first offer at the first price. You do not sell anything at the bottom because rent is due. You are, for a while, in the position the people buying dark fibre in 2003 were in, which is the only genuinely enviable position in a downturn.

The bubble is not the reason to do any of this. It is only the reason to start now instead of later.

The first seventy two hours, in order

Everything above this line is preparation, and preparation is where almost all of the value sits. This is the other half: what to actually do in the week the headlines are bleeding red and everybody around you is improvising. It is short on purpose. Panic is not a plan, and neither is refreshing the news.

Notice what is nowhere on this list. Nothing here asks you to predict anything, time anything, or sell anything. Six moves before, six moves after, and every one of them is something you can do on an ordinary Tuesday without knowing what happens next.