| |
| Case File № DRT-ED-AI · Special File · The AI Fingerprint |
| |
| 1929. 2000. 2008. We Found the Same Fingerprint on the AI Boom. |
| |
| Record spending, $1.2 trillion in AI-linked debt plus hundreds of billions more off the books, a market resting on a few names, and Washington insisting it’s a golden age. History has seen this setup before. It rarely ends quietly. |
| |
| Threat level 4/5 — Elevated |
| |
|
| |
| |
| |
| Every financial crisis of the last hundred years left the same fingerprint before it hit. Prices ran far ahead of profits. Borrowing hid in places regulators weren’t looking. Fewer and fewer stocks carried the market. And the people closest to the boom kept insisting this time was different. We went looking for that fingerprint in the AI trade. Here’s what’s on the evidence board. |
| |
| The pattern |
| |
| |
| |
The margin boom Stocks bought with borrowed money, a market that only went up, and officials who called the economy sound. The Dow eventually lost close to 90% from its peak. |
|
| |
The dot-com bubble Investment in tech hit record highs while profits lagged far behind. The Nasdaq fell about 78% from March 2000 to October 2002. |
|
| |
The leverage crisis Debt was buried in off-balance-sheet vehicles most investors never saw. When it surfaced, bank stocks lost most, and in some cases all, of their value. |
|
| |
| |
| The evidence board |
| |
| | What the AI Boom Looks Like Right Now |
|
| |
|
| |
| Exhibit A — the spending. Microsoft, Alphabet, Amazon and Meta are on track to spend roughly $700 billion on AI infrastructure in 2026, mostly data centers and chips. In early 2026, Goldman Sachs pointed to peak investment spending as the first warning sign from the dot-com era. |
| |
|
| |
| Exhibit B — the hidden debt. A draft Treasury analysis obtained by NOTUS in July put visible AI-related corporate debt at about $1.2 trillion, roughly 14% of JPMorgan’s investment-grade index. It cited Moody’s estimate of $662 billion in data-center leases held off balance sheet through special-purpose vehicles funded by private credit. Treasury called the draft “unvetted.” |
| |
|
| |
| Exhibit C — the cash squeeze. The same analysis projects hyperscalers will plow about 94% of their operating cash flow back into infrastructure by 2026–2027. This week Micron posted a record quarter and its stock still slipped after it announced higher spending: roughly $25 billion in capex in half a year. |
| |
|
| |
| Exhibit D — the thinning market. Fewer than half of S&P 500 stocks trade above their 200-day moving average, down from about 75% in mid-August. The index looks healthy only because a handful of leaders are holding it up. |
| |
|
| |
| Exhibit E — the cost of money. In 1999 the Fed cut rates and fueled the rally. In 2026 the Fed has hiked, and the 10-year Treasury hit its highest level since 2002 this week. Every dollar of AI debt now costs more to carry. |
| |
| Decoder |
| |
| | The Line Every Bubble Uses |
|
| |
| | “Artificial intelligence will be a key driver of America’s new Golden Age.” — Treasury spokesperson, responding to the draft report on AI debt |
|
| |
| What it says |
| |
| There’s nothing to worry about. |
| |
| What it means |
| |
| Our theory: AI may well transform the economy, and that’s exactly the problem. The internet did transform the economy, and the Nasdaq still lost 78%. A technology can be real and still be badly overpriced. When officials answer a warning about debt with a promise about the future, they’re not answering the question. |
| |
| Place your bets |
| |
| |
| |
| A major AI borrower struggles to refinance as yields stay above 5% 35% |
| |
|
| |
| A hyperscaler cuts its capex plan, and suppliers sell off together 30% |
| |
|
| |
| Private-credit losses tied to data-center leases surface publicly 20% |
| |
|
| |
| None of these by year-end: the boom keeps running 15% |
| |
|
| |
| Our best guesses. The four add up to 100%. |
| |
Classification Pressure, building |
|
| |
| The fingerprint is there: record spending, debt hidden off balance sheet, a narrowing market, rising rates and official reassurance. None of that tells you the day the turn comes. It tells you the kind of protection people wished they’d had in 2000 and 2008 is worth looking at before the turn, not after. |
| |
The riddle · answer at the end of the file In 2000 and in 2008, the warning signs were public for months. Why did so few investors act on them in time? |
|
| |
| |
|
| |
| Riddle answer |
| |
| Because the crowd was still making money. Warnings sound foolish while prices are rising, and the people who issue them get laughed at, right up until they don’t. Our theory: the investors who came through 2000 and 2008 intact didn’t predict the exact day. They decided in advance what they would do when the signals lined up. The signals are lining up now. |
| |
