Letter No. 1 · The Snake Eating Its Tail · Part 1 of 3
The Snake Eating Its Tail: From the Y2K Fiber Glut to the AI Hyperscaler Loop
Thirty years, one thread: every crash financed the next boom — and today's AI capex loop is the 1990s vendor-financing playbook re-skinned with GPUs.
To understand how we arrived at today's staggering AI capital expenditures, hardware shortages, and circular tech financing, you have to trace a single, unbroken macroeconomic thread.
For thirty years, the global economy has run on a repeating engine: a speculative technology boom built on massive capital expenditure, followed by a debt-fueled crash, followed by central bank liquidity injections, which then ignite the next speculative asset bubble.
This is the story of how the infrastructure of the information age was built, liquidated, repurposed, and ultimately financialized into the "singularity before the singularity."
Act I: The Post-Gulf War Hangover & The Y2K Catalyst (1991–1996)
The early 1990s opened in the shadow of a mild but grinding recession. The Gulf War oil shock, combined with post-Cold War defense cuts and a commercial real estate savings-and-loan crisis, left corporate America looking for its next growth engine.
That engine arrived via two converging forces: the commercialization of the internet and the existential panic of Y2K.
As early desktop PCs trickled into corporate offices, IT departments realized that millions of mainframe systems were hardcoded to read years using only two digits. The fear of global technological collapse on January 1, 2000, triggered an unprecedented, mandatory corporate spending spree.
Corporations worldwide ripped out legacy infrastructure and bought brand-new enterprise servers, networking hardware, and database software. Y2K acted as a massive, forced modernization campaign — subsidizing the foundational digital plumbing of corporate America.
Act II: Dark Fiber, Vendor Financing, & The Dot-Com Meltdown (1996–2001)
With corporate networks modernized, the speculative rush began. Telecom giants and upstarts like WorldCom, Global Crossing, and Qwest embarked on a historical CapEx expansion, laying millions of miles of fiber-optic cables across continents and under oceans.
To fund this, equipment vendors like Lucent, Nortel, and Cisco engaged in aggressive vendor financing — lending billions directly to unproven telecom startups so those startups could buy their routers and optical gear. Stock prices soared, and capital seemed infinite.
By 2000, supply had wildly outstripped demand. The overwhelming majority of the newly laid fiber sat "dark" and unlit — contemporary estimates ran into the high nineties as a percentage of installed strand miles, though the figure varies with how you define lit capacity and which year you measure. The dot-com equity bubble burst, corporate fraud (WorldCom, Enron) was exposed, and the telecom sector dissolved into a wave of massive bankruptcies.
The "Cents on the Dollar" Pivot
In the bankruptcy courts, distressed asset investors purchased thousands of miles of state-of-the-art glass fiber for a few cents on the dollar. The original debt was wiped out, but the physical backbone remained.
By drastically lowering bandwidth costs, this subsidized infrastructure inadvertently laid the physical highway that enabled the modern web — paving the way for broadband, e-commerce, streaming, and cloud computing.
Act III: The Greenspan Pivot & The Real Estate Financial Machine (2001–2008)
Faced with the dot-com crash, the 2001 recession, corporate accounting scandals, and the geopolitical shock of 9/11, Federal Reserve Chairman Alan Greenspan took aggressive action. The Fed slashed the Federal Funds rate from 6.5% down to 1.0% by June 2003.
This deluge of cheap money needed a place to yield returns. Capital migrated from tech stocks into residential real estate.
[Tech/Telecom Bust] ──► [Fed Drops Rates to 1%] ──► [Wall Street Financial Engineering] ──► [Global Real Estate Bubble]
Wall Street securitized this cheap liquidity by bundling mortgages into Mortgage-Backed Securities (MBS) and Collateralized Debt Obligations (CDOs). Real estate transformed from a shelter asset into a leveraged speculative instrument. When housing prices stagnated and adjustable rates began to reset, the systemic risk brought down the global banking system in the 2008 Global Financial Crisis (GFC).
Act IV: ZIRP, QE, & The Tax-Driven Capital Shift (2009–2019)
To rescue the financial system from total collapse, central banks entered uncharted territory: Zero Interest Rate Policy (ZIRP) and Quantitative Easing (QE).
For over a decade, money was essentially free. Capital flooded back into technology and real estate in search of yield. Unprofitable "blitzscaling" tech startups were subsidized by cheap venture capital, while real estate operators leveraged low borrowing costs to acquire and consolidate assets across sub-sectors.
By late 2017, fiscal policy joined monetary policy with the passage of the Tax Cuts and Jobs Act (TCJA). Corporate tax rates dropped to 21%, bonus depreciation incentivized heavy capital investments, and Section 199A pass-through deductions altered small-business cash flows.
The Inflation Pivot: When Printed Liquidity Hits Finite Reality
To understand the bridge between the ZIRP era and today's AI frenzy, we have to look past popular catchphrases and define inflation by its core economic principle:
Inflation is an expansion of the circulating money supply faster than the growth of real goods and services — literally, too much currency chasing too few physical things.
When money is injected into an economy, it is created through two mechanisms:
Monetary expansion. Central banks create bank reserves out of thin air to purchase government bonds (QE), or lower interest rates to encourage commercial banks to issue private credit.
Fiscal expansion. Governments borrow that newly created money or issue treasury debt to directly fund spending, stimulus checks, corporate subsidies, and tax cuts.
Why the Pandemic Injection Was Unprecedented
Every crisis in our thirty-year chain involved a liquidity injection, but their paths into the economy were vastly different:
2001 (post-dot-com). Liquidity entered primarily through lower mortgage borrowing rates, inflating real estate prices — asset inflation, not consumer inflation.
2008 (post-GFC). The Fed's balance sheet grew roughly five-fold, from about $0.9 trillion before the crisis to roughly $4.5 trillion by 2014. But that money largely did not reach the street: it accumulated as bank reserves, which topped $1 trillion and reached about $2.5 trillion by 2014. QE repaired depleted capital at the banks and prevented a depression. It did not land in consumer checking accounts, which is precisely why a decade of "money printing" produced so little consumer inflation and so much asset inflation.
2020 (the pandemic era). Monetary and fiscal policy fused completely. Central banks expanded M2 at record pace while governments simultaneously disbursed trillions in direct relief, forgivable loans, and stimulus spending.
For the first time in modern history, a massive monetary injection landed directly in the hands of consumers and businesses at the exact moment global supply chains snapped. The result was textbook macroeconomics: an explosion of fiat currency colliding with physical scarcity, triggering a global inflationary shock. The difference from 2008 was not the size of the injection. It was the destination.
Act V: Today's Hyperscaler Loop & The Hardware "Tax"
By the early 2020s, as central banks raised interest rates to combat broad consumer inflation, Big Tech faced a growth crisis. Saturation had arrived in consumer software — every smartphone was sold, every ad impression monetized, every basic cloud workload migrated.
Enter the AI compute boom, powered by a modernized variation of the 1990s vendor-financing playbook:
The circular loop. Big Tech hyperscalers invest billions into frontier AI research labs or "neocloud" providers. Those labs use the capital to rent compute or buy server hardware from monopoly chip vendors. The chip vendors report record earnings, fueling high stock valuations that allow them to fund more capital allocation.
The distillation crisis. Massive, capital-intensive frontier models are proving difficult to monetize relative to their compute and power costs. Meanwhile, open-weights and distilled models increasingly deliver a large fraction of frontier capability at a small fraction of the cost — the precise ratio is contested and moves with every release, but the direction is not in dispute, and it threatens near-term return on billions in hardware CapEx.
The Inflation Conversation, Full Circle
This brings us directly back to the inflation engine. Just as pandemic stimulus injected trillions into a constrained consumer supply chain, Big Tech is now injecting hundreds of billions of stock-margined capital into a constrained industrial supply chain.
Capital can be created digitally overnight; physical infrastructure cannot. Hyperscalers are bidding against the rest of the world for finite resources: silicon wafer capacity, high-bandwidth memory, electrical transformers, and power grid access.
[Trillions in Margined/Corporate Capital] ──► [Bidding on Finite Fabs & Power Grids] ──► [Structural "AI Hardware Tax" on Consumers]
Because semiconductor fabrication capacity is prioritized for high-margin enterprise AI chips, traditional consumer RAM, desktop components, and enterprise storage face severe supply deficits. Consumers and ordinary businesses effectively pay an AI hardware tax through inflated electronics prices, subsidizing the data center rush.
It is the same mechanism as 2021, aimed at a different shelf. Too much money, chasing too few physical things.
The Thread Unbroken: Reshuffling as the Great Repricing
Looking back across thirty years, the pattern reveals something fundamental about modern capitalism:
Every major technological transformation in modern history is built on an unprofitable, debt-financed capital overbuild — which must ultimately be cleansed through repricing.
The 1990s telecom boom collapsed, repricing optical fiber down to cents on the dollar and handing a cheap backbone to the early internet pioneers.
The 2000s real estate crash wiped out trillions in paper wealth, repricing housing stock and forcing a decade of zero interest rates that funded the modern cloud era.
Today's AI hyperscaler boom is eating its own tail, spending trillions on GPUs, land, and power grids without a clear path to near-term software return.
The eventual crack in the circular financing loop will not destroy the data centers, the subsea cables, or the power substations. It will force a systemic reshuffling — which is simply repricing under another name. Unpayable debt gets restructured. Overvalued equities re-rate. Physical assets change hands at valuations that reflect what they actually produce rather than what they were financed at.
That is what a repricing is: the market discovering the difference between the cost of building something and the value of what it does.
Only when these inflated assets are repriced to match real-world utility will the genuinely productive wave of the AI era begin.
Which brings us to the ultimate question: when this circular capital loop inevitably breaks and this unprecedented mountain of compute infrastructure is finally repriced on the open market, who will buy the keys to the AI backbone for cents on the dollar — and what will they build once the cost of intelligence approaches zero?
Part 2: The Blast Radius
What a Mag-7 reversion does to a retirement portfolio — and the tax moves that only work while it's down.
One email when it publishes. No list-selling, ever.
The reference behind the Letter: THE TAX CUTTERY® Guide to Federal Income Taxation — 564 pages, nearly 2,000 primary-source citations. Join the launch waitlist →
The firm behind this analysis: THE TAX CUTTERY® — back taxes, IRS letters, planning, and returns. Start a conversation any time.