An Open Letter to Anyone Still Drinking the Kool-Aid
I heard a story many years ago that is more poignant now than ever. A young man, twenty-six years of age in the spring of 2000, proudly announced that his entire 401(k) was tied up in a company that sold pet food over the internet. The mascot was a sock puppet. The stock price was $1.20. Six months later, no one was laughing. Half of the investment group he was part of saw their retirement accounts evaporate into the same thin air as that sock puppet’s promises.
I bring this up not for nostalgia, but because here we are, a quarter-century later, and the only thing that’s changed is the mascot. The sock puppet is now a chatbot, and the stakes have an extra zero—or three.
I’m not a doomsayer by nature. I’ve seen enough real revolutions—the internet did change everything—to know that bubbles and breakthroughs aren’t mutually exclusive. The dot-com crash didn’t negate the fact that the internet was transformative; it just proved that hype and reality don’t always align on the same timeline. But what I’m seeing now in the AI gold rush isn’t just hype. It’s hype with a side of environmental collapse, financial recklessness, and a willful ignorance of history’s lessons.
So let’s talk about the money, because the money is the tell.
JULY, 2026
The Cash Burn: A Fire Sale of Common Sense. The five biggest cloud companies—Microsoft, Alphabet, Amazon, Meta, and Oracle—are on pace to spend $660 to $690 billion this year alone on AI infrastructure. To put that in perspective, that’s more than the GDP of most countries. McKinsey projects the global data center buildout will hit $6.7 trillion by 2030. Six point seven trillion. I’ve tried to find a comparable historical figure that isn’t a war, and I can’t. The closest might be the transcontinental railroad—except the railroad actually connected things. This? This is just digging holes in the ground and filling them with servers.
And yet, the losses are staggering. OpenAI, the company that kicked off this arms race, burned through $3.7 billion in the first three months of this year alone—more than half of what it brought in during the same period. Internal projections put its 2026 net loss at $14 billion, with cumulative losses potentially hitting $115 billion by 2030. Let that sink in. Uber, a company famous for hemorrhaging cash, took a decade to lose $25 billion. OpenAI is on track to lose more than that in a single year.
But here’s what really frightens me: It’s not the losses. It’s the circularity.
Microsoft invests in OpenAI. OpenAI spends that money buying cloud computing from Microsoft. NVIDIA has committed $100 billion to OpenAI—money that OpenAI’s own CFO has admitted flows right back to NVIDIA in the form of chip purchases. NVIDIA, meanwhile, is a major investor in CoreWeave, a cloud company that exists largely to buy NVIDIA chips and rent them back out to companies like… OpenAI. Everyone is each other’s customer. Everyone is each other’s investor. It’s a Jenga tower of mutual dependency, and when one piece wobbles, the whole thing comes crashing down.
On January 29 of this year, Microsoft’s stock fell 12% in a single trading session—$440 billion, gone before lunch—after the company disclosed that nearly half of its cloud backlog depends on the fortunes of one single customer. If that doesn’t alarm you, I don’t know what will. You don’t need an economics degree to see the problem here. You just need to have seen a house of cards before.
THE INFRASTRUCTURE NIGHTMARE: POWER, WATER, AND THE GRID’S LAST GASPS
If the financial side of this bubble is a dumpster fire, the physical infrastructure is a five-alarm blaze.
The Electricity Crisis. Data centers already consume 4-6% of all U.S. electricity, and that number is projected to hit 10-15% by 2030. A single data center can use as much power as a small city. Training a large AI model consumes more electricity than hundreds U.S. homes do in a year.
Dominion Energy in Virginia has warned that data center growth could outpace its ability to supply power by 2028. Texas has paused new data center approvals in some areas because the grid can’t handle the load. What happens when the grid can’t keep up? Blackouts. Brownouts. Skyrocketing utility bills. In Northern Virginia, residential electricity costs have risen by 267% since data centers moved in. And that’s just the beginning.
The Water Wars
If the electricity problem is bad, the water problem is apocalyptic. Data centers need 1-5 million gallons of water per day for cooling. In Austin, Texas, data centers used 1.5 billion gallons in 2023—enough to fill 2,200 Olympic swimming pools. In Arizona, where the Colorado River is already over-allocated and shrinking, tech giants like Meta, Google, and Microsoft are building massive data centers in Mesa and Phoenix. The state is in a 20+ year megadrought, and aquifers are being depleted faster than they can recharge. Some are expected to run dry in 20-50 years.
Where’s the water coming from?
- Groundwater: Data centers drill their own wells and take as much as they want, even if it harms neighbors or the environment. In Texas, the “Rule of Capture” means landowners have near-unlimited rights to pump groundwater.
- Municipal supplies: Data centers buy water from cities, competing with residents and farmers. In Round Rock, Texas, data centers now account for 20% of the city’s water use.
- Reclaimed wastewater: Some data centers use treated sewage water, but it’s limited, corrosive, and controversial.
And the “solutions”?
- Air cooling? Less efficient in extreme heat and uses more electricity.
- Immersion cooling? Expensive to retrofit and years away from widespread adoption.
- Building in colder climates? Latency issues make it unworkable for many applications.
The hard truth? We’re building data centers in deserts and hoping for a miracle.
THE REGULATORY AND LEGAL STORM BREWING
Generative AI relies on scraping vast amounts of data—often without permission. The lawsuits are piling up:
- Getty Images vs. Stability AI: Getty is suing for copyright infringement, arguing that Stability AI illegally used its images to train its models.
- Authors, artists, and musicians are fighting back against AI companies using their work without compensation.
- The EU AI Act (2024) is slapping companies with compliance costs, and U.S. regulators are starting to take notice.
What’s the endgame? Fines. Bans. Forced changes to AI models. And that’s just the beginning.
Then there’s antitrust. Microsoft, Google, and NVIDIA are dominating the AI space, and regulators are taking notice. The DOJ and FTC are already investigating whether these companies are stifling competition. If history is any guide (Standard Oil, AT&T, Microsoft in the 90s), breakups or heavy regulations could be on the horizon.
THE CONSUMER AND ENTERPRISE BACKLASH
AI was supposed to revolutionize productivity. Instead, most companies aren’t seeing the gains. Why?
- AI is still clunky. Chatbots hallucinate. AI art is generic. Most “AI features” are gimmicks.
- Integration is hard. Companies are spending millions on AI pilots, but most fail to deliver ROI.
- Employees aren’t using it. Microsoft’s Copilot was supposed to revolutionize Office. Instead, most users ignore it.
Gartner predicts that by 2027, 50% of AI projects will be abandoned.
And consumers? They’re getting bored. After the initial novelty, chatbot usage is plateauing. Most people don’t use AI daily. And why would they? It’s not solving real problems—it’s just a shiny new toy.
THE BUBBLE’S TICKING CLOCK
No one knows the exact timing, but the triggers are clear:
| Trigger | Timeframe | Impact |
| Major AI company collapses | 2026-2027 | Investor panic → funding dries up → mass layoffs |
| NVIDIA’s stock crashes | 2026-2027 | GPU prices drop → data center investments stall → AI growth halts. |
| A black swan event (regulatory crackdown, water/energy crisis) | 2025-2028 | Public backlash → governments step in → AI adoption slows to a trickle. |
| The “Trough of Disillusionment” | 2026-2028 | Hype fades → only the strongest survive → consolidation / bankruptcies. |
Most likely scenario:
- 2026-2027: First wave of AI startups collapse (like Pets.com in 2000).
- 2027-2028: Big Tech scales back AI investments (like Cisco did after the dot-com crash).
- 2029+: AI enters the “Trough of Disillusionment”—hype dies, but real, profitable use cases emerge slowly.
THE WARNING: WHAT HAPPENS IF WE DON’T ACT?
1. Environmental Collapse
- Water shortages: Aquifers in Texas and Arizona could run dry in 20-50 years. Farmers will lose crops. Cities will ration water. Ecosystems will collapse.
- Energy crises: Data centers could consume 15% of U.S. electricity by 2030. Blackouts. Brownouts. Skyrocketing bills.
- Climate feedback loop: Data centers emit CO2 and deplete water—accelerating climate change, which worsens droughts, which makes the water crisis worse.
2. Financial Implosion
- AI stocks crash: NVIDIA, Microsoft, Google—all trading at all-time highs—could lose 50-70% of their value if AI demand slows.
- Data centers become stranded assets: If water or power runs out, these multi-billion-dollar facilities could become worthless overnight.
- Economic fallout: Job losses. Bankruptcies. A tech industry in freefall.
3. Social Unrest
- Water wars: Farmers vs. tech companies. States vs. states. Tribes vs. governments. Lawsuits. Protests. Violence.
- Energy rationing: Governments forcing data centers to cut power during heatwaves. Digital economy grinds to a halt.
- Public backlash: AI bans. Regulations. A loss of trust in Big Tech.
THE FINAL WARNING: THE PIPER WILL BE PAID
I’m not telling you to sell everything and buy gold. But I am telling you this: The structure holding this boom together—a small number of customers, a web of circular investment, optimistic depreciation assumptions, and hundreds of billions in off-balance-sheet lease commitments—is exactly the kind of structure that turns an ordinary slowdown into a disorderly one.
It didn’t happen to Cisco until 18 months after the warnings started sounding foolish. But I remember 1999 well enough to know that the people who sounded foolish in March were, by the following spring, the only ones who still had a portfolio worth discussing.
So do your own homework. Ask what happens to the company you’re invested in if its three biggest customers are also its three biggest investors, and one of them has a bad quarter. Ask what happens when the water runs out or the grid collapses. Ask what happens when the hype fades and the bills come due.
Because the piper is coming. And this time, he’s bringing a water bill, an electricity invoice, and a GPU receipt.
History doesn’t repeat, but it rhymes. And right now, AI is rhyming with 1999—but with more zeros, more desperation, and higher stakes. The question is: Will we act in time to soften the blow? Or will we let it hit us full force? The clock is ticking. And the bill is coming due.