The Crash I See Coming: Why I’ve Liquidated My Portfolio
AI bubbles, stretched valuations, and fragile economies are colliding. I don’t know the trigger or the timing — but I see the fall ahead.
AI hype, overvalued markets, and global headwinds are converging. I’ve stepped aside — because the setup for a crash is already here..
Baskar Agneeswaran
Published
Sep 8, 2025
Categories
My Personal Decision
Last week, I liquidated my entire holdings in the market. Lock, stock and barrel. That’s not a decision I made lightly, and it’s not based on a single chart or a sudden gut feeling. It’s the result of watching AI, valuations, and macroeconomic signals converge into a picture that looks eerily familiar to past bubbles — only bigger.
I want to be clear: this isn’t financial advice. This is my personal perspective as someone who’s been deeply involved in technology, SaaS, AI, and the markets. I’ve seen hype cycles come and go, but what’s happening right now feels more dangerous — because the gap between perception and reality is widening at both the micro (AI company fundamentals) and macro (global markets) levels.
When you combine unsustainable AI economics, overvalued public markets, and the structural job displacement AI is about to unleash, the conclusion for me is unavoidable: we’re headed for a crash. The only question is how deep and how soon.
The AI Bubble
Every bubble has the same DNA: capital chasing a story, while ignoring fundamentals. In this cycle, that story is AI.
a. Broken Unit Economics
Let’s start with the basics: it is expensive — very expensive — to process AI requests. Training is costly, inference is costly, and the infrastructure doesn’t scale profitably at today’s price points. Large language models like ChatGPT are losing money at the user level.
That means the unit economics don’t work. Either prices for AI services must rise significantly (which risks slowing adoption), or the cost to serve must fall dramatically (which is nowhere close to reality yet). Until one of those happens, this is classic “growth at any cost” — the same playbook that killed many tech darlings in the past.
b. The AGI Mirage
Much of the frenzy is justified by the promise of AGI — Artificial General Intelligence. But let’s be real: AGI isn’t coming anytime soon. Philosophically, I don’t even think it’s possible in the way some dreamers describe it. If humans could create beings with true, independent intelligence, that would effectively make us God — and we are not.
What’s more, the incremental benefit of new models is already tapering. Each generation of LLMs delivers smaller and smaller leaps. Without a radically different approach, we’re not marching steadily toward AGI — we’re hitting diminishing returns. The billions of dollars chasing this dream are wildly disconnected from what’s actually achievable.
c. The Agentic AI Hype
Then there’s “agentic AI.” Simple agents can work — an AI SDR, an AI LinkedIn post-writer, basic support bots. But can these agents solve the complex workflow problems of Fortune 500 enterprises? Absolutely not. Enterprise workflows need orchestration, coordination, and governance — not just autonomous agents running in silos.
Even in the SMB space, where agents are more practical, supply already outpaces demand. Every week there’s a new wave AI agent startups pitching the same idea. Consolidation is inevitable, and most will not survive.
Taken together, these realities point to a harsh conclusion: over 90% of AI startups today will not exist in the next 2–3 years. The bubble will burst — and when it does, the fallout will spread far beyond AI companies themselves.
Macro Overvaluation Signals
It’s one thing to say markets are expensive. It’s another to look at the numbers and see just how far we’ve drifted from fundamentals. Across multiple valuation models, the message is consistent: we’re at unsustainable highs.
a. The Buffett Indicator (Market Cap-to-GDP)
The Buffett Indicator is simple in concept but powerful in what it reveals. It takes the total value of all publicly traded stocks (market capitalization) and compares it to the size of the real economy (GDP).
Think of it this way: if the stock market were a house, GDP is its foundation. If the house grows far bigger than the foundation can support, it becomes unstable.
Current reading (Sep 5, 2025): 214% (i.e. the stock market valuation is 2.14 times the GDP in the US). This is the highest ever that it has been recorded.
Historical average: ~100% (meaning the market value equals the economy’s output).
Dot-com bubble (2000): ~150% before the crash.
Global Financial Crisis (2008): plunged to ~50–60%.
Why it matters: When the Buffett Indicator rises above 100%, it suggests stocks are collectively priced higher than the economy can sustain. Today’s reading is far beyond both the dot-com and 2008 levels, flashing a clear signal of overvaluation.
Source: https://buffettindicator.net/
b. Price-to-Earnings (P/E) Ratio
The P/E ratio tells us how much investors are willing to pay for every dollar of company earnings. If a company earns $1 per share, and its stock trades at $20, the P/E ratio is 20.
High P/E ratios aren’t always bad — growth companies often trade at premiums. But when the entire market’s P/E drifts too far above its historical average, it signals a collective bet that future profits will be much higher. If those profits don’t materialize, valuations collapse.
Current reading (Sep 4, 2025) of PE of NASDAQ 100: 33.17.
Long-term average: ~15–16.
Dot-com peak (2000): soared above 40 before the crash.
Pre-2008 levels: in the high 20s before the financial crisis.
Why it matters: Today’s P/E ratio suggests investors are once again pricing in perfection — assuming earnings will keep climbing indefinitely. History shows that kind of optimism never lasts.
Source: https://worldperatio.com/index/nasdaq-100/
c. Price-to-Sales (P/S) Ratio
The P/S ratio looks at stock prices compared to company revenues. Unlike earnings, which can be massaged with accounting tricks, sales are straightforward — money in the door.
If the P/S ratio is high, it means investors are willing to pay a huge premium even before a company shows profitability. That’s a hallmark of speculative bubbles.
Current reading of P/S ratio of NASDAQ 100 (Sep 04, 2025): 6.7.
Long-term average: ~1.0.
Dot-com peak (2000): large caps traded at >2.0, some far higher.
Why it matters: Paying 2–3x sales might be fine for a breakout company with defensible growth. But when the entire market trades at those levels, it’s a sign of collective over-optimism detached from business reality.
Source: https://www.macrotrends.net/stocks/charts/NDAQ/nasdaq/price-sales
d. Mean Reversion (S&P 500 vs. Historical Trend)
Markets rise and fall, but over time, they follow a long-term growth trendline. “Mean reversion” measures how far current stock prices deviate from that long-term average.
Think of it like a rubber band: the further you stretch it, the stronger the snapback when it’s released.
Current deviation (Aug 2025): 187% above trend, standard deviation of +4 (highest ever recorded in history).
Historical precedent: Similar deviations occurred in 1929, 2000, and 2007 — each followed by sharp crashes that pulled valuations back toward the mean.
Why it matters: The bigger the gap between today’s prices and the historical trend, the harsher the eventual correction tends to be. And right now, that gap is wide.
e. The Sahm Rule (Recession Signal)
Unlike valuation ratios, the Sahm Rule is about the real economy. It looks at unemployment. Specifically, it triggers when the three-month average unemployment rate rises by 0.5 percentage points or more above its lowest level in the past year.
Why is this important? Because rising unemployment tends to ripple quickly through the economy — lowering spending, increasing defaults, and weakening business investment.
Current status (Aug, 2025): 0.13.
Historical record: Every time this rule has triggered since World War II, the U.S. has entered recession. There have been no false positives.
Why it matters: The Sahm Rule is the one major signal that hasn’t been tripped yet. But when it does, history says recession is already underway. Keep watching this metric — the day it crosses 0.5, the inevitable won’t be theoretical anymore.
Source: https://fred.stlouisfed.org/series/SAHMCURRENT
The Big Picture
Each of these indicators tells the same story: stocks are not just priced for perfection — they’re outright overpriced. The Buffett Indicator, P/E, P/S and mean reversion have all flashed red before past crashes, but rarely all at the same time. Today, they’re aligned. Valuations have drifted far beyond what the economy, earnings, or revenues can realistically support. History shows that when prices and fundamentals diverge this much, the correction isn’t optional — it’s inevitable.
Structural AI-Driven Job Losses
Even if 90% of AI companies collapse in the coming years, the technology itself isn’t going away. In fact, AI is already better than humans in multiple domains — and that’s where the real economic disruption lies.
Think of Stockfish, the open-source chess engine with an Elo rating above 3400. The best human player in the world, Magnus Carlsen, is rated around 2800. Stockfish would crush him like an ant. It’s not a contest anymore. The machine simply plays in a different league.
That same pattern is repeating across industries. AI systems are already outperforming humans in areas such as:
Coding: AI-generated code is often faster and more accurate than human output. At Vajro, we’ve set a target to move toward 100% AI-written code by the end of 2025. I personally used ChatGPT to generate Google Apps Script for the AI Maturity Model, automating workflows that would have otherwise required developers.
Customer Support & Back-Office Work: Virtual agents can already handle the bulk of tickets, insurance claims, accounting entries, and document reviews. In many cases, human involvement is only for exceptions.
Content Creation: From LinkedIn posts to long-form books, AI can replicate tone, structure, and depth. Trained correctly, it can write in a voice indistinguishable from its human author.
Marketing & Sales: Agentic AI may fail at running enterprise workflows, but in narrow tasks like outreach, copywriting, and lead scoring, it scales faster than any human team.
Design: Tools like MidJourney and Figma AI can produce polished graphics in seconds, reducing the need for large design departments.
This is the paradox: most AI startups will die, but AI itself will live on — and thrive. The implication for the global economy is profound. Hundreds of millions of jobs in IT, services, back-office operations, and creative industries are at risk of redundancy.
The technology is real. The disruption is inevitable. And the transition will be brutal.
And this is where the link to the Sahm Rule becomes critical. The Sahm Rule triggers when unemployment rises 0.5% above its recent low. AI-driven displacement could be the catalyst. As AI eats into coding, support, content, and back-office jobs, unemployment will tick upward. At first slowly, then suddenly. When it crosses that threshold, the Sahm Rule won’t be theoretical anymore — it will confirm what we already know: recession has arrived.
Other Headwinds
The AI bubble and market overvaluation would be concerning enough on their own. But they’re unfolding against a backdrop of additional pressures that amplify the risk of a hard landing.
a. Interest Rates and Debt
After years of near-zero rates, the world is now facing sustained higher interest rates. That matters because global debt has never been higher. Governments, corporations, and households all carry record levels of leverage. Refinancing this mountain of debt at today’s rates will strain balance sheets and budgets everywhere.
b. China’s Structural Slowdown
For decades, China has been the growth engine of the world. But its property market — once the cornerstone of its economy — is in deep trouble. Demand is slowing, real estate developers are defaulting, and consumer confidence is weakening. A weaker China doesn’t just hurt Asia; it sends ripples through global supply chains and commodity markets.
c. Geopolitics and Protectionism
Wars in Europe and the Middle East, rising U.S.–China tensions, and the resurgence of trade protectionism are reshaping global supply chains. These shifts are inflationary, adding cost pressures at a time when economies can least afford them.
d. Climate and Insurance Costs
Climate shocks are no longer rare events. Hurricanes, floods, and wildfires are becoming more frequent and more destructive. Insurance companies are already pulling out of high-risk markets, driving costs higher for businesses and households alike. Climate-related losses are turning into a structural economic drag.
e. India’s IT and Services Vulnerability
India has been celebrated as one of the fastest-growing major economies, with much of that growth powered by IT services exports. But AI threatens the very foundation of that model. The tasks historically outsourced to India — coding, back-office processes, customer support — are precisely the tasks AI is beginning to do faster and cheaper.
Large IT firms are already freezing hiring, and whispers of deeper disruption are spreading. If AI replaces even a fraction of the millions of jobs tied to India’s services boom, the ripple effects on income, consumption, and growth could destabilize one of the world’s most important emerging markets.
The Combined Weight
High debt, rising rates, a slowing China, unstable geopolitics, mounting climate costs, and structural risks in India’s services economy — all of these forces compound the risks posed by AI and market overvaluation. When the winds are blowing this hard from so many directions, it doesn’t take much to tip the system over.
Conclusion – Why I’m on the Sidelines
I’ve liquidated my holdings not because I know exactly when the crash will come, but because I believe the setup for one is already in place.
Every major downturn in history has had a trigger:
The dot-com bust unraveled when investor faith in profitless tech stocks evaporated.
The 2008 financial crisis erupted when mortgage-backed securities collapsed.
The COVID crash was sparked by a sudden global shutdown.
In each case, the underlying imbalances were already there. The trigger was just the match that lit the fire.
This time, I don’t know what that trigger will be. It could be a sharp rise in defaults from overleveraged corporations. It could be an AI unicorn imploding and shaking faith in the sector. It could be a geopolitical event, a climate shock, or something entirely unexpected.
And I don’t know when it will happen. Markets can always rise higher before they fall — bubbles are notorious for lasting longer than anyone predicts. But history shows that when fundamentals and valuations diverge this dramatically, the eventual correction is not a question of if. It’s only a question of when.
That’s why I’ve chosen to step aside. I’d rather miss a little more upside than risk being fully exposed when the downward spiral begins. The signals are too strong, the risks too broad, and the complacency too high.
I’m not here to tell you what to do with your money. But I am here to say this: ignoring the signals has never ended well. The AI bubble, stretched valuations, structural job losses, and global headwinds aren’t abstract theories — they’re all flashing red lights on the dashboard. You don’t have to agree with my decision to liquidate, but at the very least, you should be asking yourself the hard question: what happens if the crash comes sooner than you think?
