Is the AI Bubble About to Burst? How to Profit Either Way

# Is the AI Bubble About to Burst? How to Profit Either Way (2026)

Nvidia trades at a P/E of 47.7. Microsoft just committed $13 billion to OpenAI. Palantir somehow commands a P/E north of 200. The entire AI sector added roughly $8 trillion in market cap since ChatGPT launched in November 2022 — more than the GDP of Japan and Germany combined.

I've been investing through two major bubbles — dot-com and crypto 2021 — and I'm seeing patterns that make the hair on my neck stand up. But here's the thing: I'm not going to tell you whether AI is a bubble or not. Smarter people than me have been wrong about timing. What I *am* going to do is give you a concrete playbook for profiting regardless of which direction this goes — and a 15-minute emergency kit you can set up today so you're not scrambling when the market moves.

Because the worst position in any bubble isn't being wrong about direction. It's being right about direction but unable to act fast enough.

---

Is AI Really a Bubble? The Numbers Don't Lie

Let me lay out both sides honestly, because the data tells a more nuanced story than most pundits admit.

The bear case is compelling. Sam Altman himself acknowledged in a February 2026 interview that "parts of AI are probably overvalued." Ray Dalio has publicly compared the current environment to the dot-com era, saying the parallels are "uncomfortably similar." The Buffett Indicator — total US market cap divided by GDP — currently sits above 190%, well past the 140% level Buffett once called "playing with fire." During the dot-com peak, it hit roughly 148%. We've blown past that.

The concentration is extreme. The "Magnificent Seven" tech stocks account for over 30% of the S&P 500's total market capitalization. That's more concentrated than any point during the dot-com bubble. Palantir trades at 200+ times earnings on the thesis that government AI contracts will eventually justify the valuation. C3.ai loses money every quarter but trades at a multi-billion dollar market cap because "AI" is in the name. Sound familiar? In 1999, companies added ".com" to their names and watched their stock prices double overnight. Today, slap "AI-powered" on a press release and the same magic happens.

But the bull case isn't stupid either. Nvidia's P/E of 47.7 looks nothing like Cisco's 100-150x during the dot-com peak. Nvidia is actually delivering massive revenue growth — $130 billion in trailing twelve-month revenue, with data center demand still outstripping supply. Microsoft's $13 billion OpenAI investment is generating real returns: Azure AI revenue grew 150%+ year-over-year. These aren't Pets.com burning cash on Super Bowl ads. These are real businesses with real cash flows.

The core technology works. AI is already reducing costs in customer service, drug discovery, legal research, and software development. McKinsey estimates AI could add $4.4 trillion in annual value to the global economy. That's not hype — that's measured by one of the most conservative consulting firms on the planet.

My read? The technology is real. The valuations on *some* companies are insane. This looks like 1999 — not because the internet was fake (it obviously wasn't), but because investors couldn't distinguish between Amazon and Webvan. Both were "internet companies." One was worth $1.7 trillion two decades later. The other went bankrupt. The same sorting is coming for AI.

The question isn't whether AI is a bubble. The question is: *which AI stocks are the bubble, and which are the real thing?* And more importantly: are you positioned to profit from the answer?

---

Why Most People Can't Act When Bubbles Pop

Here's something nobody talks about: even people who *correctly predicted* the dot-com crash mostly didn't profit from it. Why? Because acting on a bubble pop requires infrastructure that most people don't have set up in advance.

The 2000 crash moved fast. Between March 10 and March 15, 2000, the Nasdaq dropped 9% in a single week. By April 14, it had crashed 34% from its peak. Ordinary investors — people with 401(k)s and Schwab accounts — watched their portfolios evaporate in real-time. Even those who saw it coming couldn't do much: selling individual stocks took T+3 settlement. Short selling required margin accounts most people didn't have. By the time they repositioned, the worst damage was done.

The total destruction was staggering: $5 trillion in market value evaporated between March 2000 and October 2002. The average tech investor lost 78% of their portfolio value. Many never recovered emotionally and stayed out of the market for years, missing the eventual recovery.

Traditional short selling has absurd barriers. To short a stock through a US broker, you need: a margin account (separate application, 2-5 business day approval), a minimum of $25,000 in equity (FINRA pattern day trader rule), the ability to borrow shares (popular shorts like Nvidia can have borrow fees of 5-15% annually), and you need to be a US resident or have a US brokerage relationship. If you're reading this from Manila, São Paulo, or Lagos? You're essentially locked out of the traditional short-selling market entirely.

Timing is everything, and setup takes forever. Let's say you wake up tomorrow to headlines that Nvidia's largest customer just cancelled $10 billion in chip orders. The stock is gapping down 15% in pre-market. You know — with every fiber of your being — that this is the start of a larger crash. What do you do?

If you don't already have a funded account with short-selling capability, the honest answer is: nothing. You watch. You open a brokerage application. You wait 3-5 business days for approval. You initiate a wire transfer that takes another 1-3 days. By the time you can actually execute a trade, the move has already happened, the VIX has spiked, and you're either chasing or the opportunity has passed.

This is the problem I want to solve in the rest of this article. Not predicting when the bubble pops — but making sure you can act within 15 minutes when it does.

---

Scenario A: AI Keeps Rising — How to Ride the Wave

Let's start with the bullish scenario, because despite everything I just said about bubble risks, there's a genuine possibility AI stocks keep running for another 12-24 months. Bubbles can stay irrational far longer than your short position can stay solvent — just ask anyone who shorted Tesla in 2020.

If AI continues its trajectory, you want exposure to the companies with real revenue and defensible positions. The obvious picks:

Nvidia (NVDA) — Controls 80%+ of the AI training chip market. As long as companies keep building AI models, they need Nvidia hardware. The P/E of 47.7 is elevated but not insane for a company growing revenue at 90%+ annually.

Microsoft (MSFT) — The OpenAI partnership gives them an actual AI monetization engine through Azure and Copilot. Unlike pure-play AI companies, Microsoft has a fortress balance sheet and diversified revenue.

Apple (AAPL) — The AI-on-device play. If AI moves from cloud to edge (running locally on phones and laptops), Apple's hardware ecosystem becomes the distribution channel.

The problem for most international investors is *access*. You can't buy US stocks from most countries without dealing with minimum deposits, currency conversion fees, and settlement delays. This is where stock tokens on platforms like OKX come in — they let you buy fractional exposure to these stocks starting from $1, settle instantly in USDT, and are available in 100+ countries. No wire transfers, no currency conversion headaches, no $500 minimum positions.

I'm not going to walk through a full buying tutorial here — I've already written detailed guides for buying Tesla and other US stocks through stock tokens. The short version: register, complete KYC (takes about 5 minutes with automated verification), fund with USDT via P2P or direct deposit, and buy. Total process: under 15 minutes.

The key advantage in the bullish scenario is fractional ownership. You don't need $900+ to buy a single Nvidia share. You can buy $50 worth, or $10 worth, and add to the position gradually. If AI keeps running, you participate. If it doesn't, your downside is limited to what you put in — no leverage, no liquidation risk.

---

Scenario B: The Bubble Pops — How to Profit in 15 Minutes

This is the scenario most people are unprepared for, and the one where speed matters most. If the AI bubble bursts — a major customer cancellation, a regulatory crackdown on AI training data, a sudden investor panic — prices will move 20-40% in days, not weeks. You need to already be set up.

Why OKX perpetual contracts are the fastest way to short AI stocks

Traditional short selling and crypto-based shorting via perpetual contracts aren't even in the same universe when it comes to speed. Here's the comparison:

How to short AI stocks on OKX — step by step:

Step 1: Register with referral code BUYSTOCK. Go to OKX and create your account. This takes about 2 minutes. The referral code gets you reduced trading fees, which matters because you'll be paying funding rates on perpetual contracts (more on that below).

Step 2: Complete KYC. OKX uses automated identity verification. Upload your ID, take a selfie, and you're typically verified within 5 minutes. This is required for derivatives trading.

Step 3: Buy USDT. Go to the P2P marketplace or use a direct deposit method. P2P lets you buy USDT from local sellers in your own currency — bank transfer, mobile money, whatever works in your country. Buy at least $100 to have working capital ready. This is *not* money you need to trade today; it's your emergency fund for when the market moves.

Step 4: Navigate to perpetual contracts. Find NVDA-USDT-PERP (or whichever stock token perpetual contract you want to short). The interface shows the order book, chart, and order entry panel.

Step 5: Open a short position. Select "Sell/Short." Choose your leverage — I'd recommend starting with 2-3x maximum. Enter your position size. Hit confirm. You're now short Nvidia.

Step 6: Set a stop-loss immediately. This is non-negotiable. If you're shorting NVDA at $120, set a stop-loss at $132 (10% above entry). Shorting has theoretically unlimited loss potential — your stop-loss is the only thing preventing catastrophic damage. I've written an entire guide on avoiding liquidation with stock tokens that you should read before placing any leveraged trade.

The cost breakdown for shorting $1,000 of Nvidia:

Let's say Nvidia trades at $120 per token and you want $1,000 of short exposure.

• Position size: $1,000 notional value


• At 3x leverage, you need: ~$333 in margin (USDT collateral)


• Opening fee: 0.05% = $0.50


• Funding rate: varies, typically 0.01-0.03% every 8 hours (you'll *receive* funding when shorting a premium asset — see my funding rate explainer)


• Closing fee: 0.05% = $0.50


• Total round-trip cost: ~$1.00 + funding rate (which may actually pay *you*)

---

Scenario C: The Smart Hedge — Long AI Revenue, Short AI Hype

This is the strategy I personally find most interesting, because it doesn't require you to predict direction. It requires you to predict *quality* — which AI companies are real and which are hot air.

The premise is simple: not all AI stocks are created equal. Microsoft generates billions in actual AI revenue. C3.ai generates losses. Both are labeled "AI stocks." If you go long the real ones and short the fake ones, you profit regardless of market direction:

• If AI keeps rising: The real companies (Microsoft, Nvidia) rise faster than the hype companies (C3.ai, SoundHound). Your long gains exceed your short losses.


• If AI crashes: Everything falls, but the hype companies fall *harder* and *faster* because there's no floor of real revenue underneath them. Your short gains exceed your long losses.


• If the market stays flat: You still make money if the quality gap between real and fake AI companies widens, which it tends to do over time as earnings reports separate fact from fiction.

Pair 1: Long Microsoft / Short C3.ai
Microsoft's AI revenue through Azure and Copilot is real, growing, and diversified. C3.ai has been around since 2009, rebranded as an "AI company," and still hasn't achieved consistent profitability. If AI sentiment turns negative, Microsoft drops 20% and C3.ai drops 60%. If AI sentiment stays positive, Microsoft grinds up 30% and C3.ai might match it — but probably won't, because the next earnings miss will crush it.

Pair 2: Long Nvidia / Short AMD (smaller position)
This is more nuanced. AMD isn't a bad company — they make real products with real revenue. But in the AI chip race, they're a distant second. Nvidia owns 80%+ market share in AI training. AMD's MI300 chips are decent but haven't dented Nvidia's dominance. If AI spending contracts, companies cut the #2 supplier first. If AI spending expands, Nvidia captures disproportionate share. Either way, the gap between them widens.

Pair 3: Long Big Tech AI / Short Pure-Play AI SPACs
Several AI companies went public via SPAC mergers in 2023-2024 with minimal revenue and astronomical valuations. These are the modern equivalents of dot-com era companies that IPO'd with a business plan and a prayer. I won't name them all, but look at any AI company trading at 50x+ revenue with declining growth rates.

Using the Buffett Indicator as your throttle. This is where the Buffett Indicator tool becomes genuinely useful beyond academic interest. Here's my framework:

• Buffett Indicator below 150%: Normal market, run pairs at 1:1 ratio (equal long and short exposure)


• Buffett Indicator 150-180%: Elevated, shift to 1:1.5 ratio (50% more short exposure than long)


• Buffett Indicator above 200%: Danger zone, shift to 1:2 ratio (double the short exposure)

---

The 5 AI Stocks Most Likely to Crash (and How to Short Each)

Let me be blunt: predicting *which* stocks will crash is harder than predicting *that* some stocks will crash. But based on valuation, revenue quality, and competitive positioning, these five stand out as the most vulnerable if AI sentiment reverses.

1. Nvidia (NVDA-USDT-PERP)

The elephant in the room. Nvidia is simultaneously the *best* AI company and one of the most *dangerous* to own at current prices. The bull case is undeniable: they own the AI chip market and are growing revenue at rates that make the valuation arguably reasonable. The bear case is equally undeniable: they're priced for perfection. Any miss — a quarter of slowing growth, a competitor breakthrough, export restrictions to China tightening further — and the stock drops 30%+ overnight. Nvidia is the single largest contributor to the S&P 500's AI premium. If AI sentiment turns, Nvidia leads the way down.

*How to short:* NVDA-USDT-PERP on OKX. Tight stop-loss (8-10% above entry) because this stock can squeeze violently on positive news. Start small — 2x leverage maximum.

2. Tesla (TSLA-USDT-PERP)

Tesla's inclusion in "AI stocks" is itself a sign of froth. Yes, they have autonomous driving technology and a humanoid robot program. No, neither generates meaningful revenue today. Tesla trades at 80x+ earnings on the promise that it will become an AI/robotics company rather than a car company. Meanwhile, the actual car business faces margin compression from Chinese competitors, price cuts in major markets, and brand damage from Elon Musk's political activities. If the AI narrative cracks, Tesla's AI premium (which accounts for roughly half its market cap by most analyst estimates) evaporates. I wrote a deeper analysis on whether Tesla is worth buying in 2026.

*How to short:* TSLA-USDT-PERP. Be warned — Tesla is the most volatile large-cap stock in the market. Use 2x leverage maximum and a 12-15% stop-loss. Musk's tweets alone can move the stock 5% in minutes.

3. Palantir (PLTR)

Palantir is a real company that does real things for government and enterprise clients. It's also trading at a P/E above 200, which prices in roughly a decade of perfect execution. The thesis: AI will multiply Palantir's government contracts and make its AIP platform the standard for enterprise AI. The risk: government contracts are lumpy, budget-dependent, and subject to political shifts. Enterprise AI adoption is real but competitive — Microsoft, Salesforce, and Google are all building similar capabilities into existing platforms that enterprises already use. At 200x earnings, Palantir needs everything to go right for ten consecutive years. How often does that happen?

*How to short:* Palantir doesn't always have a direct perpetual contract on OKX — check availability. If unavailable, use it as a signal: when Palantir cracks, the broader AI trade is cracking too, and you should be scaling into your Nvidia/Tesla shorts.

4. AMD (AMD-USDT-PERP)

AMD is the "other AI chip company," and that's precisely the problem. In a growing market, being #2 is fine — you still grow. In a contracting market, being #2 is devastating. Customers cut the marginal supplier first. AMD's MI300 chips have been well-received but haven't taken significant share from Nvidia. If AI capex spending plateaus or declines, AMD's AI revenue (which is driving most of its stock price appreciation) gets cut first. The stock has already run from $90 to $160+ on AI hype. A reversion to pre-AI-hype levels would mean a 40%+ decline.

*How to short:* AMD-USDT-PERP. This is actually one of the cleaner shorts because AMD doesn't have the same cult following as Nvidia or Tesla — fewer short squeezes, more rational price action. 3x leverage is manageable here with a 10% stop-loss.

5. C3.ai (AI)

The ticker symbol tells you everything you need to know. C3.ai literally trades under the ticker "AI" — and that's been its primary investment thesis. The company was founded in 2009 as C3 Energy, pivoted to C3 IoT, and then pivoted again to C3.ai when artificial intelligence became the hot narrative. Revenue growth has decelerated. Operating losses continue. Customer concentration is high. The stock trades at 15x+ revenue for a company that isn't growing particularly fast and isn't profitable. In any AI downturn, C3.ai is the canary in the coal mine — it'll crash first and hardest because there's nothing underneath the AI label except hope.

*How to short:* Check OKX for availability. If not directly available, this stock is worth tracking as a sentiment indicator. When C3.ai breaks below key support levels, it's a signal to add to shorts on the bigger names that *are* available as perpetual contracts.

---

Your 15-Minute Emergency Kit: Set Up Now, Act Later

Everything above is useless if you can't execute when the moment comes. So here's what I want you to do today — not tomorrow, not "when things look shaky," but today. It takes 15 minutes and costs $100.

Step 1: Register on OKX (2 minutes)

Go to OKX and create your account. Use referral code BUYSTOCK — it gives you reduced trading fees, which directly increases your profit on every trade. All you need is an email address and a password. Two minutes.

Step 2: Complete KYC verification (5 minutes)

Navigate to identity verification. Upload a photo of your government ID (passport, national ID, or driver's license). Take a selfie. OKX's automated system typically verifies you within minutes, not days. This step is required for derivatives trading — you can't access perpetual contracts without it. Five minutes.

Step 3: Buy $100 USDT and leave it (10 minutes)

Go to the P2P marketplace. Find a seller in your local currency. Buy $100 worth of USDT. Transfer it to your trading account (not your funding account — you want it ready for derivatives trading). This $100 isn't a trade recommendation. It's an insurance premium. It sits there, costing you nothing, until the day you need it. Think of it as a fire extinguisher: you hope you never use it, but you're damn glad it's there when the kitchen catches fire.

If you want more context on how to short US stocks with USDT, I've written a detailed tutorial that walks through the entire process with screenshots.

Step 4: Bookmark these resources

Save this article somewhere you can find it fast. Bookmark the Buffett Indicator tool — check it weekly. When it crosses 200%, start paying close attention. When financial news starts using phrases like "this time is different" and "new paradigm" with increasing frequency, reread the shorting section of this article.

Step 5: Set a price alert

On OKX, you can set price alerts for NVDA-USDT-PERP. Set one at 15% below the current price. If Nvidia drops 15% from wherever it is when you read this, that alert will wake you up and remind you that the short playbook exists.

Here's the core message: you don't need to trade today. Maybe AI stocks keep rising for another year. Maybe two. Maybe I'm completely wrong and Nvidia goes to $200 before any correction. That's fine. The entire point of this emergency kit is *optionality*. You're spending 15 minutes and $100 to buy the option to act in minutes when others need days. In a crash, that speed differential is worth thousands.

The best time to set up your account was six months ago. The second best time is now. The worst time is the morning after a crash, when everyone is trying to register simultaneously and systems are overloaded.

---

⚠️ Risk Warning

Read this section carefully. It is not a formality.

Short selling and leveraged trading can result in losses exceeding your initial investment. When you short a stock, your theoretical maximum loss is unlimited — a stock can rise infinitely, and you owe the difference. A $1,000 short position on a stock that doubles costs you $1,000. On a stock that triples, you lose $2,000. With leverage, these losses multiply.

Perpetual contracts carry specific risks. Funding rates can swing against your position. In strongly trending markets, holding a short perpetual contract means paying funding to long holders — this can erode your position even if the price stays flat. Liquidation can happen suddenly if price moves against you and your margin is insufficient.

Past patterns do not guarantee future outcomes. Just because the dot-com bubble burst doesn't mean the AI bubble will burst on the same timeline, in the same way, or at all. AI may be fundamentally different from the internet in ways that make historical comparisons misleading.

These are not investment recommendations. I'm describing strategies and tools. I am not a licensed financial advisor. I don't know your financial situation, risk tolerance, or investment goals. Everything in this article is educational content, not personalized investment advice.

Never trade with money you cannot afford to lose entirely. The $100 emergency kit suggestion assumes you can comfortably lose $100 without affecting your rent, food, or essential expenses. If you can't, do not do it.

Start small if you trade at all. If you've never shorted a stock before, your first trade should be tiny — $10-$20 of exposure — just to understand the mechanics. Learn on small money, not large money. The market will still be there when you're ready for bigger positions.

---

Frequently Asked Questions

Is the AI bubble going to burst in 2026?

Nobody knows with certainty, and anyone who claims otherwise is selling you something. What I can tell you is that multiple indicators — the Buffett Indicator above 190%, P/E ratios on pure-play AI companies exceeding 200, and legendary investors like Ray Dalio drawing dot-com comparisons — suggest the market is in elevated territory. That doesn't mean a crash happens tomorrow or even this year. The dot-com bubble started inflating in 1995 and didn't pop until 2000. Five years of "this is clearly a bubble" before it actually burst. The productive approach isn't predicting the exact date — it's being positioned to profit when a correction happens, whenever that may be, while also profiting if it doesn't. That's what the hedging strategies in this article are designed to do.

How do I short Nvidia without a US broker?

The most accessible method for international investors is through USDT-settled perpetual contracts on platforms like OKX. The contract ticker is NVDA-USDT-PERP. You don't need a US brokerage account, a margin account, or the ability to borrow actual shares. You're trading a derivative that tracks Nvidia's price, settled in USDT (a stablecoin pegged to the US dollar). Registration takes about 2 minutes at OKX with the referral code BUYSTOCK for reduced fees. After KYC verification (typically 5 minutes with automated review), you can fund via P2P in your local currency, convert to USDT, and open a short position immediately. The entire process takes roughly 15 minutes from scratch. Note that leverage is available but should be used conservatively — I recommend 2-3x maximum for stock token perpetual contracts.

What is the minimum amount to short AI stocks on OKX?

The minimum position size varies by contract, but for most stock token perpetual contracts on OKX, you can open a position with as little as $10-$20 in margin (collateral). With 3x leverage, a $10 margin lets you short approximately $30 worth of a stock. This is dramatically different from traditional short selling, which typically requires a $25,000 minimum margin account balance in the US. The low minimum makes it practical to experiment with small positions while learning how perpetual contracts work before committing real capital. I strongly recommend starting with the minimum possible position size on your first trade, purely as a learning exercise.

Can I short AI stocks from India, the Philippines, or Brazil?

OKX is available in 100+ countries, including the Philippines and Brazil. For India, crypto regulations are more complex — OKX is accessible but Indian users should be aware of the 30% crypto tax on gains and 1% TDS on transactions above certain thresholds introduced in Indian tax law. For Philippines and Brazil users, there are no major regulatory obstacles. You can register, verify your identity with a local government ID, and buy USDT through P2P using local payment methods (GCash, bank transfers in the Philippines; Pix, bank transfers in Brazil). The entire process works the same regardless of country — stock token perpetual contracts on OKX are identical for all users globally. Just confirm that OKX is available and permitted in your specific jurisdiction before starting, as regulations can change.

What if AI keeps going up and I shorted it?

This is the single most important risk scenario to plan for, and it's why stop-losses are absolutely non-negotiable. If you shorted Nvidia at $120 and it rises to $150, that's a 25% move against you — with 3x leverage, that's a 75% loss on your margin. Without a stop-loss, you could get liquidated (lose your entire margin). With a stop-loss set at 10% above your entry ($132), your maximum loss is capped at 30% of margin with 3x leverage. Painful but survivable. Beyond stop-losses, the pair trading strategy in Scenario C provides natural protection: if AI keeps rising, your long positions (Microsoft, Nvidia) appreciate and offset some or all of your short losses on weaker AI companies. The key principle: never have a naked short position without both a stop-loss and ideally a corresponding long position to hedge. And never, ever short with money you can't afford to lose completely, even with stop-losses — in extreme gap moves (like a stock opening 20% higher on news), stop-losses can fill at worse prices than set. Plan for the worst case, not the expected case.