Bitbaby Research | May 13, 2026 | 12 min read
How Does This Rally Compare Historically?
Dalio studied 10 major market bubbles. Average duration: 2.5 years. Average gain: 244%.
This rally started in late 2022. We’re in year three. S&P 500 up 79%. Nasdaq 100 up 130%.
Duration: past the historical average. Magnitude: not there yet.
But one number has no historical precedent: the top five companies now account for nearly 30% of the S&P 500. One bad earnings report from any of them moves the entire index.
How Close Are We to 2000?
Buffett Indicator 221%: measures total stock market cap relative to GDP. It has crossed 200% exactly once before — early 2000, right before the bubble burst. Now at 221%, an all-time high.
Shiller CAPE 39.7: inflation-adjusted P/E using 10-year average earnings. In 150 years of data, only the dot-com peak in 2000 (44.2) has been higher. Currently the second-highest reading ever recorded.
The difference from 2000: companies this time have real revenue and are investing from earnings, not debt. That’s JPMorgan’s argument for why this isn’t a bubble. But the Buffett Indicator is already higher than it was in 2000. The Shiller CAPE is 4.5 points away from matching it.
Grantham (GMO) puts the probability of a bubble burst at “close to 100%,” comparing AI to the 1840s railway bubble — railways changed the world, but most railway investors lost everything. JPMorgan says not a bubble. Dalio puts it most precisely: AI euphoria is at roughly 80% of the levels seen before 1929 and 2000. He doesn’t recommend selling, but warns to expect low returns.
Who Is Actually Winning This Rally
The big winners in this rally aren’t AI application companies, and they aren’t the Big Tech companies spending the money, they’re the infrastructure supply chain.
NVIDIA is up over 450% in two years. Broadcom, TSMC, and ASML followed. Power and cooling company Vertiv more than doubled. Data center REITs surged.
NVIDIA converts those purchase orders into earnings. The stocks going parabolic are AI infrastructure suppliers — not AI itself.
The core risk: NVIDIA’s 53% growth rate isn’t something NVIDIA controls. It’s determined by whether the four big buyers keep spending. If they pull back, the assumptions under a 43x P/E are gone.
Why the Math Doesn’t Work Yet
Two specific calculations:
Goldman Sachs: To maintain the returns on capital investors expect, these companies need to generate $1 trillion in annual profit from AI. Wall Street’s current consensus estimate is $450 billion. The gap is more than double.
Sequoia Capital: Take NVIDIA’s revenue, double it for data center operating costs, double again for end-user profit margins. That’s $600 billion — the annual revenue the AI industry needs to justify current infrastructure spending. Total AI industry revenue right now? About $100 billion. A 6x gap.
What’s driving continued investment isn’t ROI — it’s FOMO. Fear that competitors will invest and you won’t. Same psychology as the late-stage dot-com bubble: not investing because returns are visible, but because stopping feels more dangerous than continuing.
Five Indicators Worth Watching
Earnings Growth: Market is discounting it
NVIDIA’s projected 53% annual growth makes 45x P/E look reasonable. But the stock is up just 15% this year, vs. +450% over the prior two. The market is saying: growth is real, but the most profitable stretch is already priced in.
Earnings Quality: Diverging
Google Cloud: 48% growth, rewarded. Microsoft: $150B on infrastructure, market skeptical about utilization. Same AI narrative, opposite verdicts — classic late-cycle behavior.
Valuations: One assumption holding it all up
NVIDIA at 43x, below the semiconductor average of 46x — looks fine. But it assumes Big Tech keeps spending at current rates. If growth slows from 53% to 30%, 43x gets expensive fast.
Cash Flow: Yellow flag
Amazon FCF down 71%. Meta projected down 90%. Earnings being consumed by buildout, not returned to shareholders. If AI revenue doesn’t materially improve before 2027, spending cuts become inevitable.
Interest Rates: Working against them
The Fed says inflation hasn’t come down enough. Rate cuts slower than expected. Higher rates increase the cost of every capex dollar being deployed.
Fidelity’s read: no systemic crash signals yet, but cash flow is flashing yellow.
Other Signals Flashing Simultaneously
Beyond Fidelity’s framework, a string of independent indicators are lighting up at the same time:
Moody’s AI recession model: 49%. Historically, crossing 50% means recession within 12 months. One point awayVIX at 27. Nasdaq in correction territory. S&P 500 below 200-day moving averageBank of England publicly warned of correction risk, naming AI stocks specificallyBIS warned AI amplifies the speed and complexity of financial stability risks, with concentration as the core concernTop 5 companies = ~30% of S&P 500. Highest in half a century. One bad earnings report moves the entire index
None of these individually proves a crash. But the last time they all activated simultaneously was 2000.
If It Pops, How Does It Pop?
Regulatory intervention. Governments restrict AI capital deployment or data usage, compressing growth expectations directly.
Market saturation. Enterprises find AI produces little return and start cutting purchases. With 95% currently reporting zero ROI, this path has the highest probability if nothing changes.
Economic downturn. AI doesn’t need to fail on its own merits. A broader recession means budget cuts, and the first thing to go is investment that hasn’t produced returns yet.
One risk fewer people are talking about: NVIDIA sells chips to Microsoft, Microsoft invests in OpenAI, OpenAI runs on Azure, Meta and Google are simultaneously NVIDIA’s largest customers. A problem at any one node propagates along the supply chain — the same dependency structure that made 2008 so severe.
Oliver Wyman quantified the worst case: an AI equity bubble burst could erase $33 trillion in market value. Over $1 trillion in AI-related debt sits in private credit — opaque, off-balance-sheet structures where banks may be underestimating their exposure.
Three Conditions to Track
The top can’t be predicted. But three preconditions can be tracked. When all three show up together, start paying attention:
Big Tech starts cutting AI capex guidance. Currently: not happening. All four still increasingMoody’s recession model crosses 50%. Currently: 49%AI companies begin failing at scale. Currently: isolated cases
Right now: somewhere between 0/3 and 1/3.
The distance between 49% and 50% might be one jobs report.
To track this yourself:
Capex direction: Quarterly earnings capex guidance — focus on Amazon and MicrosoftRecession probability: Moody’s AI recession model, NY Fed probability modelValuation levels: Shiller CAPE, Buffett IndicatorMarket temperature: VIX, S&P 500 vs. 200-day moving averageBTC linkage: BTC–S&P 500 rolling correlation (above 0.6 = elevated spillover risk)
This article does not constitute investment advice. All data from public sources.
Sources:
Goldman Sachs, Tracking Trillions & Why AI Companies May Invest More than $500 Billion in 2026Goldman Sachs / Fortune, *FOMO has proven a stronger incentive than poor stock performance*, May 2026Sequoia Capital, AI’s $600 Billion QuestionGMO (Jeremy Grantham), *Valuing AI: Extreme Bubble, New Golden Era, or Both*, Jan 2026Ray Dalio / Fortune, *AI is in the early stages of a bubble*, Jan 2026Fidelity, *Is AI a bubble? 5 signs to watch for*, 2026Yale SOM (Jeffrey Sonnenfeld), This Is How the AI Bubble BurstsOliver Wyman, *How An AI Bubble Burst Could Shake Global Financial Markets*, Jan 2026BIS, *The financial stability implications of artificial intelligence*, Jan 2026CNBC, Tech AI spending approaches $700 billion / Big Tech capex topping $1 trillion in 2027Bitfinex, AI, Repricing Risk and the Outlook for Bitcoin in 2026CME Group, Why is Bitcoin Moving in Tandem with Equities?Phemex, Bitcoin-S&P 500 Correlation Hits 94%Master of Code, AI ROI: Why Only 5% of Enterprises See Real Returns in 2026
How Far Is the Stock Market From a Bubble? Breaking Down the Real Risks of This AI Bull Run was originally published in Coinmonks on Medium, where people are continuing the conversation by highlighting and responding to this story.