The Buffett Indicator shows that the US Stock market is now at 250% of GDP, US Stocks are worth 2 and a half the value of everything the economy produces in a year. But is that a sign of overvaluation?
Well, there are several indexes which are reminiscent of previous crashes, to give just one, the Shiller PE is within touching distance of 2000 Dot Com Crash.
However, although many indexes are flagging red, record corporate profitability, and AI hype mean some still remain bullish.
This is the stock market, reaching near record levels. And some may argue that the reason the stock market is booming, is because in the last decade the Fed created money for QE. However, even if we adjust stock market for the money supply, the ratio is above 2000 peak. Another evidence, the stock market is overvalued is the dividend yield from shares.
This basically is the effective income you get from holding shares. You can see it is very near a record low, and lower than pre-dotcom bust of 2000 and 2007. It is worth noting the 2000-09 bear market couldn’t be rescued by dividends because they had collapsed, today dividends are even lower.
At the same time, something else is happening in financial markets, bond yields are rising sharply. When bond yields are rising, shares look relatively less attractive. But, actually rising bond yields also highlights concerns over inflation and US government debt. You can see US government debt is substantially higher than 2000 and 2007, which doesn’t on its own affect share valuations, but if there was a crisis, there is less room for manoeuvre than in the past, especially with US Treasury currently concerned about rising bond yields.
Margin Debt
Another warning sign of something that happens before a financial crash is a rise in margin debt. Over-exuberance is encouraging investors to borrow money to use leverage in buying shares. This was the major factor behind Wall Street Crash of 1929, and modern platforms make it easier than ever, in fact, US households have never been so exposed to the stock market, which means a stock crash would have a bigger effect on economy than in the past . Margin debt has risen quite sharply in the past years, and if we look at borrowing to buy shares as a share of GDP, the level is now above 2000 and 2007.
AI Bubble
However, the biggest fear is the potential bubble in AI. The Mag 7 now make up about 34% of the S&P 500, having first passed 30% in 2023. But Joachim Klement at Panmure Liberum looked at accounts of AI firms and even using generous terms such as zero costs, only Amazon has a clear path to profitability.
Klement’s wider calculation is that hyperscalers would need $2–5 trillion in additional annual revenue to earn even a 10% return
But, there’s more issues. AI Debt is being hidden off-balance sheets and so is actually worse than it looks
Less damaging measures
Now, we could stop here. But that would be selective. Firstly, margin debt growth was faster in 2000 and 2007. The Shiller CAPE was slightly higher in 2000, though I actually don’t think that is much comfort because after reaching peak it fell pretty sharply. A third measure is that the Vix Fear Gauge is actually quite low compared to 2000 and 2007. Though bears may well claim that lack of fear and low volatility more likely reflects high levels of complacency.
Also, you could point to a low unemployment rate, and a relatively low Federal funds rate Also the corporate bond spread which reflects risk on corporate borrowing is lower than 2000. That means fear of default is currently low. Also, another factor behind the bull run, is Company profit margins are still increasing. And finally if you look at the Sahm Rule it is historically a pretty good indicator of recessions arriving, but at the moment, the outlook seems relatively benign.
AI Bubble (Again)
However, there are issues with all of these indexes not telling the full story. The Bank for International Settlements recently issued a warning that we are seeing an AI investment bubble which rivals the canal mania, railways and dotcom bubbles. Basically genuine technological breakthroughs tend to pull in more capital than returns can justify. The Stock boom has been predominantly based on the Magnificent seven.
But, whereas the buildout was once based on cash, cashflows are going negative, which means more borrowing to fund investment. In 2025, there was $354bn of net new debt. And across the wider AI complex, Morgan Stanley expects around $570 billion of new borrowing this year alone. And the concern is that markets are becoming less willing to lend to AI firms. We mentioned credit rating risk is low. That is true for most of the market, but not AI firms.
The increased speculative nature of AI and uncertainty revenues will meet expectations means banks demand higher borrowing costs, and with interest rates rising, it is going to be even more expensive to borrow and invest. In 2008, it was higher interest rates that were the tipping point to end the growth of a bubble, in that case housing.
The AI boom has led to a huge circularity of money flows. Chip makers invest in the AI labs. The labs buy computing power from the cloud giants. The cloud giants buy chips, so the money goes round — and every step gets booked as revenue. The BIS found 96% of chip makers’ announced AI deals are circular. It isn’t fraud, but it means the sector’s ears is primarily coming from its own eco-system.
Some more positive signs
If there is a slowdown or pull back in investment, we will see a sharp slowdown in economic growth and more job losses. Also, earlier we mentioned unemployment is low at 4.1%, but this masks a fall in employment growth and rise in people leaving the market, labour market participation tells a different story. Also, inflation may seem relatively low at 3.5%, but the problem is that there is great uncertainty about future inflation. With oil supplies disrupted by conflict, there still remains risk of further rises in the oil prices. It is this risk of inflation, that is pushing up bond yields and interest rates. Given the greater leverage of the AI buildout, rising interest rates will be a big concern.
Nevertheless, it is worth bearing in mind, that some warning signals have flashed before, and no crash arrived. In 2011, 2015 and 2018, we had a rise in earnings ratios and uncertainty, but the stock market shrugged it off to continue with its record bull run. Also, it is worth putting the stock market in historical terms. This shows that 2009 to 2026 bull run has given returns of 10 times, but the 1978-2000 bull run lasted 22 years and delivered 16.7x. It also shows how the bear markets are more about the stock market going sideways or worse.
But can we compare to the 1978-2000 bull run? No, because in 1978, stocks were really cheap and it was a still an era of relatively high economic growth. The outlook for future economic growth is really very different. Growth is slowing down and we are getting to a situation where the effective interest rate on debt exceeds nominal growth. (r > g). AI may give a productivity boost, but the ageing population will have a fairly negative impact on nominal growth. Simply there will be more retired and fewer workers. Also, remember we said corporate profitability is at an all-time high. But how sustainable is that? The labour share of income has fallen. The housing market is stagnating at the bottom end and booming at the top. Outside the top 10%, consumer spending is weak.
In terms of AI Stock performance, one of the most important trends is the growing share of Chinese open source AI. This is eating into potential profitability of US tech firms, and therefore the whole AI infrastructure. And also in AI we do see signs of what economists call irrational exuberance in the race for AI singularity.
So to summarise, there are very real warning signals about the valuations of stocks. It is near record levels, and AI will need to generate very substantial revenues to make the gamble pay off. The things to look out for are. Will AI credit spreads continue to rise. Can OpenAI and anthropic actually grow their revenue to meet promises and will the US be able to prevent more rises in interest rates throughout the economy. The problem is that this will be a challenge, this next video explains the outlook for US borrowing.
Sources and data
All charts in this video were produced by EconomicsHelp from the data listed below.
Valuation
- Buffett Indicator (US stock market value as % of GDP) — Federal Reserve Z.1 Financial Accounts ÷ BEA nominal GDP. Current Market Valuation
- Shiller CAPE ratio — Robert Shiller Data Library, Yale University.
multpl.com - S&P 500 dividend yield, monthly since 1871 — Standard & Poor’s and Robert Shiller.
multpl.com - S&P 500 historical prices —
multpl.com - Money supply (M2), used for the stocks-to-money-supply ratio —
FRED, series M2SL
Leverage and borrowing
- Margin debt (borrowing to buy shares) — FINRA monthly margin statistics.
The Trading Tools - US federal debt as % of GDP — US Treasury and BEA.
FRED
Credit markets and volatility
The wider economy
- Unemployment rate — US Bureau of Labor Statistics.
FRED - Federal funds rate — Federal Reserve.
FRED - Sahm Rule recession indicator —
FRED
AI investment and financing
- Bank for International Settlements, Annual Economic Report 2026, Chapter I (28 June 2026) — circular AI financing, AI credit spreads, and debt-financed capital expenditure.
bis.org - Joachim Klement, Panmure Liberum — implied returns on hyperscaler AI investment, 2025–30, as published by the Financial Times.
- Morgan Stanley forecast of AI-related debt issuance — reported by Forbes, 17 July 2026.
Forbes - Hyperscaler free cash flow — BofA Research Investment Committee. Figures shown are approximate, read from the published chart.
Notes on the data
- Dot-com and pre-financial-crisis figures are taken at each cycle peak (March 2000 and 2007), not annual averages.
- The stocks-to-money-supply ratio uses share prices only and excludes dividends. The official definition of M2 changed in May 2020, when savings deposits were reclassified, so there is a small break in the series at that point.
- Secular bull and bear market multiples are calculated from monthly average prices. Figures quoted elsewhere using daily data will differ slightly.
- Total return figures are nominal and are not adjusted for inflation.
This video is for general information and education. It is not financial advice.
About The Author
