The great contribution to investing by American Nobel laureate for economics Robert Shiller was to devise a way to judge whether US stocks are overvalued or undervalued.

While many valuations techniques are based on future (‘forward’) earnings that might never eventuate, the Yale University professor born in 1946 looked to the past. His ‘cyclically adjusted price-to-earnings ratio’ is calculated by dividing a share price by the average of the past ten years of earnings adjusted for inflation. If high, the ratio warns of lower returns to say the least.

Shiller added credence to his analysis by calculating his ratio for the US share market on a monthly basis back to 1881 – the ratio’s average over these 144 years is 17.29. At the end of October, the ratio stood at 41.20, the highest since 1881 except for when during the dotcom bubble when it reached 44.19 in 1999.

That means the Shiller PE ratio is higher than its 31.48 peak before the 1929 crash, when stock values plunged nearly 90 per cent. It’s almost at dotcom heights from when the S&P 500 Index sagged 49 per cent and the tech-heavy Nasdaq Composite index dived 78 per cent. Other indicators say US stocks are more overvalued than in 2000.

Such numbers explain why so many, from the IMF and the Bank of England to even tech CEOs, talk of a US stock bubble, as do investors. A survey in October found a record 54 per cent of global fund managers (who each manage more than US$400 billion) think artificial intelligence stocks are in a bubble.

For, of course, it’s a frenzy over AI-themed stocks that has propelled the US bourse (and others) to record highs and likely set up a crash.

Since the release of the ChatGPT AI chatbot in 2022, the S&P 500 has surged more than 65 per cent, and the ‘Magnificent 7’ stocks of Alphabet, Amazon, Apple, Broadcom, Meta Platforms, Microsoft and Nvidia have driven more than half those gains. These stocks, some now worth US$4 trillion while Nvidia tops US$5 trillion, command valuations that meet the definition of bubble in that profit expectations appear unrealistic.

Other AI-themed stocks look just as precarious. Ten loss-making AI start-ups have gained about US$1 trillion in valuation over the past 12 months. Startups have reached market caps of US$12 billion before announcing any products. Energy companies with no revenue deemed essential to the AI transformation have soared too.

The problem is the danger AI-themed stocks pose extends beyond their lofty valuations.

One is that deal-making, especially by private start-up OpenAI, is fanning the bubble, especially as many of the transactions are incestuous. Advanced Micro Devices, Broadcom, Nvidia and Oracle are among stocks that soared on deals to deliver computing power to OpenAI. The first risk is these companies have tied their fates to OpenAI even though this firm competes against the tech giants; OpenAI’s long-term success is not assured.

The other risk is the incestuous nature of many deals – investment becomes income for other tech companies – smacks of the ‘vendor finance’ or ‘circular’ deals of the dotcom mania. The OpenAI deal that boosts demand for Nvidia chips, for instance, is due to an investment of US$100 billion in OpenAI by Nvidia. The other Mag 7 have invested billions of dollars in soaring AI startups too. And so on.

A second AI threat to stocks is the questionable returns on the huge investments being made in advanced chips and data centres. The biggest tech companies are investing US$400 billion in AI this year, and trillions of dollars more by 2030.

While the capital expenditures of Big Tech are mostly funded from retained profits, many AI companies are borrowing to undertake investments with dubious returns. Much of the borrowing is via junk bonds, syndicated loans and private credit – under-regulated and opaque lending sourced from fund managers, insurers and wealthy individuals – that is behind corporate failures and scandals.

A third threat to stocks is that young investors, enthused by ‘neomania’ and aided by technology that gives them unprecedented reach, are propelling AI stocks as they chase trendy assets including cryptos. These youngsters have an appetite for borrowing and seem to overlook fundamentals such as AI being unproven as a profit-spinner.

Lastly is the question of how much AI is hype. Artificial intelligence is a brilliant marketing term for code written by humans that proponents define as formulas that mimic human intelligence. For all the feats AI can perform, code can’t ‘self-learn’, reason, think, feel or comprehend. Nor has code intentionality or a concept of truth – hence the new word ‘botshit’. Talk of artificial general intelligence and AI as magic that will solve intractable problems and lead to abundance is marketing puff often spouted to propel AI stocks. Investors might soon see AI for what it is – sophisticated data-based pattern matching – and choke on their dashed AI hopes.

The danger of the AI bubble is global and economies are more exposed to US stocks than they were in 2000. Over the past 15 years, US households and foreign investors have invested in US stocks such that they comprise 60 per cent of global market cap compared with 40 per cent in 2010.

The IMF estimates a market correction of dotcom magnitude could destroy more than US$20 trillion of US household wealth. The loss of about 70 per cent of US GDP, which is several times the damage of the dotcom bust, would lop GDP growth by at least 2 percentage points and usher in a recession.

Foreign investors could suffer losses exceeding US$15 trillion, or about 20 per cent of world GDP ex-US, the IMF calculates. This would compare with dotcom-crash foreign losses of about US$4 trillion adjusted for inflation, which was less than 10 per cent of rest-of-world GDP at the time.

Who knows if AI stocks will plunge soon or the extent of the wider damage wrought. But if AI stocks do crash, no one can say they weren’t warned.

The US stock surge could persist for some time, especially if the Federal Reserve keeps cutting rates, to be sure. Some AI stocks will prove good long-term investments. Other risks could trigger a stock crash. But any damage would be centred on AI stocks.

One thing investors will learn is how accurately the Shiller PE ratio predicts busts. Likely spot on.