Central banks stoked the dotcom bubble. They’re doing it again for AI

After a few decades as a daily stockpicking column, Questor returns to its 60-year-old roots by taking a weekly view of the markets – what is moving them, what lies ahead and how all of this could affect your portfolios and financial goals.

Blow-out quarterly earnings estimates from American software and cloud services specialist Oracle, a $100bn (£75bn) investment in ChatGPT developer OpenAI by silicon chip designer Nvidia and, in reverse, the purchase of a 10pc stake by OpenAI in another semiconductor maker, AMD, are all keeping the artificial intelligence (AI) pot boiling, so far as stock markets are concerned.

The magnificent seven of Alphabet, Amazon, Apple, Meta, Microsoft, Nvidia and Tesla now have an aggregate stock market capitalisation of $21tn – or 36pc of the S&P 500 index’s total value. Their spending on AI and their share prices are firing US economic growth, which continues to shrug off tariff turmoil, corporate earnings – and thus, the US stock market.

In theory, where the US goes, the rest are sure to follow. America is the world’s largest economy and home to its biggest stock market, biggest bond market and its reserve currency the dollar, all of which count for rather a lot. Many headline stock indices are cruising to new highs, including our very own FTSE 100.

Yet someone is worried. The minutes of the latest meeting of the Bank of England’s Financial Policy Committee (FPC), released last Wednesday, offer warnings of increased risks of a sharp pullback in global share prices. The question now, as one email sent to this column immediately reflected, is whether this is a timely alert or a repeat of Alan Greenspan’s “irrational exuberance” speech of 1996. The then-Federal Reserve chair cautioned about the dangers of a galloping stock market, only for share prices to continue rising, not peaking until 39 months later.

The FPC’s frets about the sky-high valuations of anything AI carry uncomfortable echoes of the technology, media and telecoms (TMT) – or dotcom – bubble that inflated over the 1990s, popped in 2000 and bottomed in 2002. It then took America’s technology-laden Nasdaq Composite index, the poster child for the bubble, until April 2015 to return to its March 2000 high, teaching investors a painful lesson of the perils of choosing narrative over valuation.

Greenspan had, in the eyes of some, tried to calm animal spirits before the bubble even got going – and with good reason. Just eight months after the Fed chair spoke, trouble blew in from an unexpected quarter when Thailand devalued the baht. Indonesia, South Korea and Malaysia quickly followed with the rupiah, won and ringgit, and the Asian debt crisis began to roil markets.

Russia’s own debt crisis and rouble devaluation followed in August 1998, which in turn prompted the collapse of Long-Term Capital Management (LTCM). It had borrowed heavily to fund its long and short positions and strategies that had been drawn up by its mathematical models.

But no sooner had the Fed chair been proven correct, he helped to orchestrate the $3.6bn bail-out of the stricken LTCM hedge fund and, for good measure, threw in two emergency interest cuts from the Fed – one of a half-percentage point and one of a quarter – to take the Fed funds rate to 4.75pc from 5.5pc.

Markets got the message, lapped up the liquidity and implicit central bank support, and latched on to the prospect of how the internet and 3G mobile telephony could change the world. Cue the TMT bubble, which some argue Greenspan helped to stoke, despite his initial warning. Perhaps that is what the gold price is telling us now.

Policymakers may worry but they will bail out and provide liquidity in the event of another accident, come what may.

In the end, the forecasts that underpinned the TMT boom more than two decades ago proved conservative. The problem was that the S-curve of demand did not develop as rapidly as valuations required to maintain the momentum at that time.

The big test for AI is therefore the same one: when will enough customers show up and pay up to provide a return on the substantial upfront investments being made? Broker research puts the total capital investment budget of the so-called AI hyperscalers (Meta, Microsoft, Amazon, Alphabet and Oracle) at $389bn, up 63pc from last year, which was in turn 60pc higher than 2023’s spend. Another 10pc hike – to $428bn – is forecast for 2026.

At some stage, a return is needed on this investment for investors to keep the faith.