AI - not quite everywhere

AI has driven the US stock market to new highs. Now it’s driving the US economy. In the first half of this year investment in IT and data centres accounted for all of the growth in the US economy. Tech investment accounts for just 6.1% of total US GDP yet between January and June this small sector kept America growing.

Just four companies, Google, Microsoft, Amazon and Meta, account for the lion’s share of this capital spending. They are leaving the rest of the world behind. Stanford University estimates that in 2024 US private investment in AI was 11 times Chinese levels and 5 times European levels.

Today’s surge in AI-related stocks and investment has drawn comparisons with earlier technology-driven booms. The railways, the telephone, electric lighting and power, electronics – each successive technology has driven equity valuations and investment. The last major tech boom was in the 1990s when capital flooded into computing, networks and software, and tech stocks soared.

That set the modern standard for tech booms, one which, in some respects, AI is exceeding. IT capex accounts for a larger share of US GDP today than at the height of the dotcom boom. Tech valuations are higher. Meanwhile tech shares are even more dominant in equity indices. The Magnificent Seven tech stocks account for a third of the US S&P 500 index today, more than twice the share accounted for by tech behemoths including Lucent, Cisco and Dell in the late 1990s.

The dotcom boom turned to bust in March 2000, with the NASDAQ index of US tech stocks losing almost 80% of its value in the ensuing two and a half years. Investors were burned and companies failed. But the boom created new infrastructure and business models that brought lasting benefits. By the eve of the financial crisis, in 2007, the change wrought by the tech boom of the 1990s were widely thought to have raised productivity and growth rates.

AI companies are hoping for similarly transformative effects today. While share prices and investment among AI providers have soared, the test is whether AI can raise productivity for the rest of the economy. A widely cited MIT survey of 300 companies released last month found that 95% of AI pilot projects fail. The report said that “just 5 per cent of integrated AI pilots are extracting millions in value, while the vast majority remain stuck with no measurable [profit and loss] impact”. The Economist suggests that data security concerns, shortages of technical expertise and poorly organised data make it harder for businesses to extract value from AI.

This could be due to lags. History shows it often takes time to fully harness the benefits of new technologies. It took more than 50 years from Faraday’s pioneering experiments to the early commercial exploitation of electricity in the 1880s. In 1987 the Nobel laureate Robert Solow famously said that “you can see the computer age everywhere but in the productivity statistics”. It was not until later, in the 1990s and early 2000s, that the productivity statistics caught up.

A less benign explanation is that the promise of AI has been over-hyped. The Economist notes that not all tech booms lead to transformational change, pointing out that, “much of the capex by Japanese electronics firms in the 1980s ultimately served no useful function”.

You see AI in the stock market and in US GDP data. Now it needs to prove it can move the productivity statistics too.