Why the AI boom is not a dotcom redux

The media hype over artificial intelligence since Microsoft announced its investment in ChatGPT in January inevitably calls to mind the excesses of the dotcom bubble.

The sense of déjà vu was reinforced last week as the market capitalisation of Nvidia, whose chips power AI applications at ChatGPT among others, briefly topped $1tn. So is it a case of here we go again?

In fact, no. There is much about this AI buzz in the markets that is healthy.

The plunge in Big Tech stocks last year was substantially to do with central banks raising interest rates. Applying a higher discount rate to distant future cash flows in the tech sector shrank the present value of those cash flows. This year’s bounce, far from being driven by central banks, reflects something real.

The simulation of human intelligence in machines has dramatic potential to change the way the economy works. Some people will profit greatly from the process. In the case of Nvidia they have already made a considerable killing this year.

It is easy, now the monetary tightening cycle has been under way for some time, to forget just how artificial market conditions have been and for how long. A new report from the McKinsey Global Institute points out that before the turn of the millennium, growth in global net worth largely tracked growth in gross domestic product. But then something unusual happened.

Around the year 2000, with timing that varied by country, net worth, asset values and debt began growing significantly faster than GDP. In contrast, productivity growth among G7 countries has been sluggish, falling from 1.8 per cent per year between 1980 and 2000 to 0.8 per cent from 2000 to 2018. AI has the potential to help take us beyond this world of asset price levitation and debt-dependent growth through its capacity to improve productivity.

Dario Perkins of TS Lombard suggests that two mechanisms will drive this improvement. First, AI can make current processes more efficient. It is already helping workers make better informed decisions, optimise their processes and remove mundane tasks. The resulting increase in the efficiency of the workforce should boost overall output.

And then AI can help workers invent new things, make new discoveries and generate technological progress that can raise future productivity. Meantime a number of studies have shown that Generative AI, which is capable of self-learning and performing several tasks, will boost the efficiency of workers and companies that use it.

Note, too, that this could all happen much faster than anything in the dotcom bubble. The public facing version of ChatGPT reached 100mn users in just two months. Data analytics firm GlobalData (which recently acquired TS Lombard) estimates the global AI market will be worth $383bn in 2030, a 21 per cent compound annual growth rate over 2022.

Much media commentary has harped on the scope for AI to cause unemployment to rocket — a fear that has been encouraged by AI enthusiasts talking about driving down labour costs. Yet Perkins points out that the ultimate impact of technology on labour markets is theoretically ambiguous.

This is because technological advancements have two contradictory effects: a substitution or displacement effect, where labour-saving technologies can displace workers, and an income or compensation effect, where technology makes all goods and services cheaper, raising real incomes and generating new sources of demand in other sectors of the economy. Throughout history the compensation effect has consistently outweighed the displacement effect.

No one can be certain whether AI will buck that historical trend or indeed reach or exceed human levels of comprehension. In its current stage of development it can be untrustworthy and even spew out nonsense. Equally imponderable is whether AI’s deflationary impact will outweigh the current inflationary forces of supply shortages and tight labour markets and the future upward price pressure from shrinking workforces in the developed world and in China.

Nvidia chief executive Jensen Huang detected last week “the tipping point of a new computing era”. He could be right. It seems likely that Big Tech will continue to march to a different beat to more conventional companies in the S&P 500 index that are more sensitive to monetary policy. One lesson investors should recollect from the dotcom era is that much dross goes up alongside companies of real substance. At today’s valuations, we may not be far from the winnowing of the dross.

john.plender@ft.com