There have been historically only 3 capex frenzies in US. The first was peak railroad spending in 19th century and the next was the peak telecom spending around the dot come bubble. Now we are seeing a similar trend in capital expenditures on AI data centres. Globally, data centre capital expenditure (capex) driven by AI is projected to reach $1.2 trillion by 2029 with the top four hyperscalers (Amazon, Google, Meta, and Microsoft) are expected to account for half of the total spend of 1.2 TN USD. So far this year, AI capex, which is defined as information processing equipment plus software has added more to US GDP growth than consumers' spending. Without AI datacenter investment, Q1 GDP contraction could have been closer to –2.1%. (actual was -.5%). But where is this gigantic funding coming from. Initially it was internal cash flows of the Mag-7 but now VCs & PEs i.e. the private pool of capital has overtaken the public of capital. Private credit funding of artificial intelligence is running at around $50 billion a quarter, at the low end, for the past three quarters. Even without factoring in the mega deals from Meta and Vantage, they are already providing two to three times what the public markets are providing. The amount of CMBS backed by AI infrastructure is already up 30%, to $15.6 billion, from the full year total in 2024. But by spending GDP-moving amounts of money on GPUs and such, it is not, by definition, being spent on something else. The telecom capex bubble led to a sharp decline in "other" infrastructure spending, one that is still playing out. The datacentre spending frenzy will almost certainly do the same, starving other infrastructure for money. But is such large spending worth the expected productivity growth? We decided to compare the current productivity growth of AI against similar trend in the telecom capex in the 2000s. We used the release of ChatGPT 3.5 in November 2022 for AI, and that of Netscape Navigator 2.0 in September 1995. Both are almost performing same. But now a days the US economy requires a considerably larger degree of capital spending to continue to generate the same sort of labor productivity growth that we got nearly three decades ago. Substantially-higher capital investment that yields identical labor productivity implies that total-factor productivity growth has been substantially inferior to that observed during the dot-com era. Interestingly, as AI usage has ramped over the last year, total-factor productivity has pretty much flat-lined. That’s a big contrast to the Internet era and perhaps raises a few questions about the zillions of dollars that are piling into AI investment, as well as what the actual prognosis might be for US trend growth. Summary: The only fact we can agree upon now is that capital is being aggressively reallocated—from venture funding to internal budgets—at the expense of other sectors. Entire categories are being starved of investment, and large-scale layoffs are already happening. The irony is that AI is driving mass job losses well before it has been widely deployed. So even if AI fails to deliver the productivity gains it promises, job losses will have swelled by the time markets realise. And if AI does in fact provide productivity gains, more job losses will happen. In either case, AI leads to job losses whether or not productivity gains materialise.
