How much dotcom is in AI?
There are important parallels between the current data center build-out and the glass fiber investments of the 1990s.
TLDR Summary
Appetite for investment is a good thing for an economy. It triggers growth in the present and sets the foundation for more growth in the future. However, a misallocation of capital during the investment phase can come with painful consequences when consumption doesn’t follow, or at least doesn’t follow as quickly as anticipated.
The glass fiber build-out in the 1990s fits into that category. Hundreds of billions of dollars were poured into upgrading the physical telecommunications infrastructure in the US to prepare for customers that didn’t come until about 10 years later.
What happened 30 years ago differs from today’s AI bonanza in some important aspects. But there are also curious similarities that suggest that history may rhyme in this case.
And to the extend it doesn’t rhyme, it might not do so in a good way. The glass fiber build-out paid off eventually by serving as the foundation for trillion dollar companies that wouldn’t exist otherwise. It’s unlikely GPUs can find a similar use case later on if they are being overbuilt right now. Glass fiber cables became core infrastructure. To some extent GPUs are more akin to…consumables.
Investment vs. consumption
The ‘economy’ is typically defined as GDP, gross domestic product, i.e. the sum of all goods and services produced in a given period. GDP doesn’t care what happens with those goods and services, whether they are immediately consumed, whether they are capitalized & depreciated or whether they are simply put into inventory. As such, a production boom can be investment-driven or consumption-driven, depending on the type of economic output demanded by buyers.
Since World War II, the US has gone through various cycles of acceleration and deceleration of real GDP growth. The 1960s were primarily consumption-driven, particularly from families with baby boomer children. The 1980s were a mix of both as falling interest rates stimulated both consumption and investment.
Consumer demand was strong in the 1990s. Internet adoption was rapid. But from a macro perspective, it was primarily an investment-driven boom. Most importantly in communications equipment. I would even go as far as arguing that almost the entire acceleration highlighted in yellow in the chart above can be attributed to it.