What The Railroads Teach Us About The AI Buildout
Inside today’s Daily Journal…
Essay: Another Great Depression Is Coming
Outside money pours into stocks
Consumer savings dwindle
Musk and Cook: chips prices are sky high!
Chart Of The Day… Credit Acceptance (CACC)
Reader poll
Today’s Mailbag
Today’s Daily Journal is about Bitcoin.
In this month’s Complete Investor, I gave readers my Bitcoin pricing model and explained how it works. As I said to subscribers: because of Bitcoin’s halving cycle, we’re at a low price right now. Bitcoin is, more or less, as cheap as it will ever be in this cycle. And, there’s a good chance that it will move up 10x from here in the next cycle.
But to understand why we believe that’s inevitable, you need to know what happened the last time our country financed an infrastructure project as big as the expected build out of artificial intelligence (“AI”).
The post-Civil War railroad buildout was the largest private-capital infrastructure buildout, by share of GDP, in U.S. history. In the years immediately after the Civil War (between 1865 and 1873) total railroad mileage doubled. In the peak construction years (1871 and 1872), roughly 4% of America’s entire GDP was spent on railroad construction.
The railroads were financed, almost entirely, by debt. At its peak, cumulative outstanding railroad debt equaled a third of America’s economy. Expressed in today’s dollars, railroads issued about $10 trillion in bonds.
The AI data-center buildout is now at 1.9% of GDP for the hyperscalers alone – Microsoft (MSFT), Alphabet (GOOG), Amazon (AMZN), and Meta (META). AI infrastructure spending is projected to reach $5 trillion cumulative by 2030, per consulting firm McKinsey & Co. By the time it’s completed, it will likely become as large as the railroad buildout – or even bigger because there’s one very important difference between the two types of infrastructure.
Railroads are primarily physical infrastructure – tracks, land, and trains – that are durable, long-lasting assets with a utility that changes slowly over time. Once laid, a railroad remains productive for decades or even a century with minimal technological upgrades.
In contrast, data centers are high-tech, capital-intensive facilities defined by rapid technological obsolescence. Their value is derived from the computing hardware – chips, servers, and cooling systems – that lose relevance as processing power improves and energy requirements shift. Consequently, while a railroad’s value lies in its physical geography, a data center’s value is tied to technology that inevitably demands replacement every 10 to 20 years – greatly increasing the capital commitment.
