Are We Being Railroaded by AI?
Just how much are we spending on AI?
Compared to other massive infrastructure projects, AI is the sixth largest in US history, so far.
World War II dwarfs everything else at 37.8% of GDP. World War I consumed 12.3%. The New Deal peaked at 7.7%. Railroads during the Gilded Age reached 6.0%.
AI infrastructure today sits at 1.6%, just above the telecom bubble’s 1.2% & well below the major historical mobilizations.
Companies like Microsoft, Google, & Meta are investing $140B, $92B, & $71B respectivelyin data centers & GPUs. OpenAI plans to spend $295B in 2030 alone.
If we assume OpenAI represents 30% of the market, total AI infrastructure spending would reach $983B annually by 2030, or 2.8% of GDP.
To match the railroad era’s 6% of GDP, AI spending would need to reach $2.1T per year by 2030 (6% of projected $35.4T GDP), a 320% increase from today’s $500B. That would require Google, Meta, OpenAI, & Microsoft each investing $500-700B per year, a 5-7x increase from today’s levels.
And that should give you a sense of how much we were spending on railroads 150 years ago!
Sources
World War I & II:
New Deal:
Recommended by LinkedIn
Railroads:
- Melville J. Ulmer, “Capital in Transportation, Communications, & Public Utilities: Its Formation & Financing” (1960), via Silicon Valley Capital Partners
- NBER: Railroads & Economic Growth in the Antebellum United States
- American-Rails.com: Railroads In The Gilded Age
Telecom Bubble:
Apollo Program:
Manhattan Project:
AI Infrastructure:
- Microsoft, Google/Alphabet, Meta earnings reports & CAPEX guidance (FY24-25)
- Bureau of Economic Analysis: US GDP Data
Methodology
All historical spending figures are adjusted to 2025 dollars using Consumer Price Index (CPI) inflation data. Each figure represents peak single-year spending in the year indicated. Percentages show spending as a share of GDP in that specific year, not as a percentage of today’s GDP.
For example, WWII’s $1,152B represents actual 1944 defense spending ($63B nominal) adjusted for inflation, which consumed 37.8% of 1944’s GDP ($175B). This differs from asking “what would 37.8% of today’s $30.5T GDP cost?” which would yield $11.5T.
- Assuming 2.5% annual GDP growth to $35.4T in 2030
Absolutely insane, added for reference point the % spent on fighting climate crisis (makes me really wish we considered it ww3)
Not talking about the size, I believe it's the most advanced one, and comparable with the railroads. Growth of AI investments carries with it: scientific development, pure infrastructure and compute-power developments, talent acquisition, hardware advances, etc. - a lot of positive things that will impact industry for a long time.
Tell me you have no clue without telling me that you have no clue. 🤣
Tomasz, amazing article! The WWII spend/AI Infrastructure multiple of 23.6x is a great starting point. However, it's presented in a vacuum. Adjusting for structural differences reveals a more intense picture: Fiscal Space: 1944's 37.8% spend happened as Debt/GDP rose to 98% from 44%. Today, we start at 123% Debt/GDP, with total federal spending at 24%. Adjusting for this existing burden: 1.6% / (1-0.24) = 2.1% (18x multiple). Industrial Base: The economy was 44% goods-producing in 1944 vs. 18% today. Comparing relative to the mobilizable base: 37.8%/0.44 = 85.9% vs 1.6%/0.18 = 8.89%. New multiple: 9.7x. Global Dominance: The U.S. was 50% of Global GDP in 1944 vs. 25% now. Adjusting: 37.8%/0.5 = 75.6% vs 1.6%/0.25 = 6.4%. New multiple: 11.8x. Your analysis also excludes labor, R&D, and government investment. Conclusion: The real multiples are 9.7x to 18x. This isn't a little thing—we're embarking on an expense equivalent to 5.5%-10.3% of the WWII effort without a war, while in fiscal dominance and facing supply chain threats. God help us!