AI:The 4th Great Infrastructure Gamble in Human History

Railways, Roads, the Internet—and Now AI: The 4th Great Infrastructure Gamble in Human History

This earnings season, tech giants quietly unveiled a staggering plan:
This year they will pour US$342 billion into building AI infrastructure.

The figure has Wall Street asking:
Is this the greatest investment of all time, or the most spectacular bubble?

How mind-blowing is the number?
💰 After inflation, it eclipses the entire Apollo moon program
💰 It equals 1.2–1.4 % of U.S. GDP
💰 In real terms, it is 50 % higher than global telecom investment at the peak of the dot-com bubble

Only a few periods in history have seen infrastructure spending on this scale:

  • 19th-century Railway Mania: 3–4.5 % of U.K. GDP

  • Early-20th-century roads & cars: 2 % of U.S. GDP

  • 1990s fiber-optic build-out: 1.5 % of U.S. GDP

  • 2020s AI infrastructure: 1.2 % of U.S. GDP

Every earlier episode ended in a bust. Will this time be different?

Gamble #1 – Railway Mania

1840-1850: U.K. private railway investment reached 3–4.5 % of GDP.

The logic was identical to today:
“Railways will change everything!” (today: “AI will change everything!”)
“First-mover wins—miss the boat and you’re finished!” (today: “AI arms race”)
“Losing money now doesn’t matter; profits will come!” (today: “land-grab first”)

In 1845 alone, 12,875 km of rail lines were submitted for approval—only slightly less than Britain’s entire 16,000 km network today.

Result?
📉 1847: Railway bubble bursts
Hundreds of rail & finance firms collapse
Investors wiped out; Britain slides into recession

✅ But the rail network survives
Becomes the backbone of the Industrial Revolution
Transforms freight & passenger transport
50 years later, hailed as the greatest investment ever

Lesson 1: Over-investment creates bubbles, but infrastructure endures.

Gamble #2 – Auto Fever

1910-1920: The automobile ignited the second tech-fueled frenzy.

Same script:
“Cars will change everything!”
“Every family will own one!” (today: “AI will be standard”)
“Expensive & unreliable now, but mass adoption is inevitable!” (today: “land-grab first”)

1910: 1 car per 1,000 Americans
1929: 200 cars per 1,000 – a 200× surge in 20 years

1900-1929: >1,800 U.S. carmakers
By the 1930s, fewer than 20 survived the Great Depression.

Result?
📉 Late-1920s bubble bursts
Thousands of carmakers bankrupt or merged
Investors wiped out; survivors (Ford, GM) bruised

✅ But the auto industry & highway system endure
Become the engine of 20th-century growth
Reshape life, cities, commerce
30 years later (with 1950s Interstates), hailed as the greatest investment ever

Highways avoided rail-style busts because governments, not speculators, paid.

Lesson 2: Application-layer bubbles burst, but sector champions survive.

Gamble #3 – Fiber Bubble

1996-2001: Telcos burn US$1 trillion (today’s dollars) laying fiber.

Forecast: “Internet traffic doubles every 90 days!”
Reality: Doubled annually4× slower

Peak capex: US$120 bn in 2000US$213 bn today
Massive over-capacity → glut of fiber

Result?
📉 Bubble bursts 2000-2005
80–95 % of the 80 million miles of fiber sits “dark”
Global Crossing, WorldCom bankrupt; fiber prices crash 90 %
Cisco stock –89 % from peak

✅ Yet demand arrives 2006-2015
YouTube (2005), Netflix streaming (2007), cloud boom
“Dark fiber” turns to gold; WDM tech boosts capacity 100×
Buyers of cheap fiber (Level 3) become winners

Lesson 3: Real demand may be late, but it never fails to show up.

Gamble #4 – AI Investment

We are now at the starting line of the fourth great infrastructure gamble.

Where does AI spending stand?
🚂 19th-c railways: 3–4.5 % of U.K. GDP
🚗 20th-c autos & roads: 2 % of U.S. GDP
💻 1990s fiber: 1.5 % of U.S. GDP
🤖 2020s AI: 1.2 % of U.S. GDP

AI investment is moving from “rational” to “euphoric”—and the totals keep climbing.

May be an image of textMay be an image of text

So, will this time be different? History teaches three things:

  1. Over-investment → inevitable bubble
    Not a valuation issue; a human-nature issue.
    When everyone believes “AI changes everything” and fears being left behind, over-building is guaranteed.
    Subsidies, cheap money and FOMO only fan the flames.

  2. Bubbles burst, but infrastructure stays
    Short term: investors lose

  • Railways: hundreds of rail firms dead

  • Autos: thousands of carmakers gone

  • Fiber: trillions vaporized

  • AI: ?

Long term: the rails, roads, fibers—and AI data centers—remain.
1840s tracks still carry trains; 2000 fiber still carries bits; 2025 data centers will still serve 2050.

  1. The next-decade winners may NOT be the “pick-and-shovel” stars
    2000: Cisco is the most valuable firm on Earth
    2020: The world is reshaped by then-tiny Google (born 1998) and money-losing Amazon

$Cisco(CSCO)$ today: US$0.28 trn
$Alphabet(GOOGL)$ today: US$3.4 trn (12× Cisco)

Will the 2035 or 2045 champions be:

  • An obscure AI-app start-up?

  • A loss-making, unloved company?

  • A firm that hasn’t even been founded yet?

The AI era’s true winners may surface only after the build-out bubble bursts.

May be an image of text that says 'Capital expenditure share of GDP much higher for hyperscalers today vs. telecom companies during dot-com bubble % GDP 1.6% 1.4% APOLLO Capital expenditure 1.2% Telecom companies during dot-com bubble 1.0% Hyperscaler companies today 0.8% 0.6% 0.4% 0.2% 0.0% 1996/2021 Note 1997/2022 companies indlude Xo 1998/2023 Meta, Amazon andGoogle Telecom companies Factset. 1999/2024 2000/2025 2001/2026 Level3 3Communications WorldC.om, Global Crossing Nortel Networks, Verizon, AT&T Nokia 2002/2027 2003/2028'May be an image of text that says 'Capital expenditure share of GDP much higher for hyperscalers today vs. telecom companies during dot-com bubble % GDP 1.6% 1.4% APOLLO Capital expenditure 1.2% Telecom companies during dot-com bubble 1.0% Hyperscaler companies today 0.8% 0.6% 0.4% 0.2% 0.0% 1996/2021 Note 1997/2022 companies indlude Xo 1998/2023 Meta, Amazon andGoogle Telecom companies Factset. 1999/2024 2000/2025 2001/2026 Level3 3Communications WorldC.om, Global Crossing Nortel Networks, Verizon, AT&T Nokia 2002/2027 2003/2028'


Every infrastructure mania looks like a bubble to contemporaries—and a godsend 50 years later.

1840s railways, 1920s roads, 2000 fiber: all still in use.
What will survive from the 2020s AI data-center boom?

History says:
Bubbles pop, investors lose, humanity keeps marching.
Railways, cars, the Internet—each proved the point.
AI will likely do the same.

Perhaps the US$342 billion is not a bubble,
merely this generation’s down-payment on the next 50 years
we just can’t yet see what it will buy.

So, what do you think will be the lasting legacy of today’s AI spending spree?


🧘Welcome to Open a Cash boost Account @Tiger_CashBoostAccount today and enjoy access to a trading limit of up to SGD 20,000 with upcoming 0-commission, unlimited trading on SG, HK, and US stocks, as well as ETFs.

Disclaimer: Investing carries risk. This is not financial advice. The above content should not be regarded as an offer, recommendation, or solicitation on acquiring or disposing of any financial products, any associated discussions, comments, or posts by author or other users should not be considered as such either. It is solely for general information purpose only, which does not consider your own investment objectives, financial situations or needs. TTM assumes no responsibility or warranty for the accuracy and completeness of the information, investors should do their own research and may seek professional advice before investing.

Report

Comment2

  • Top
  • Latest