The Question That Won’t Go Away

You’ve seen the headlines. “$100 billion in AI data centers.” “GPU shortages driving up costs.” “OpenAI burning through billions while struggling to monetize.”

And everywhere, the same anxious question: Is this another dot-com bubble?

Here’s why that question misses the point entirely.

We’re not witnessing just another tech revolution. We’re watching the second wave of a pattern that has transformed civilization twice before—and the second wave always plays by different rules.

Think about it: The Industrial Revolution had two waves. First came steam engines and railways—built by scrappy entrepreneurs, wild speculation, massive failures. Then came electricity—methodically deployed by established corporations like General Electric and Westinghouse who’d learned from the railway chaos.

The Intelligence Revolution follows the same pattern. First wave: personal computers and the internet—built by garage startups, funded by venture capitalists, ending in the dot-com crash. Second wave: AI—deployed by Microsoft, Google, Meta, Amazon with trillion-dollar market caps and decades of experience.

This changes everything about how the bubble unfolds.

In dot-com, startups with PowerPoint decks and no revenue raised billions. Pets.com famously spent $300 million and went bankrupt in 268 days. The companies building the internet were fragile, speculative, desperate.

But today’s AI wave? It’s Microsoft generating $245 billion in annual revenue while investing in OpenAI. It’s Google with $307 billion in revenue methodically building AI infrastructure. These aren’t desperate startups—they’re profitable giants making strategic bets they can afford to lose.

Yes, there’s massive spending with unclear returns. Yes, people use expensive AI for trivial questions while companies absorb the costs. But that’s exactly what happened when General Electric and Westinghouse “wasted” money electrifying America—the infrastructure survived long after the initial inefficiencies were forgotten.

To understand what happens next, we need to travel back 180 years to a muddy field in Britain, where the first “second wave” was about to reshape civilization.

The Framework: Two Revolutions, Two Waves Each

Before we dive into history, let’s define the pattern that explains everything about AI’s current trajectory.

The Industrial Revolution (1760s-1920s)

The transformation from human/animal power to machine power for physical work.

• First Wave: Steam engines and railways (1760s-1850s)

  • Built by entrepreneurs and speculators
  • Massive financial bubbles and crashes
  • Railway Mania: wild overbuilding, spectacular bankruptcies
  • Result: Physical transportation infrastructure

• Second Wave: Electricity and electrification (1880s-1920s)

  • Deployed by established corporations (GE, Westinghouse)
  • Methodical infrastructure investment
  • Gradual consolidation into utility monopolies
  • Result: Distributed power infrastructure

The Intelligence Revolution (1940s-present)

The transformation from human thinking to machine processing for information work.

• First Wave: Computers and the internet (1940s-2000s)

  • Built by startups and venture capital
  • Massive financial bubble (dot-com crash)
  • Garage entrepreneurs, PowerPoint decks, spectacular failures
  • Result: Information transportation infrastructure

• Second Wave: AI and machine intelligence (2010s-present)

  • Deployed by tech giants (Microsoft, Google, Meta, Amazon)
  • Strategic infrastructure investment by profitable corporations
  • Gradual consolidation expected
  • Result: Intelligence processing infrastructure

The key insight: We’re not in another dot-com bubble. We’re in the electricity phase of the Intelligence Revolution. The financial dynamics, competitive landscape, and likely outcomes are completely different.

1840s Britain: When the Madness Began

Picture London in 1844. The air thick with coal smoke, horse-drawn carriages clattering through narrow streets, and in every coffee house, pub, and drawing room—a new kind of fever spreading.

Railway fever.

The first successful passenger railways had proved themselves. The Liverpool and Manchester Railway was printing money. Suddenly, everyone wanted in. Aristocrats, merchants, clerks, washerwomen—all scrambling to invest in railway schemes.

Here’s where it gets wild: In just five years (1844-1849), Parliament authorized the construction of 9,500 miles of railway. To put that in perspective, Britain built more railway in those five years than exists in the entire London Underground system today—multiple times over.

The money involved was staggering. At the peak of Railway Mania in 1846, railway investments represented 6.7% of Britain’s entire GDP. Today, that would be equivalent to spending roughly $1.7 trillion in a single year on one technology.

And here’s the part that should sound familiar: most of these railway companies were burning through capital faster than steam locomotives burned coal. About a third of the authorized railways were never even built. Investors lost fortunes. Companies collapsed spectacularly.

It looked, by every measure, like disaster.

The Infrastructure Paradox

But something strange happened after the bubble burst.

Yes, investors lost money. Yes, railway companies went bankrupt. Yes, the financial wreckage was enormous. But when the dust settled around 1850, Britain found itself with something unprecedented: a 6,000-mile, gleaming railway network that formed the backbone of its transportation system.

The infrastructure survived the financial catastrophe.

This is the paradox that repeats throughout history: Financial bubbles in infrastructure create lasting value even when they destroy investors.

Think about what those railways meant for Britain. Suddenly, coal from Newcastle could reach factories in Birmingham in hours instead of days. Fresh fish from coastal towns could reach London markets while still fresh. People who’d never traveled more than ten miles from their birthplace could visit other cities.

The railway network didn’t care that the companies that built it went bankrupt. The physical infrastructure—the rails, the bridges, the stations—remained valuable. Someone would use it. Someone would profit from it.

Just not the original investors.

The Pattern Emerges: Five Universal Truths

Once you see the Railway Mania pattern, you start noticing it everywhere throughout history. Let me share the five truths that emerge again and again:

Truth 1: Infrastructure Survives Financial Failure

The railway companies failed, but the railways remained. This happens because infrastructure has intrinsic, enduring utility. A railway isn’t like a speculative stock—it’s a physical asset that enables economic activity whether or not the company that built it still exists.

When Lehman Brothers collapsed in 2008, their office buildings didn’t vanish. Someone bought them and used them. That’s infrastructure.

Truth 2: Wasteful Overbuilding Accelerates Progress

Here’s a counterintuitive truth: The “waste” during infrastructure bubbles serves a hidden purpose.

Yes, Britain didn’t need 9,500 miles of railway in 1849. Many routes were redundant or unprofitable. But this overbuilding meant that by 1850, Britain had railway coverage that would have taken 30-40 years to build “efficiently.”

The market crash forced railway companies to sell off assets at fire-sale prices. Suddenly, communities that couldn’t have afforded railway service got it anyway, because bankrupt companies were desperate to offload anything that might generate even minimal revenue.

Think about fiber optic cables in the 1990s. Telecom companies massively overbuilt—laying way more cable than necessary. Most went bankrupt. But that “excess” fiber became the cheap bandwidth that enabled YouTube, Netflix, and video calling. The oversupply created possibilities no one had imagined.

Truth 3: Consolidation is Inevitable

Competition among railway companies was brutal. Some went bankrupt. Others sank heavily into debt. To survive, companies began consolidating—merging with competitors, buying up bankrupt rivals’ assets.

By 1922, Britain’s hundreds of railway companies had consolidated into just four major groups. Sound familiar?

This pattern repeated with electricity. By the 1920s, ten utility systems controlled three-fourths of US electric power. Samuel Insull’s Commonwealth Edison grew by aggressively acquiring competitors. By 1930, his holding company provided more than 10% of America’s electric power.

The pattern: Many companies enter during the exciting growth phase. A brutal shakeout follows. Three to five mega-corporations emerge controlling most of the infrastructure.

Truth 4: Government Eventually Steps In

Here’s where it gets political.

Private railway companies served profitable routes. But what about the remote Scottish Highlands? Poor rural villages? These areas couldn’t support a profitable railway, so private companies ignored them.

Eventually, public pressure forced government intervention. Similar pattern with electricity—private utilities refused to serve rural areas because running power lines to sparse populations was unprofitable. So in 1935, the U.S. government created the Rural Electrification Administration.

By the late 1950s, 97% of American farms had electricity—not because private markets wanted to serve them, but because the government decided electricity had become essential infrastructure.

The timing is predictable: Government intervenes after consolidation, once it’s clear who the dominant players are and what regulation is needed.

Truth 5: From Miracle to Commodity

The final stage is almost sad in its predictability.

Railways started as miraculous technology. People would travel just to watch trains pass by. The first passengers were terrified—doctors warned that traveling at “20 miles per hour” might be fatal.

But by 1900, railways were boring. Just infrastructure. The miracle had become mundane.

Same with electricity. In 1907, only 8% of American homes had electric power. It was expensive, exotic, thrilling. By the 1960s? Electricity was so ordinary that people only noticed it when it stopped working. The technology that once amazed everyone had become invisible commodity infrastructure.

The Electrification Era: The Pattern Repeats

Let me show you how precisely this pattern repeats by jumping forward 70 years from Railway Mania.

By 1910, electricity was the hot new infrastructure. Factory owners were converting from steam power to electric motors. But here’s the key detail: electricity required massive, concentrated investment in generation plants and distribution networks.

This was expensive, risky, and required vision. Many small electricity providers competed. Most failed. But the ones who survived and consolidated became giants.

Now here’s where it gets interesting for our AI story: The companies that won big during electrification weren’t the ones building power plants.

The Real Winners: Infrastructure Users

Let me tell you about three companies that understood something crucial—they didn’t need to build the infrastructure to profit from it.

Sears, Roebuck & Co. started with Richard Sears as a railroad station agent. He noticed wholesalers had excess inventory and bought watches below cost to resell. His first “office” was a rented room for $10/month with a kitchen table desk and a 50-cent chair.

But Sears understood something: the railway network meant he could mail catalogs to rural America and ship products directly to customers who had no other access to variety or competitive prices.

By 1895, his sales exceeded $750,000. By the early 1900s, he was earning $10 million annually. His secret? He leveraged existing infrastructure aggressively.

When the Rural Free Delivery Act of 1896 expanded mail routes to rural areas, Sears pounced. When parcel post became available in 1913, his sales increased fivefold in the first year.

He didn’t build railways. He didn’t build the postal system. He just used them brilliantly.

Ford Motor Company did the same with electricity. In 1913, Ford’s moving assembly line reduced car assembly time from 12+ hours to under 3 hours. By 1916, it was down to 1 hour 33 minutes.

The secret weapon? Electric motors.

Old factories used a single massive steam engine turning shafts and belts throughout the building. Unreliable, inflexible, dangerous. Electric motors allowed decentralized control. Each workstation could have its own motor. Assembly lines could speed up or slow down independently. Electric lighting meant factories could operate night shifts.

Ford didn’t build power plants. He just understood how to use electricity to transform manufacturing. Production went from 10,660 Model Ts in 1909 to over 1 million by 1920. The price dropped from $600 to $290, making cars accessible to regular people.

General Electric is fascinating because they were both infrastructure builder and infrastructure user. They started as an electrical goods wholesaler in 1886, opened their first factory within three years, and by 1900 had established the first industrial research lab in the United States.

They understood that as more factories electrified, demand for electrical equipment would explode. They positioned themselves to capture that demand—providing 300+ electric motors each to Portsmouth Dockyard and the London subway, creating subsidiaries for telephones, supplying equipment for the UK National Grid.

But here’s the crucial insight: These companies shared common patterns that had nothing to do with the specific technology.

The Universal Principles: What Never Changes

After studying Railway Mania, electrification, and now looking at AI, certain patterns become undeniable. These aren’t about technology—they’re about human nature and economic forces.

Start Where Infrastructure Already Exists

Sears didn’t wait for perfect railway coverage nationwide. He started where trains already ran and expanded as the network grew.

Ford didn’t wait for every factory to have perfect electrical service. He built where power was available and proved what was possible.

This matters because there’s always a temptation to say “I’ll wait until the infrastructure is better.” But the winners move while infrastructure is still imperfect.

Solve Real Problems With Real Economics

None of these companies built “cool technology.” They solved expensive, painful problems.

Sears solved: “I live in rural Nebraska, the general store charges outrageous prices, and I have almost no product selection.”

Ford solved: “Cars cost as much as a house, so only rich people can own them.”

GE solved: “My factory needs flexible, reliable power that I can control station-by-station.”

Notice what’s missing? No vision statements about “revolutionizing society.” Just: “Here’s a painful problem. Here’s how this infrastructure lets me solve it profitably.”

Obsess Over Efficiency

Richard Sears started with a $10/month office, kitchen table desk, and 50-cent chair. Minimal overhead meant he could survive lean times and reinvest profits.

Ford’s entire genius was efficiency obsession. That’s what the assembly line was—a ruthless focus on eliminating wasted motion, wasted time, wasted material.

When you’re using expensive infrastructure, efficiency isn’t optional. It’s survival.

Move at Infrastructure Speed

Here’s a pattern that repeats with eerie precision: Winners adapt to infrastructure changes in weeks or months, not years.

When parcel post launched in 1913, Sears was ready immediately. His sales increased 500% in year one. He didn’t “study the opportunity” or “develop a strategic response.” He pounced.

When electricity became reliable and affordable, Ford redesigned his entire factory around it. Competitors took years to fully electrify. Ford won those years.

The companies that said “let’s wait and see how this develops” got crushed by competitors who moved fast.

Build Trust Through Service

In rural America, nobody knew Richard Sears. Why would they mail money to a stranger for products they couldn’t see or touch?

Sears solved this with fanatical customer service. Money-back guarantees. Free trials. His catalog said: “We solicit honest criticism more than orders.”

That trust turned one-time customers into lifetime customers who’d tell their neighbors.

Trust was infrastructure too—invisible but essential.

Vertical Integration Comes Later

Notice the progression:

Sears started by reselling excess inventory. Only after he had revenue and understood his customers did he create private label brands like Kenmore (appliances) and Craftsman (tools).

Ford started with assembly innovations using standard parts. Only after proving the model did he build the River Rouge Complex, controlling everything from raw materials to finished cars.

The pattern: Start lean using other people’s infrastructure. Once you’re profitable and understand your market, then consider building your own infrastructure for competitive advantage.

Reinvest in Innovation

GE established the first industrial research lab in 1900. This wasn’t altruism—it was strategic survival. They knew continuous innovation would create barriers to competition.

Ford nearly bankrupted himself building the River Rouge Complex in 1921, but it made him the most efficient manufacturer for the next two decades.

The companies that extracted profits without reinvesting rarely survived the next infrastructure shift.

Adaptation or Death

Here’s the sobering part: Most “winners” eventually become losers when the next infrastructure shift arrives.

Ford’s lesson: By the mid-1920s, competitors had caught up to his assembly line innovations. They were offering cars in colors other than black, with more comfort features. Ford stubbornly refused to change, insisting the Model T was perfect. He lost market dominance within a few years.

Sears’ lesson: Dominated mail-order retail for decades. When automobiles reduced mail-order demand, they adapted—opening retail stores. By 1929, had 300+ stores. But when the internet arrived? They failed to adapt quickly enough. Filed for bankruptcy in 2018, brought down by Amazon doing to them what they’d done to general stores a century earlier.

The universal truth: The infrastructure that made you successful will eventually be replaced. The question isn’t whether you’ll face disruption—it’s whether you’ll see it coming and adapt before it’s too late.

What This Means for AI: The Second Wave Advantage

So where does that leave us with the great AI spending spree of the 2020s?

Here’s the crucial insight: This is a second wave revolution, not a first wave.

That means the playbook is different. When electricity replaced steam power, it wasn’t scrappy startups leading the charge—it was established industrial giants with deep pockets and hard-earned wisdom from the railway chaos.

Today’s AI follows the same pattern. Microsoft, Google, Amazon, and Meta aren’t desperate dot-com wannabes. They’re second wave players who watched the internet revolution unfold, learned its lessons, and now control the infrastructure everyone else needs.

This creates a completely different competitive landscape. Let me walk you through how this pattern is likely to unfold:

Phase 1: The Mania (2022-2027)

This is where we are now. Massive spending, unclear ROI, lots of debt. Microsoft, Google, Amazon, Meta pouring billions into GPUs, data centers, and AI research. Hundreds of startups chasing every conceivable AI application.

It feels chaotic because it is chaotic. That’s normal for infrastructure revolutions.

Most people focus on the inefficiency—“Why are we giving everyone access to expensive AI for trivial questions?” But remember Railway Mania: the inefficiency is part of the process. It accelerates deployment and learning.

Right now, companies are figuring out what works. Most experiments will fail. That’s how you find the few approaches that succeed.

Phase 2: The Shakeout (2027-2032)

Some major debt defaults are coming. Not all AI infrastructure companies will survive. We’ll see acquisitions and bankruptcies.

But—and this is crucial—the infrastructure remains valuable.

Those data centers don’t vanish. The trained models don’t disappear. The research knowledge stays in the world. Someone will buy the assets and use them, just like bankrupt railway companies’ tracks were bought and operated by survivors.

Expect 3-5 mega-corporations to emerge as dominant. Probably some combination of the current tech giants plus maybe one or two newcomers who positioned themselves perfectly.

Phase 3: The Consolidation (2032-2040)

The survivors buy bankrupt competitors’ assets at fire-sale prices. Government regulation arrives—late, as always, but probably focused on preventing monopoly abuse and ensuring public access.

AI transitions from “exciting technology” to “essential infrastructure.” Companies stop competing on raw model capability and start competing on specialized applications and services.

Market barriers solidify around dominant players who control the base infrastructure. It becomes extremely difficult for new entrants to compete.

Phase 4: The Commodity Era (2040+)

AI becomes everywhere, boring, essential. You notice it only when it doesn’t work. Like electricity today.

New innovations happen on top of the AI infrastructure rather than in the infrastructure itself. The next revolution begins—quantum computing? biological computing? We won’t know until we’re in the middle of it.

The Second Wave Survival Guide: Competing Against Giants

Here’s the uncomfortable truth about second wave revolutions: Most of the infrastructure gets built by giants, not startups.

During electrification, you didn’t compete with General Electric by building power plants. You won by being brilliant at using electricity before your competitors figured it out.

The same applies today. Microsoft and Google are spending hundreds of billions building AI infrastructure. You’re not going to out-infrastructure them. But you can absolutely out-execute them at specific applications.

The Sears Strategy: Leverage Infrastructure Aggressively

Remember Richard Sears? He didn’t build railways or the postal system. He just used them more cleverly than anyone else.

Today’s equivalent: Don’t try to build better AI models. Use existing AI infrastructure to solve expensive problems in ways that established giants can’t or won’t.

Example: While Google builds general-purpose AI, you build AI specifically for veterinary diagnostics. While Meta focuses on social media AI, you create AI for commercial real estate evaluation. The giants provide the computing power—you provide the domain expertise and specialized focus.

The Ford Strategy: Efficiency Through Specialization

Ford didn’t invent the automobile or even the assembly line. He perfected efficient automobile manufacturing using electrical infrastructure better than competitors.

Today’s equivalent: Take expensive human processes and make them dramatically more efficient using AI infrastructure that already exists.

The key insight: Giants optimize for scale and generality. You optimize for specific efficiency gains in narrow domains where you can measure ROI precisely and move faster than large bureaucracies.

Why Second Wave Timing Matters

Here’s what most people miss: Second wave timing is more forgiving than first wave timing.

During the dot-com era, being six months late meant your competitor got all the venture funding and market share. But second wave infrastructure gets built gradually and predictably by established companies following quarterly earnings schedules.

You can watch Microsoft’s AI investments, Google’s model releases, Amazon’s infrastructure buildout—and time your applications accordingly. The infrastructure providers want you to succeed because your success validates their infrastructure investments.

The Cash Flow Reality

Unlike dot-com startups that burned VC money until they died, successful second wave companies generate revenue while the infrastructure improves.

Start where AI already works profitably. Expand as the infrastructure gets better and cheaper. You don’t need to bet your company on AI infrastructure that doesn’t exist yet—you can build sustainable businesses on AI infrastructure that’s deployed today.

This is exactly what Sears did with railways and Ford did with electricity: Start profitable, stay profitable, and grow with the infrastructure rather than ahead of it.

The Patterns That Never Change

After this journey through 180 years of infrastructure revolutions, here’s what I keep coming back to:

Greed drives overbuilding → Society benefits from infrastructure → Financial players lose money → Infrastructure remains valuable

It’s wasteful. It’s chaotic. Fortunes are made and lost. But somehow, civilization ends up with valuable infrastructure that gets used for decades.

First movers rarely win → Cash-rich survivors acquire the pieces → Monopolies form → Government intervenes too late

The companies making headlines during the mania phase usually aren’t the long-term winners. It’s the ones with deep pockets who survive the shakeout and buy assets at bankruptcy auctions.

Every revolution looks “different this time” but follows the same cycle

Railway Mania, electrification, automobiles, telephones, radio, television, computers, internet, mobile—every single one felt unprecedented while it was happening. Every single one followed roughly the same pattern.

Essential technologies always become regulated utilities eventually

Private profits in growth phase. Public control or regulation in maturity phase. The timeline varies, but the destination is predictable.

A Final Thought: Why Second Waves Matter

You asked me about AI infrastructure and whether we’re in a bubble. The answer is yes—and understanding what kind of bubble makes all the difference.

This isn’t the dot-com bubble 2.0. This is the electricity bubble of our time—a second wave infrastructure revolution driven by established giants who learned from the first wave’s mistakes.

Railway Mania gave Britain the transportation backbone that powered the Victorian era. The electrification “waste” gave us the industrial economy. The dot-com crash left us fiber optic networks that enabled streaming.

The AI bubble will leave us something equally transformative: Intelligence infrastructure. Data centers that think. Models that reason. Research advances that compound for decades.

But here’s what the historical pattern teaches us: Second wave bubbles are different. They’re more methodical, more survivable, and they create more lasting value because they’re built by companies that can weather the storms.

When the AI shakeout comes—and it will come—Microsoft, Google, and Amazon won’t vanish like Pets.com did. They’ll consolidate, adapt, and emerge stronger. The infrastructure will remain. The capabilities will endure.

And unlike the dot-com era where only technical experts could build websites, the AI era will democratize intelligence itself. That’s what second waves do—they take revolutionary technology and make it ordinary, accessible, boring.

I suspect we’ll realize the true impact around 2030, when AI has become so embedded in everything we do that we barely notice it anymore.

That’s when you know a second wave revolution succeeded—when the miracle becomes mundane, and the infrastructure becomes invisible.

---

(Written by Human, improved using AI where applicable.)