The More Successful AI Becomes, the More Likely Investors Are to Lose Money: Century-Long Lessons from Railroad, Oil, and Fiber Optic Bubbles
Author: Packy McCormick
Translated by: TechFlow
TechFlow Summary: From the British railway mania, early oil gold rushes, to the late-century fiber broadband bubble, the more miraculous a technology is and the more it lends itself to expansion through heavy capital investment, the more likely it is to attract surges of capital, ultimately resulting in severe oversupply and a crash in prices.
Faced with Anthropic’s trillion-dollar valuation and hundreds-of-billions-worth bets on compute capacity by major tech giants, Packy McCormick draws on several eras of business history to argue that miraculous technologies never automatically translate into profitable ventures. When the capital tide recedes and consumers pocket 98% of the gains, only those who build genuine moats survive.
Dismissing something as a “bubble” is neither constructive nor dignified. Those who endlessly warn of bubbles are mostly pessimistic killjoys: having missed out on the frenzy themselves, they dismiss foundational technologies as overhyped and assume their adherents are mindless followers.
Moreover, sounding the alarm about a tech bubble runs entirely counter to my usual stance. Anthropic is very likely to see its market cap surge well above its IPO price by the end of its first week, potentially gobbling up an entire economic sector and becoming the first company in human history to reach a ten-trillion-dollar valuation.
The future economic landscape could indeed fracture into an extreme “K-shaped” split, much like the letter K written by my child: a tiny elite of Anthropic employees will control prime real estate across the San Francisco Bay Area.
And the rest of us will have to bow respectfully to Claude, hoping the “benevolent machine” keeps dispensing crumbs of prosperity (unless you manage to front-run the Duet prediction markets on Kalshi to cover your basic expenses).
While I previously wrote that AI is more like oil than divinity (which looks remarkably prescient looking back at that article’s cover art), and I still hold that view, the objective facts are equally clear:
Oil cannot refine itself; you can’t use oil to recreate Microsoft Office; and it hasn’t even managed to solve Navier-Stokes equations.
History typically unfolds along one of two opposite narratives. In the first scenario, years from now people will look back at Anthropic’s IPO pricing with disdain: How did we ever justify paying roughly $1.4 trillion for an old-school general-purpose large language model?
Perhaps there isn’t another S-curve growth trajectory as steep as the software boom after programming. Looking backwards, computers were always good at “speaking computer languages”—an ability that may not generalize to everything.
Or perhaps such a curve does exist, but a brutal pincer attack from OpenAI, Google, SpaceX AI, Meta, Chinese open-source models, and emerging labs has slowed revenue growth, forcing it to sell to Amazon or Apple under the weight of compute contracts signed during the boom.
In the second scenario, Dario decides that outsiders’ mockery has hurt the models’ feelings, so he outright bans everyone from laughing, given that he himself now embodies the entire economy.
The bull case for Anthropic is actually quite straightforward: It holds the best models, which are currently self-improving at an unprecedented pace, widening the gap to competitors.
This is likely a winner-take-all market unlike any single-track industry in the past. This time, the winner-take-all dynamic extends across all sectors. Whether it applies smart models to productivity software or pharmaceutical R&D, it wins in both cases.
Anyone claiming they know with absolute certainty which of these two scenarios will come to pass is talking nonsense. I have plenty of flaws, but I hope I’m not one of them.
The Paradox of Anti-Bubble Narratives: The First Three Steps Are Actually the Starting Line of the Cycle
Nevertheless, three arguments against the “AI is not a bubble” thesis are currently circulating widely. They all sound plausible and well-reasoned, yet perfectly align with the classic precursors to every major bubble in history.
To clarify: I’m temporarily borrowing the term “bubble” as a blanket label, but what I’m actually referring to is closer to a “commodity supply cycle,” not pure speculative mania like tulip crazes where prices completely detach from reality. These three classic rebuttals are:
• Unlike past bubbles (lol, WebVan and certain meme coins), AI is an incredibly useful technology being widely and efficiently adopted in the real economy.
• These top companies are generating genuine revenue, growing at terrifying speeds.
• Top labs have already proven profitability; reported losses are simply intentional investments made to capture future incremental demand.
Even if all three premises hold true, they precisely demonstrate that the current situation is the perfect incubator for generating—and eventually bursting—a bubble. The full script usually plays out like this:
1. A miraculous disruptive technology emerges; 2. Early supply is extremely scarce, allowing first-mover suppliers to rake in profits;
3. Early metrics make it instantly obvious that consumer demand seems endless.
4. High margins and rigid demand attract massive capital and competitor influx; 5. All players simultaneously scale capacity. Even knowing oversupply is coming, they lock in core resources at scarcity premium prices to launch first and seize initial advantage.
Then the script continues: 6. Massive supply comes online all at once, further technology iteration lowers barriers, and product prices crash overnight; 7. Capital suffers huge losses and exits, ordinary consumers reap the dividends, and the technological train rolls forward unscathed.
8. Capital panic drives a stampede for the exits, new entrepreneurs can’t raise funding; 9. A handful of survivors or external giants with deep pockets acquire the wreckage en masse, becoming ultra-low-cost suppliers; 10. Real demand finally catches up with excessive supply, and the ultimate victor ends up larger than the sum of all fallen failures.
All of this is hardly novel. It almost perfectly maps onto Carlota Perez’s framework of technological revolutions and financial capital and Edward Chancellor’s capital cycle theory. Byrne Hobart and Tobias Huber even wrote an excellent book praising the historical role bubbles play in breaking growth stagnation.
The truly fascinating part is that the three core pillars of today’s anti-bubble narrative coincide exactly with the first three steps of this classic ten-step cycle.
The reason this playbook repeats across generations is precisely because every wave of new technology is genuinely astonishing enough that those living through it sincerely believe: “This time, it’s different.”
The first seven steps align with ironclad laws of human nature and capital markets with near mathematical precision. Steps eight through ten also frequently unfold, sometimes evolving into consumers feasting directly on the “carcasses” of an entire industry as they enjoy technological dividends.
The Aviation Miracle: Benefiting Humanity While Wiping Out Shareholders
Take commercial aviation as an example. If there’s anything in human history comparable to magical breakthrough technologies that let humans fly through the sky like birds, this definitely ranks at the top.
Before flight was born, the idea seemed so absurd that even The New York Times infamously predicted humanity wouldn’t achieve powered flight for another one to ten million years—just over two months before the Wright brothers’ successful test flights.
So here’s the question: Did the Wright brothers become obscenely wealthy? Did the airlines that put live passengers into the blue skies ultimately prove to be enviable, lucrative businesses?
Berkshire Hathaway founder Warren Buffett cut straight to the chase at the 1999 Sun Valley conference: “Karl Marx did less damage to capitalists than Orville Wright.”
In his 2007 shareholder letter, Buffett revealed the business logic behind that sharp jab: “The worst kinds of businesses are ones that grow rapidly, require enormous capital commitments just to sustain that growth, and ultimately earn almost nothing.”
“Think about the airline industry. Since the days of the Wright brothers, lasting competitive advantages in this sector have been as elusive as mirages. Honestly, if a visionary capitalist had stood on Kitty Hawk beach back then, he would have done posterity a massive favor by shooting Orville down on the spot.”
Uttering that line today might not be popular, but zooming out on the timeline reveals a fundamental truth: For a company to endure forever, it needs two things simultaneously: products customers desperately want to buy, and lasting competitive advantages that fend off rivals hungry for margin share. Businesses need moats, and airlines precisely do not.
Ironically, even Buffett himself couldn’t resist temptation. In 2016, Berkshire spent roughly $7 billion to $8 billion taking massive concentrated positions in American Airlines, Delta Air Lines, Southwest Airlines, and United Airlines.
At Berkshire’s 2017 annual meeting, he defended the move: “This is a fiercely competitive industry. The question is whether it remains as deadly as it used to be.”
By April 2020, he painfully liquidated all four holdings. Granted, the pandemic black swan was the direct trigger, but his actual decision framework still boiled down to supply-and-demand laws: “Even if passenger traffic rebounds to 70% or 80%, all those idle planes won’t magically vanish.”
Therefore, not every great industry built on miraculous technologies and riding massive tailwinds can spawn enduring money-printing machines. But without exception, their long-term impact on human society vastly exceeds even the most aggressive early forecasts: consumers win handily, and humanity takes a giant leap forward.
Ultimately, there is no more optimistic attitude toward life than this: As long as humans devise any new miraculous technology, profit-driven capital will always flood in to create severe oversupply, driving that miracle down to commodity prices so that everyone benefits. That is precisely the definition of “miracles without moats.”
The British Railway Mania: Knowing Overinvestment Was Coming Yet Unable to Stop
What drives this pattern isn’t stupid capital—in fact, it’s the exact opposite. Only genuinely earth-shattering technologies can mobilize global capitalists to pour sums representing a significant portion of world GDP into the same track.
This applies to oil, railways, automobiles, and the internet. Their ultimate societal impact all far exceeded early prophets’ predictions, much like how oil initially replaced whale oil in lamps, later propelling the entire modern automotive and aviation civilizations.
And insiders weren’t blind to the fact that the industry was careening toward overinvestment and digging their own graves. But from a game theory perspective, what else could a player at the table do? The British railway frenzy of the 19th century is a perfect case study.
On January 26, 1846—years before I was born, curiously on my birthday—Charles Russell, chairman of the Great Western Railway, stood before the British Parliament urging colleagues to “expose and rigorously curb all speculative railway schemes,” calling on the government to slow railway approvals.
Just four years earlier in 1842, early UK railways like those under Russell’s banner were literal printing presses. Pioneer lines like Liverpool & Manchester and London & Birmingham routinely paid staggering dividends of at least 10%, while the slightly less stellar Great Western still handed out 6% to 8%.
Rail freight and passenger demand seemed bottomless at the time. Due to difficult construction and uncertain prospects, only a handful of trunk lines enjoyed this windfall. But within the next two or three years, countless credible and wildly speculative promoters submitted hundreds of new railway proposals, everyone desperate for a slice of early supernormal profits.
By autumn 1845, Britain had succumbed to railway fever. The Railway News ran editorials attacking The Times for nine consecutive weeks, solely because The Times kept warning that railway stock speculation had become dangerously bubbly.
On November 17, The Times published a full-page special edition calculating the astronomical capital required for 1,263 proposed railway projects. And yet? The British Parliament still approved a staggering 4,540 miles of new track in a single session in 1846.
Charles Russell desperately wanted to “curb all speculative schemes,” because every newly approved route introduced a fresh competitor that would crush freight rates. Ironically, however, he was no conservative investor opposing rail expansion—he was fully betting on it himself.
At the 1848 Paddington shareholders’ meeting, facing questions about why the Great Western itself was mortgaging heavily to aggressively buy up surrounding branch lines, The Times recorded his classic reply: “It is for self-defense.”
Outsider challengers have nothing to lose, while incumbent giants live in terror of losing everything. Two entirely different fears ultimately push everyone at the table toward the same outcome: You must spend recklessly to expand at any cost.
This is exactly how Google thinks today. Founder Larry Page reportedly said plainly: “I’d rather go bankrupt than lose this race.” CEO Sundar Pichai also explicitly stated to Wall Street analysts during the Q2 2024 earnings call: “For us, the risk of investing too little far outweighs the risk of investing too much.”
Mark Zuckerberg feels the exact same pressure. In September 2025, during an interview with Access podcast host Alex Heath, he conceded that “there is absolutely possibility that AI is a bubble.”
He added: “If we end up misallocating a few hundred billion dollars, that would certainly be regrettable; but I’d argue the risk on the other side of the scale is higher. For a company like Meta, the risk of not moving aggressively outweighs the risk of moving slightly too aggressively.”
History rhymes beautifully.
Earlier this year, Anthropic CEO Dario Amodei told Dwarkesh Patel that he firmly believed a “nation of geniuses in data centers” was imminent, but also admitted “if you miscalculate the timing by even a single year, you’ll destroy yourselves completely.”
He immediately laid out the balance sheet clearly: “I could easily decide to procure $1 trillion worth of compute annually starting in late 2027. But what if our revenue at that point doesn’t hit $1 trillion—even if it’s $800 billion? There is literally no force or hedging instrument on Earth that can save me from bankruptcy when saddled with that many compute contracts.”
At the time, Anthropic’s forward compute commitments hovered around a relatively restrained $100 billion. His remarks subtly mocked competitors (ahem, pointing at OpenAI) for being reckless: “I always feel some peers haven’t properly run their Excel models, or don’t fully grasp the magnitude of the risk they’re carrying.”
Fascinatingly, the very next week after that interview, OpenAI reset its guidance with investors, cutting its targeted compute spend by 2030 from a previously publicized $1.4 trillion down to around $600 billion (whatever, big numbers blur together anyway).
But checking the latest S-1 filing reveals Anthropic itself has already committed to compute contracts totaling a staggering $900.9 billion through 2045, with roughly half of that maturing before the end of this century.
Yet this might actually be the only correct chess move right now! Five reasons: One, who am I to negate it; Two, its revenue is indeed skyrocketing; Three, if we don’t sign, a competitor who doesn’t do math will blindly jump in anyway.
Four, Anthropic literally built its own spreadsheet software; Five, the underlying calculation decisions are undoubtedly made by smart executives paired with frontier models that are “too smart to safely disclose publicly.” Plus, Anthropic booked operating profits in Q2, and turned a profit in Q3 even after excluding ~$2 billion in non-cash expenses.
Only Dario himself admits how razor-edge this thing is, and it sounds suspiciously similar to the British Railway Mania, the American Railway Boom, and even… North American cannabis cultivation fever. Indeed, many North American marijuana growers ruined by severe oversupply are now converting greenhouses into AI data centers to survive.
The Fiber Optic Broadband Bubble: Real Explosion, Unavoidable Pricing Crash
Not to mention the bubble at the turn of the millennium.
Modern commentators often scoff at linking the 2000 dot-com crash to today’s AI boom, arguing that back then, any company slapping “.com” onto its name without revenue would instantly surge 420%—pure retail investor mania.
But we really should separate “internet tech stock euphoria” from the “broadband infrastructure bubble.” The latter essentially refers to pre-building infrastructure capacity driven by early surging demand signals.
Looking back through hindsight makes it seem comical, but every telecom company frantically laying underground fiber optics back then was making highly rational business decisions from their independent perspectives.
Internet traffic was indeed exploding. In 1990, almost nobody in the US used the internet; by 1995, penetration hit 9%; and just five years later in 2000, that figure jumped to 43%.
That’s just US domestic data. The rest of the world connected soon after, spawning wave after wave of compounding growth curves. Besides, bandwidth experience was stuck in the ultra-low-speed dial-up era!
I personally endured the torturous snail-speed of early dial-up: loading a static image took a full minute, downloading an MP3 took twenty minutes or indefinite waiting, and you constantly prayed no one in the house would suddenly pick up a landline to kick you offline.
Imagine what happens if everyone in the future can access virtually unlimited high-speed bandwidth on demand: unimaginably new use cases would emerge. People would stream shows flawlessly, attend cloud meetings, trade goods without leaving home, and even spawn internet species unimaginable at the time.
These weren’t random geeks envisioning this; even Bill Gates, sitting on his throne, made almost verbatim predictions in his 1995 #1 New York Times bestseller The Road Ahead. You can click the excerpt to see it yourself. Gates nailed it, and frankly, those bold visions seem almost conservative compared to how things actually unfolded.
Millions shared that faith. Long-haul fiber giant Qwest Communications even ran a classic ad asking audiences to imagine a magical motel: “Every room can order any movie ever filmed in any language, at any time, day or night.” That dream hotel was the high-speed internet running directly into millions of bedrooms.
In that climate, nobody could foresee where internet traffic ceilings would emerge. In 1998, the US Commerce Department even released a heavyweight report Emerging Digital Economy, authoritatively declaring global internet traffic was “doubling every 100 days.”
Later, scientist Andrew Odlyzko traced the number back and found it was quoted from a 1997 Inktomi whitepaper, which itself cited WorldCom subsidiary UUNET’s chief scientist Mike O’Dell.
The funniest part? O’Dell originally meant WorldCom’s own fiber capacity expansion rate (supply-side growth), which got misreported and morphed into actual user demand growth!
But back then in the frenzy, did anyone care about such pedantry? The internet was barreling forward regardless. At a public conference in 1998, when investors worried that peers simultaneously digging trenches for fiber might cause oversupply, Qwest’s then-CEO Joseph Nacchio snapped back:
“The explosive demand driven by the internet’s phenomenal adoption absolutely guarantees there will never be a bandwidth glut.” He confidently added: “As long as our executive team is merely competent, we will inevitably succeed brilliantly.”
Originally spun off from Southern Pacific Railroad’s department, laying fiber alongside existing rights-of-way, Qwest thrived in its first few years.
Thanks to a pricey acquisition of long-distance carrier LCI in June 1998, Qwest’s reported revenue climbed from $2.24 billion in 1998 to $3.93 billion in 1999, turning an operating loss of $754 million into a net profit of $324 million.
Of course, post-fact SEC investigations revealed that polished financials contained some fabrication fluff. By peak 2000, the company’s market cap once hit $91 billion; but fortune didn’t last. By August 2002, Qwest’s market cap was bloodied down to a paltry $1.9 billion.
Chart: Qwest historical stock price and market cap crash trajectory, source companiesmarketcap.com
What happened in between? You’re comfortably reading this article via the internet right now, likely with at least two networked devices within arm’s length, possibly swearing repeatedly to quit screens due to addiction.
The real widespread adoption of the internet undeniably achieved a massive explosion. After surging 10x in 1996 alone, global traffic maintained extremely high-frequency growth, roughly doubling every year. The only problem is, it didn’t double every 100 days as the “mistranslated” Commerce Department report boasted.
Chart: Comparison of actual US internet backbone traffic and "doubles every 100 days" forecast from 1990 to 2002, data sourced from Coffman & Odlyzko
Blisteringly contrasting supply-side leaps occurred. Stimulated by the 1996 Telecommunications Act relaxing regulations, total US long-haul fiber mileage doubled from 1996 to 2001 (surging from 100,000 miles to 200,000 miles).
Meanwhile, electronic equipment upgrades multiplied single-fiber transmission capacity by up to 40x (from 2.5 Gbps to 100 Gbps). Telecom giants collectively poured ~$500 billion into frantically building optical networks.
Under the weight of massive new supply, a 2002 TeleGeography research report showed lease prices for main transatlantic/transpacific bandwidth routes plummeted cliff-like by ~70% annually for three consecutive years. In 2000, a 155 Mbps dedicated line between New York and LA commanded an annual rent of $1.8 million; by early 2002 it fell below $150,000.
Granted, humanity eventually fully digested that prematurely constructed fiber, and our appetite for bandwidth now vastly exceeds expectations; but that brings zero comfort to the infrastructure pioneers who paid upfront.
WorldCom filed for bankruptcy amid accounting fraud scandals in 2002; Williams Communications collapsed that same year under $5.9 billion in debt; Global Crossing similarly declared bankruptcy, reducing debt from $11 billion to $2 billion during 2003 restructuring before clawing its way out, while original shareholders went to zero.
Cruelly enough, Global Crossing’s traffic exploded again by 130% in its first year post-bankruptcy restructuring. Too bad bandwidth had already become worthless—the company’s annual revenue dropped another 6%.
Even the sharpest-eyed ultimate winner Level 3 didn’t walk away scot-free. It raised a war chest right before closing in February 2000, securing extreme unit-cost advantages by pre-laying 12 ducts. Post-crash, it secured Buffett capital and swallowed rivals cheap.
Still, even then, Level 3’s sale to CenturyLink in 2017 closed at a price 97% lower than its peak. Even if you successfully pick the winning horse early in a bubble, as long as your entry price priced in delusions of ignoring competition, the outcome may still amount to nothing.
With that said, returning to human-game dynamics: Could you have resisted jumping into this round back then? The internet is undeniably a miracle. Via screens, you can video-call anyone across the globe, tether to genius hubs in data centers via copper wires, shop and watch films without stepping outside. It’s hard to imagine a world without the internet. God bless the internet!
Early Oil Rush: How Rockefeller Smiled Last
Those frantic fiber-laying telecom tycoons merely moved too aggressively. They weren’t fools, just as the early Pennsylvania Titusville oil drillers flooding fields in 1860 weren’t stupid either.
A year prior, Colonel Drake drilled the region’s first industrial oil well (ironically dying destitute in his later years). Initially crude sold for a solid $20 a barrel.
But once news spread and derricks covered the hillsides, oil prices crashed to 50 cents by June 1861, halving again to a pitiful 10 cents by December of that year, instantly wiping out batches of early blind-drillers.
Chart: Rollercoaster trajectory of early crude prices 1860–1865, data from Appendix 57 of Ida Tarbell’s 1904 History of the Standard Oil Company
The collapse then wasn’t fundamentally due to lack of demand. Even in those earliest days when crude served merely as lamp fuel, societal demand for oil was practically bottomless.
After the 1861 gusher stampede triggered a panic crash, depleted older wells restricted supply while exploding demand pushed prices back above $8/barrel in 1864, immediately attracting more frantic drillers and wilder speculative froth.
In the Pennsylvania oil town of Pithole, once-worthless wasteland farmland fetched $1.3 million in July 1865, skyrocketing to $2 million upon resale by September. On Christmas Eve that same year, the desolate wilderness briefly housed nearly 20,000 residents.
But the tide receded shockingly fast: By the 1870 census, only 237 souls remained. By 1878, that once-$2-million golden plot had been reclaimed by the town council for a mere $4.37 in tax penalties.
Chart: Raw frontier expansion: The Pithole oil field at its peak in 1865
According to famed circus magnate PT Barnum’s 1865 publication The Great American Fraud, roughly 600 petroleum companies attempted to raise a combined $500 million at the time.
Barnum wryly commented: “Petroleum truly exists, undeniably adding major wealth to our nation; but the commercial hype peddling that oil market is itself a genuine swindle.”
Early speculators weren’t wrong about the grand prospects of the oil sector: Petroleum genuinely changed the world, later even seeing a single titan dwarfing the entire industry’s $500 million initial fundraising by orders of magnitude. Early investors correctly sized the market pie but completely ignored brutal industry competitive dynamics.
But one person absolutely didn’t make that mistake: John D. Rockefeller. If we accept Ida Tarbell’s accounts, President Theodore Roosevelt’s condemnation, and the DOJ’s historical charges, this oil baron was arguably overly sensitive to competitive dynamics.
As we recap in Power in the Age of Intelligence, Rockefeller entered Cleveland’s refining scene in 1863 and founded Standard Oil in 1870, spotting the systemic bottleneck from the start: the poison killing long-term oil profitability was violent cyclical volatility.
Pacifying volatility required extreme scale and chain control, while obtaining control demanded crushing costs to steal share. Consequently, he optimized kerosene yields, repurposed discarded waste streams, and ruthlessly executed vertical integration.
By the 1880s, Standard Oil momentarily controlled up to 90% of US refining capacity. Even facing antitrust breakup dawn in 1911, it still dominated 64% of the market. Its ~$1.1 billion valuation then equaled 6% of the entire US stock market, exceeding twice the entire 1865 US oil sector bubble’s total fundraising.
Others chased short-term supernormal profits; Rockefeller obsessively fortified durable moats. Ultimately, not only did Rockefeller win massively, ordinary consumers became huge beneficiaries too. Until today, industrial civilization still runs on refined oil; only those countless pioneer rivals who perfectly predicted the track but exhausted ammo in the meat-grinder perished halfway.
Three Moats Surviving Oversupply, and the Golden Age of Consumer Surplus
Rockefeller epitomizes the type of entrepreneur who ultimately stands atop the wave: He foresaw inevitable oversupply and positioned himself to leverage it to cull rivals and harvest the entire market.
Swap to a completely different commercial context, and auto mogul Henry Ford did the exact same thing. Before America’s auto industry kicked off, ~2,000 carmakers emerged. During 1900–1919, as US household car ownership skyrocketed from 0% to 1/3, nearly all those 2,000 vanished.
By the early 1940s, when half of American families drove cars, 90% market share belonged to the Big Three: Ford, GM, and Chrysler.
Ford long ago figured out: In inevitable oversupply battles, only overwhelming quality paired with lowest-ever industry prices to furiously scale volume, using greater scale to further crush costs and prices, serves as the only path to victory.
Look at railroad titan Cornelius Vanderbilt, hailed by biographer TJ Stiles as “the first magnate.” This top US railroad winner harbored zero romance for tracks; his lifelong goal was singular: Use the most efficient tools to move people and goods faster, cheaper, from A to B.
He started with sailboats, struck gold with steamships, then invaded rails. Whether water or land transport, he fixed his gaze on corridors combining speed and cost dominance: First carving out the Nicaragua route to shave hundreds of miles off Panama’s competitor line, forcing rivals to pay heavily to buy him out.
Sensing rails would replace waterways, he bluntly donated his largest steamship to the Federal Navy, exited shipping entirely, and re-bet everything on rails. He picked distressed assets with Manhattan’s core scarce rights-of-way like the Harlem Railroad, slashed redundant costs, and extended along the Hudson River corridor to Chicago; Amtrak’s mainline still uses that exact corridor today.
Vanderbilt随时准备 switched tech carriers, but clung tightly to only “lasting competitive advantages.” While mass peers who similarly bet “steamships and rails will change the world” went bust, he left ~$100 million in net worth at his death in 1877.
Sounds like a slightly more respectable Silicon Valley seed round today, but by contemporary monetary scale: If cashed out entirely on his final day, he would instantly withdraw one-twentieth of all circulating US currency.
Zooming out to today, ordinary Americans enjoy relatively cheap airfare, efficient logistics, private vehicles, affordable refined products, and ubiquitous high-speed internet daily. Even though early trailblazers who bled sweat and cash to drive these resources into commoditized bargain prices were mostly wiped out, that represents capitalism’s coldest yet most romantic undertone.
Yale economist William Nordhaus quantifies in Schumpeterian Profits in the American Economy: “Between 1948 and 2001, only a minuscule fraction of total social returns generated by technological innovation stayed with producers; the vast majority flowed to the masses.”
His specific figures are even more stunning: Innovators captured cumulatively only ~2.2% of total social surplus. A staggering 97% to 98% transformed into lower prices, better products, and rising real incomes for residents. Just thinking about it gets one’s blood pumping.
A 2004 NBER paper abstract precisely captures Nordhaus’s diagnosis of the dot-com crash: “New economy bubbles stemmed from investors overestimating innovators’ ability to exclusively monetize results within vertical tracks.” In other words, what got built back then wasn’t valueless—investors just severely overestimated what percentage they could skim into their own pockets.
Nothing new under the sun. Almost every promising revolutionary technology that ultimately changes the world undergoes this cycle baptism. It’s practically hardened into a testing threshold for verifying “whether it’s truly revolutionary,” aside from extremely rare exceptions.
As analyzed in When to Dig a Moat, startups must build defensive walls before certainty evaporates. But once entering open-info domains like oil or AI where “everyone instantly sees massive, certain demand,” the fallacy that “build it and they will come” immediately ignites cutthroat industry-wide slaughter.
In highly deterministic demand tracks, historical exceptions surviving internal competition almost entirely relied on franchise-based barriers: pharma, monopoly routes pre-1978 deregulation, Bell Telephone pre-patent expiry in 1894, and ALCOA holding exclusive electrolytic aluminum patents.
But all these belong to Cornered Resources sustained by exclusive rights. Once protection loosens or expires, floodwaters of competition still pour in, instantly replaying history scripts.
The remaining marginal survival cases either hide obscurely unrecognized initially (like Amazon refusing to separately list AWS profits for nine years, with cloud pioneers dying in droves before), or stem from decades-polished “super craft moats,” much like ASML’s lonely march.
In plain terms, wherever objective conditions dictate you “can’t simply throw money at it to quickly jump in,” some moat breathing space remains. Everywhere else, if you can just throw money at it to play, the whole domain gets thrown into Darwinian jungle rules, endlessly tearing at each other. Given a sufficiently long window, jungle rules forever stand on consumers’ side.
One of the primary reasons investors are currently most fervently obsessed with AI is precisely because “this industry truly suits throwing money to manufacture miracles.” Worse, AI renders bypassing and cloning many traditional software/tech moats ridiculously easy, guaranteeing pushing the aforementioned oversupply cycle tenfold faster across all industries.
Even Anthropic itself is using cutting-edge Claude models to assist punching through NVIDIA’s proudly guarded CUDA ecosystem moat, enabling smoother switching away from this global GPU compute monopoly boss later on.
Condensing thousands of words, I firmly believe humanity is striding into an unprecedented “Golden Age of Consumer Surplus.” Although this piece focuses on AI, from embodied robots, energy grids, advanced manufacturing to biopharma, every sector currently faces breathtaking technological explosions.
Many once sci-fi fantasy scenarios (flying cars to home robots) now attract countless sharp, well-capitalized teams charging forward simultaneously. More often than not, suspense shifts from “can it be done?” to “when will it land, and who ultimately profits?”
Even clinical-stage novel anti-cancer therapies, once considered divine medical breakthroughs, now regularly get overtaken by fresher tech pathways within months.
As an ordinary human and everyday consumer, witnessing and living in this era of relentless miracles fills me with profound excitement.
Switching back to an investor lens… this is undoubtedly the most awe-inspiring yet exhilarating investment age in business history. In my view, markets do host a batch of companies with specific DNA. Regardless of whether this super-boom rides high momentum, or one day capital chills triggering a fire-sale discount dump of miracle technologies, they’ll consistently stand on the winners’ side.
I particularly favor integrators with Vanderbilt, Rockefeller, or Ford bones in their spines: Of course the journey gets bumpy, but whenever stronger, cheaper models or technologies surface, they seamlessly pivot, delivering sharper products at lower costs.
The internet age let network effects win peacefully; but as new technologies resoundingly walk back into physical reality, extreme economies of scale and concrete absolute cost advantages will reclaim the crown.
Interestingly, I just finished writing those words and scrolled across social media to find a tweet quoting an Economist heavy-hitter feature: 'Could AI’s Leading Men Become as Powerful as Ford or Rockefeller?'
Check the top-ranked figures, and results align almost perfectly with this article’s deduction: Leading the pack are industrial oligarchs forging extreme cost moats via M&A post-bust, lone heroes nailing vertical integration (Elon Musk), followed by Zuckerberg, Bezos, and Google founders riding oversupply cheap fiber.
Compared to those, the Wright brothers, the great inventors who first lifted humanity to the skies, languish ranked 50th and 51st on this power list.
Chart: Economists’ Historical Business Titans Power Rankings, reposted from Callum Williams' tweet
As for how Anthropic’s IPO ultimately plays out, or whether the entire AI macro-cycle is a bubble? Honestly, I’m completely clueless. Whether inside AI itself or broader physical battlegrounds outside, AI both accelerates rivals mimicking and dismantling walls at supersonic speeds, and helps leaders snowball dominance at unprecedented paces.
Ultimately, no matter how futuristic wrapping a technology wears, deciding whether a commercial company and its backers actually pocket genuine cash from its universal adoption boils down to two primal iron laws: Whether you truly possess unassailable, lasting competitive advantages, and exactly what chips you staked at entry.
Remember this: The closer your technology approaches miracle status, the deeper your moat must be dug—deeper than life itself.