Investing Classics: Tech & the InternetTech, the Internet, and the Economics of Digital Moats

Distinguish network effects from mere scale: scale brings a cost advantage; network effects make the product itself better as more users join (Metcalfe's Law: value scales roughly with the square of the user count). They come in two forms: direct (social, messaging) and indirect/two-sided (payments, app stores, marketplaces — more on one side attracts the other). Past the tipping point, the market often goes "winner-take-all." But the key point is that this force is symmetric: once a network starts to bleed users, the collapse can be as fast as the rise.

In 2006 MySpace was the most-visited website in the U.S.; it had been bought by News Corp in 2005 for about $580 million, while Facebook had only just left campus. But Facebook's product experience and real social graph were stronger, and by 2008 it had overtaken MySpace in global monthly users. MySpace changed hands in 2011 for just about $35 million. The same network effect, once reversed, collapses as spectacularly as it once rose.

Network effects fail when: users can cheaply "multi-home" (use WeChat and other apps at once), switching costs are low, subnetworks are fragmented (regional platforms can't go global), or interoperability is mandated by regulators. Misuse: ① treating "many users" as proof of a network effect; ② ignoring reversibility and paying a premium for "eternal monopoly."

• Is this a direct network effect, or the more fragile indirect/two-sided kind?
• Can users and suppliers cheaply use rivals at the same time (multi-homing)?
• Over the past 3 years, is the network densifying or showing signs of attrition?
• Does the premium I'm paying for this moat assume it is "never reversible"?

The market awards big-tech ever higher valuations because a handful of companies genuinely achieved "asset-light + high return on capital + extendable optionality" — compounding spectacularly. But Buffett's 1999 warning still holds: a fast-growing industry doesn't mean shareholders get paid; the durability of the advantage is what counts. As Damodaran puts it in Narrative and Numbers (2017), even the most stirring growth story must eventually land on cash flow. The price you pay itself sets the range of your return.

The 2000 dot-com bubble left three timeless footnotes:

Conversely, excessive "valuation fear" makes you miss true compounding machines — Buffett long avoided tech, missed Amazon, and admitted it was a mistake. The point isn't "too expensive, don't buy," but whether the price you pay already discounts years of near-perfect growth. Misuse: substituting a TAM (total addressable market) story for cash-flow math, treating "growth" as a get-out-of-jail card that overrides price.

• How many years of perfect execution does the current valuation's growth and margin assumption require?
• If growth comes in at half the expectation, does this investment still make money?
• Am I buying "a great company," or "a great company at a reasonable price"?
• Will this company's competitive advantage most likely still be there in 10 years?

Economist Carlota Perez argues in Technological Revolutions and Financial Capital (2002) that each revolution passes through "installation — frenzy — turning point — deployment": financial capital floods in during the frenzy and inflates a bubble, and only after it bursts does real deployment begin. Separate two things: the infrastructure gets built (society benefits) versus the early backers make money (usually not). When an industry is fast-growing, capital-intensive, and competing on undifferentiated products, that is precisely Buffett's "recipe for destroying shareholder value" (he cites early aviation and autos). The rational question is: which link in the chain holds the pricing power?

The telecom and fiber frenzy around 2000: under the story that "internet traffic will grow without limit," carriers sank hundreds of billions into laying fiber, and the vast majority (industry estimates put it at only about 2–3% lit) became "dark fiber." Supply far outran demand, prices collapsed — WorldCom went bankrupt in 2002 (about $107 billion in assets, the largest in U.S. history at the time), and Global Crossing fell the same year. The infrastructure stayed, later picked up cheaply to power the streaming and cloud eras; but the original backers were wiped out — just like the 19th-century railroad mania.

But not every boom is pure bubble — the internet really did change the world, and "this time is different" is sometimes partly true. The key isn't whether the trend is real, but whether the valuation already discounts years of perfect growth and whether returns on capital can persist. Misuse: not valuing because the trend is grand, or assuming the "picks-and-shovels" (compute/equipment) are necessarily safe — shovel sellers are also governed by cyclical capex and overcapacity.

• Does the link I favor hold pricing power, or is it an undifferentiated capital sink?
• Does the current valuation assume demand and capex "never slow down"?
• If this is a bubble, who benefits from the infrastructure left behind — is it the company I hold?
• Does the "picks-and-shovels" logic ignore its own cyclicality?

Aggregation theory explains where a platform's power comes from: unlike a traditional linear business, a platform need only own the user relationship and the demand side to pressure suppliers in turn. Its economics are near-zero marginal cost, increasing returns to scale, and a powerful "optionality" — a platform that owns the users and infrastructure can extend cheaply into adjacent businesses, which is also why its valuation often exceeds a traditional firm's.

Amazon is the model of optionality: from an online bookstore, to onboarding third-party sellers (third-party goods already exceed half its platform sales), to opening its internal IT capabilities as AWS — AWS revenue was about $90 billion in 2023 and has long contributed the bulk of the company's operating profit. The same users and infrastructure are extended again and again into new markets. Apple's App Store, in turn, shows demand-side leverage: to reach a vast user base, developers must accept roughly a 30% platform cut.

The deeper the moat, the bigger the target. Regulation is a platform's biggest tail risk: in 2024 a U.S. court ruled Google an illegal monopoly in search; the EU's Digital Markets Act (DMA) forces "gatekeepers" to open up interoperability; lawsuits like Epic v. Apple keep challenging the platform cut. On top of that, too high a take rate provokes supplier revolt or "disintermediation." Misuse: treating a platform as an unconstrained, perpetual rent collector, ignoring that regulation can directly rewrite its business model.

• Does this platform's power come from owning the demand side, or is it just one channel among many?
• Is its high take rate / high profit already provoking supplier revolt or regulatory action?
• If antitrust or forced interoperability lands, would it rewrite the core business model?
• Is its "optionality" genuinely extendable, or just a valuation story?