The dot-com bubble is the part of internet history where the technology stopped being a curiosity and the money stopped being rational. The five-year stretch from roughly 1995 to 2000 produced more lasting infrastructure than any other period in the medium’s history: most of the protocols, most of the architectural patterns, most of the companies that still matter (Amazon, Google, eBay, plus the survivors who got bought into something larger) were either founded or had their formative growth in this window. The bubble also produced spectacular speculative excess, the kind that ends with a CFO in handcuffs and a sock puppet starring in your IPO road show.
This is the story of how the internet got commercialized, how Wall Street got over-excited, and how the crash that followed produced the (more careful, more disciplined, often more profitable) internet industry that emerged after 2002. Foundations through recovery in five phases, plus what it did to the people who lived it and the security shortcuts it baked into systems we still operate.
Phase one: foundations (1969 to 1993)#
ARPANET went live in October 1969 as a four-node packet-switching network funded by the U.S. Department of Defense’s Advanced Research Projects Agency. The original purpose was resource sharing among researchers, not surviving nuclear war. (The nuclear-survivability narrative is a recurring myth that conflates ARPANET with Paul Baran’s earlier 1964 RAND work on generic survivable packet-switching networks. ARPANET’s actual design memos by Bob Taylor, Larry Roberts, and the BBN team are about expensive mainframes being scarce and remote researchers needing to share them.) Ray Tomlinson added networked email in 1971, hacking together SNDMSG and CPYNET while choosing the @ symbol from the available teletype characters. Vint Cerf and Bob Kahn published the original TCP specification in 1974 (RFC 675); the v4 TCP/IP that the modern internet still runs got formalized in 1981 (RFCs 791 and 793).
The 1980s saw the slow expansion. NSFNET launched in 1986 as the National Science Foundation’s network connecting supercomputing centers, eventually absorbing ARPANET’s role and becoming the primary internet backbone before being decommissioned in 1995 when commercial backbones took over. The WELL (Whole Earth ‘Lectronic Link), founded in 1985 by Stewart Brand and Larry Brilliant, was the bulletin board system that produced most of the early online-community vocabulary the industry still uses. By the late 1980s the term “information superhighway” was starting to appear in trade press.
The medium became consumer-facing in 1993. Mosaic, released that year by Marc Andreessen and Eric Bina at the National Center for Supercomputing Applications (NCSA), was not the first web browser (Tim Berners-Lee’s WorldWideWeb came in 1990, and ViolaWWW, Erwise, and Cello were earlier graphical attempts). It was, however, the first graphical browser that ran on the platforms a normal person might use (Windows, Mac, Unix) and the first that was easy enough to install that the popularization actually happened. Andreessen left NCSA in 1994 to found Mosaic Communications, renamed Netscape Communications after the University of Illinois sued over the Mosaic trademark.
That rename is roughly the point where the commercial era starts.
Phase two: commercial awakening (1994 to 1997)#
Netscape Navigator shipped in late 1994, hit 70%+ market share on the web through 1996, and went public on August 9, 1995 in what’s still the IPO most people point to as the start of the bubble. Netscape had never made a profit. The IPO raised about $140 million at the offering price; the stock more than doubled on the opening day, closing at a market cap of roughly $2.9 billion. The financial press realized that a year-old company with no earnings could be worth what an established industrial firm was worth, and the rest of Wall Street drew the obvious conclusion.
Microsoft bundled Internet Explorer with Windows 95 in August 1995 and started giving it away. The first browser war ran from 1995 to about 2002, ending with Netscape’s market share collapsed and Microsoft facing an antitrust lawsuit (Department of Justice 1998, decided 2001) for the bundling tactics. Netscape’s open-source counterpunch (releasing the Navigator source code as Mozilla in 1998) became Firefox after a long, painful decade of rewrite work; that’s a separate story.
The other formative companies of this era:
- Yahoo (Jerry Yang and David Filo, founded 1994 as Jerry and David’s Guide to the World Wide Web). Web directory, then portal, then global media company. IPO April 12, 1996, valued the company at about $848 million.
- Amazon (Jeff Bezos, founded July 5, 1994, launched July 1995). Online bookstore that became an everything store, then a cloud-computing utility, then a logistics company that also runs Hollywood productions. IPO May 15, 1997.
- eBay (Pierre Omidyar, founded September 1995 as AuctionWeb). Person-to-person auction marketplace that demonstrated that the internet could enable economic activity that hadn’t been possible before. IPO September 24, 1998.
- HotWired (Wired magazine’s web operation, launched October 27, 1994). Carried the famous AT&T “You Will” campaign banner ads that started the entire web advertising industry. The often-cited 44% click-through rate (or 78%, accounts vary) for those initial banners is the high-water mark that nobody has approached since.
The infrastructure work was equally important. The World Wide Web Consortium (W3C) was founded by Tim Berners-Lee at MIT in October 1994 to standardize HTML, CSS, and HTTP. HTTP/1.0 (RFC 1945) was published in May 1996; HTTP/1.1 (RFC 2616) in June 1999, adding persistent connections and chunked transfer encoding that made modern web performance possible.
By 1997 the foundations were in place: a web everyone could browse, a small but rapidly growing set of e-commerce companies proving that money could change hands online, an advertising market generating real revenue, and a Wall Street that had decided internet companies were the next thing.
Phase three: irrational exuberance (1998 to 2000)#
The phrase “irrational exuberance” comes from a speech Federal Reserve Chairman Alan Greenspan gave on December 5, 1996, asking whether asset values had become detached from fundamentals. By the time anyone took the question seriously, the answer was unambiguously yes.
The bubble’s mechanics. Venture capital firms (Kleiner Perkins, Sequoia, Benchmark, the entire Sand Hill Road set) funded thousands of internet startups, often based on a business plan and a team rather than a working product. Companies measured themselves by “burn rate” (how fast they spent investor money), “eyeballs” (how many people visited the site), and “first-mover advantage” (a doctrine that the first company in a category would inherit the market). Profitability was treated as a problem to solve later. IPOs absorbed companies at valuations that traditional metrics (price-to-earnings, price-to-book) couldn’t even compute, because the companies had no earnings.
The NASDAQ Composite Index, the heavily tech-weighted exchange, tells the story. The index traded around 1,000 in 1995. It crossed 2,000 in July 1998, 3,000 in November 1999, 4,000 in December 1999, and peaked at 5,132.52 on March 10, 2000.
The cultural and linguistic legacy of this period is significant out of proportion to its duration. Terms that entered the business vocabulary: B2B, B2C, click-through, content portal, eyeballs, get big fast, information superhighway, netizen, surfing the web, sticky, viral, walled garden, web 1.0 (retroactively coined to distinguish it from what came next). Magazine covers featured young CEOs and confident proclamations that the rules of business had changed. The famous Pets.com sock puppet (operated and voiced by Michael Ian Black) starred in a Super Bowl ad in 2000 and became the symbol of the era’s marketing-over-substance ethos.
Famous failures whose names became shorthand:
- Pets.com: $300 million spent on advertising and infrastructure to sell pet supplies online. Founded 1998, IPO February 2000, liquidated November 2000. The sock puppet outlived the company.
- Webvan: Online grocery delivery with $830 million in funding and an aggressive expansion into 26 cities. Founded 1996, peak market cap roughly $7.9 billion in 1999, bankrupt July 2001. (The grocery delivery thesis was eventually proven correct, twenty years later, by Instacart, Amazon Fresh, and the pandemic era; Webvan was right about the demand and wrong about the timing and the unit economics.)
- Boo.com: London-based fashion e-commerce. $135 million in funding, technically impressive site, terrible Flash-heavy user experience nobody could load on 1999 internet connections. Bankrupt May 2000.
- Kozmo.com: One-hour delivery of small items in major cities. Free shipping, no minimum order, no realistic path to profitability. Bankrupt April 2001.
- eToys.com: Toy retailer. IPO May 1999 at $2 billion valuation, bankrupt March 2001.
- Geocities: Web hosting for personal pages, with around 38 million user-created pages at its peak. Acquired by Yahoo for $3.57 billion in stock, announced January 28, 1999. Yahoo shut it down October 26, 2009, taking a chunk of the early web’s cultural history offline (much of it was preserved by the Archive Team’s volunteer rescue effort).
The bubble produced 457 IPOs in 1999 alone, more than the entire decade of the 1980s combined. The average first-day return for those IPOs was over 70%. Several closed up 1000%+ on the opening day.
Phase four: the crash (2000 to 2002)#
The NASDAQ peaked March 10, 2000 at 5,132.52 and started falling almost immediately. The decline was steady through the spring (a 27% drop from peak to mid-April), accelerated through summer, and continued for over two years. The closing low was 1,114.11 on October 9, 2002 (intraday low 1,108.49 on October 10), a 78% loss from peak. Total market-value destruction is commonly cited at $5 trillion, the figure that appears in most academic summaries and post-crash reporting.
The trigger was nothing dramatic. A combination of (1) Microsoft losing the antitrust case in April 2000, (2) several large dot-com companies missing earnings or revenue projections in early 2000, (3) tax-related selling in the first quarter, and (4) a few of the biggest IPO-era survivors (Cisco, Lucent, Nortel) reporting disappointing results. Each of these by itself would have been a market correction; together they triggered a cascade.
What followed was the textbook bursting-bubble pattern. The companies with the worst business models died first (Pets.com, Boo.com, Webvan, etoys, Kozmo all gone by mid-2001). The next tier (Lycos, Excite, AltaVista) merged into larger companies or got acquired at fractions of their peak valuations. The major tech firms that had been buying these dot-coms (Yahoo, AOL, AOL after the catastrophic 2000 merger with Time Warner) wrote down billions in goodwill. AOL-Time Warner’s January 2001 merger was widely considered the symbolic peak of dot-com excess and the symbolic end of it; the combined company eventually unwound the merger in 2009, with a combined market cap roughly $200 billion lower than the deal value.
The non-dot-com side of the story was the telecom collapse. Companies that had invested heavily in fiber optic infrastructure to handle the projected internet growth (Global Crossing, WorldCom, Qwest) found themselves with massive overcapacity and falling demand. WorldCom went bankrupt in July 2002 after an $11 billion accounting fraud was discovered; Global Crossing went bankrupt in January 2002. The dark-fiber overcapacity built during this period would, ironically, become the infrastructure for the next decade of internet growth once the optical equipment to light it caught up with the cable that was already in the ground.
The human cost was real. Roughly half a million tech workers lost jobs between 2001 and 2002. Stock options that had been the bulk of many employees’ compensation became worthless. The “fail fast, fail often” mantra that emerged later was partly the survivors processing how badly the all-in entrepreneurial culture had hurt the people who’d bought into it.
Phase five: recovery and Web 2.0 (2003 to 2005)#
The crash bottomed in October 2002 and the recovery began slowly. The companies that survived had something in common: they were focused on solving actual problems with sustainable unit economics, not on land-grabbing market share with investor money. Amazon spent 2001 through 2003 cutting headcount and tightening operations; Jeff Bezos famously sent a memo demanding the company be self-funded going forward. eBay was already profitable. Google, founded in 1998 but still private, had a working ad-revenue model that was about to become an extraction engine the size of the GDP of mid-sized countries.
The venture capital community came back with more discipline. The “burn through investor cash for market share” pitch no longer worked; the new requirement was a credible path to profitability and a real customer-acquisition cost story. Sand Hill Road still funded ambitious startups, but the term sheets and the board oversight tightened significantly.
The cultural mark of the recovery was Web 2.0, a term Tim O’Reilly coined for a 2004 conference and that came to mean a specific shift in what the web was for. Web 1.0 was publishing: companies had websites, users visited them, content flowed one way. Web 2.0 was participation: Flickr (2004, photo sharing), del.icio.us (2003, social bookmarks), WordPress (2003, blogging), MySpace (2003, social networking), Wikipedia (founded 2001 but mass-adopted in this window), YouTube (founded February 2005), and Facebook (founded February 2004, opened to the general public in 2006) all leaned on user-generated content rather than editorial publishing. The underlying technical shift was AJAX (the term coined by Jesse James Garrett in February 2005), which let web pages update in place without full page reloads and made dynamic-feeling web apps practical.
The Web 2.0 companies built sustainable businesses where the dot-com companies hadn’t. Google’s IPO in August 2004 was the first major internet IPO since the crash; it raised $1.67 billion and was deliberately structured as a Dutch auction to avoid the bubble-era investment-bank price-pump pattern. By the end of 2005, the broad internet industry was profitable again, the survivor companies were healthy, and the lessons of the bubble had been thoroughly absorbed into how the next wave of companies got built. The infrastructure overcapacity from the telecom bubble (fiber in the ground, data center space) was now being absorbed at sustainable prices, fueling a decade of growth that would produce the mobile, social, and cloud computing eras.
The spectacular implosion had, paradoxically, cleared the ground for the durable internet economy that grew on top of the wreckage.
The human cost#
The financial story gets told most often. The human side affected hundreds of thousands of lives and shaped a generation’s attitudes toward technology careers in ways the index charts don’t capture.
Mass layoffs and career disruption#
The crash triggered one of the largest waves of layoffs in technology history. Companies that had hired aggressively during the boom shed employees at unprecedented rates: net tech employment losses ran into the hundreds of thousands across 2001 and 2002, concentrated in the Bay Area, Seattle, Austin, Boston, and New York. Workers who had been recruited with stock-option packages and promises of wealth found themselves unemployed with options that were worth less than the paper they were printed on.
The psychological toll was real. Many employees had invested life savings in their company’s stock during the boom because the company-purchase plans seemed like sure things; when valuations collapsed, those investments collapsed alongside their jobs. The pattern of executives cashing out options at the peak while rank-and-file employees rode the stock down to zero created lasting cynicism about Silicon Valley culture that the industry has never fully shaken off.
Entrepreneurial burnout#
The bubble destroyed not just companies but entrepreneurial lives. Founders who had mortgaged homes, cashed out retirement accounts, and borrowed from family to fund their startups found themselves personally bankrupt. The stigma of failure in the venture capital community of the early 2000s made it difficult for these entrepreneurs to secure funding for follow-on ventures; the “fail fast, fail often” mantra that emerged later was deliberate VC marketing aimed at undoing exactly that stigma.
The shift in entrepreneurial attitudes was substantial. The next wave of founders (the Web 2.0 generation) was visibly more cautious, more focused on early customer validation, less inclined to make billion-dollar bets on unproven thesis. Many of them had watched the bubble crash from inside a failed startup and weren’t interested in repeating the experience.
Geographic impact#
The bubble’s collapse hit technology hubs disproportionately. San Francisco saw a meaningful population decline as unemployed tech workers moved out. Seattle, Austin, and Boston had similar though smaller shifts. Commercial real estate in these markets crashed: office buildings that had been built or leased for dot-com companies sat empty for years. Some entire neighborhoods that had been built up to serve the influx (South of Market in San Francisco, the South Lake Union area in Seattle) had ghost-town feels through 2003 before the recovery filled them back up.
The social fabric of these communities was disrupted. Tech workers who had formed tight communities around their companies found themselves isolated when the companies dissolved. The loss of professional networks made it harder to find new opportunities; the loss of social networks made the layoff experience harder to recover from emotionally.
Long-term career impacts#
The bubble produced a lost generation of technology professionals. Some left the industry entirely for more stable fields (finance, government, healthcare). Those who stayed carried the scars: more risk-averse career choices, more skepticism of equity-heavy compensation packages, a general unwillingness to bet everything on a single employer’s stock.
The generational divide this produced is visible decades later. People who lived through the dot-com bubble at the worker level are systematically more cautious about speculative technology investments than those who came of age after; the difference shows up in how they respond to AI investment rhetoric, cryptocurrency promotion, and similar speculative-bubble patterns.
Lessons for modern tech workers#
A few things the dot-com generation has been passing along to the cohorts that followed:
Diversify away from employer equity even when the employer’s stock is appreciating. The bubble taught the industry exactly how fast that appreciation can reverse, and the people who put their life savings into company stock at the peak still talk about it.
Build transferable skills. Specialization in a single company’s stack makes you vulnerable when that company evaporates. Fundamentals carry across employers; framework-of-the-month specialization doesn’t.
The 24/7 startup culture of the bubble era burned a lot of people out and produced lasting health consequences. Some of the survivors still won’t talk about the hours they kept in 1999. The work isn’t worth what it costs you.
Build a network outside your current employer. Professional connections in other companies are what make career transitions survivable when the company you work for stops existing.
Plan for contingencies. Economic cycles hit the technology industry too. Assume a downturn at some point in your career and have a plan that doesn’t depend on your current job continuing to exist.
These came at a cost. The dot-com generation paid for them with their savings, their health, and the houses they had to sell.
The cultural impact#
The bubble reshaped how the broader culture thinks about technology, work, and entrepreneurship. The aftershocks are still propagating.
The language of the internet#
The bubble normalized a vocabulary that’s now embedded in popular speech. “Surfing the web,” “netizen,” “information superhighway,” “dot-com” itself all entered the mainstream during this window. The marketing language (“get big fast,” “first-mover advantage,” “disruptive innovation,” “eyeballs”) moved from VC pitch decks into business journalism and then into general usage. Some of these terms outlived the bubble; others (the “information superhighway” framing) sound dated now in ways that places them firmly in 1996.
Silicon Valley mythology#
The bubble cemented Silicon Valley as the place where young people could go to become unreasonably wealthy in a hurry. Stories of garage startups becoming global giants attracted talent from around the world, often at the expense of the local industries those people came from. The brain drain pattern (engineers from Detroit’s automotive industry, from Texas’s oil industry, from New York’s finance industry, from India and China and Israel) all leaning toward the Bay Area is a direct consequence of the mythology the bubble built.
That mythology, while often exaggerated, was real enough to drive real migration. The Bay Area population shifts of the late 1990s and the post-2003 recovery period are the demographic shadow of the story the bubble told.
Work culture transformation#
The bubble popularized “internet time”: decisions made fast, work hours extended, weekends and evenings treated as productive territory. The 24/7 startup culture that became normalized during the bubble influenced work expectations across the broader economy. While this produced genuine innovation and rapid product iteration, it also produced burnout at scale and created the work-life balance issues that are still part of the technology industry’s reputation.
Media and popular culture#
The bubble became a recurring theme in film, books, and television. The Pirates of Silicon Valley (1999) dramatized the Microsoft-Apple personal-computer rivalry that predated the bubble but informed its mythology. Startup.com (2001) documented the rise and collapse of govWorks.com in real time, ending up as the canonical dot-com-crash documentary. The Social Network (2010) revisited the post-bubble Web 2.0 mythology of Facebook’s founding. The bubble’s crash became a cautionary tale that subsequent technology booms (mobile, social, crypto, AI) have all been compared to.
Generational attitudes#
The bubble produced a generational divide in how technology and entrepreneurship get viewed. People who experienced the crash firsthand became cautious about speculative investments and hype-driven valuations; younger generations born after 1995 sometimes view the dot-com era as historical, the way the bubble generation viewed the 1929 stock market crash. The result is that crypto boom-and-bust cycles, NFT mania, and AI investment patterns play out differently for the cohort that remembers 2000 firsthand versus the cohort that doesn’t.
Global technology culture#
The bubble’s American model (venture capital funding, founder-led startups, IPO-as-exit, equity-heavy compensation) became a global template. European, Asian, and Israeli tech ecosystems all incorporated elements of the Silicon Valley approach. The dot-com era’s lessons about scalability, user acquisition, and monetization became universal vocabulary in technology entrepreneurship worldwide.
The cultural legacy is still in the air in 2026. The vocabulary the dot-com era invented is the vocabulary every subsequent technology bubble has been described with, and the comparison points subsequent booms get measured against (good and bad) all trace back to this window.
What survived#
The bubble produced more lasting infrastructure than any other period in internet history. The survivors:
- Companies. Amazon (down 95% from peak, survived), eBay, Yahoo (acquired by Verizon 2017, then split off), Priceline (now Booking Holdings), Akamai, Cisco (the original infrastructure beneficiary). Plus the post-bubble winners (Google, founded 1998, IPO 2004; Facebook, founded 2004; Salesforce, founded 1999, IPO 2004) that built on the rubble.
- Protocols and standards. HTTP/1.1, SSL 3.0 and the early TLS revisions, DNS as deployed at the scale it now runs at, CDN architecture (Akamai launched commercially in 1999), the entire IPv4 internet routing infrastructure as it actually exists today.
- Architectural patterns. Server-side templating (Active Server Pages 1996, PHP 3 1998, JSP 1999). Database-backed dynamic web applications (the LAMP stack: Linux + Apache + MySQL + PHP/Perl/Python). RESTful APIs avant-la-lettre. The three-tier web architecture (browser + web server + database) that almost every web application still uses.
- Search engine technology. Google’s PageRank, the inverted index, distributed crawling, the entire concept of web-scale relevance ranking. Yahoo, AltaVista, and Excite did the early version; Google did the version that won.
- E-commerce infrastructure. Payment processing, shopping carts, fulfillment integrations, recommendation engines. The technical work that Amazon and eBay pioneered is what every modern e-commerce platform is still built on.
- CDNs. Akamai started in 1998 (MIT computer science research, commercialized 1999). The basic CDN model (edge servers caching content closer to users) is what made the modern web possible at scale.
- The basic operational discipline of running an internet service. Capacity planning, redundancy, monitoring, incident response. The first generation of site reliability engineering was named after the bubble had ended (Google’s SRE term emerged around 2003), but the practices were forged during the bubble’s late-1999 “Y2K plus the holiday traffic spike plus the 60 Minutes piece” reality.
The security legacy#
The bubble baked a set of security shortcuts into the internet’s infrastructure that operators still deal with daily.
E-commerce security in the late 1990s was an afterthought. Companies shipping web applications had business pressure to launch first and fix later, no standardized testing methodology (OWASP wasn’t founded until December 2001), and no shared concept of what a “secure web application” should look like. The result was the original vulnerability classes that still plague web applications: SQL injection (the term itself dates from this era), cross-site scripting (XSS, named by Microsoft researchers in 2000), CSRF, broken authentication, insecure session management. The CD Universe breach in December 1999 (300,000 credit card numbers stolen by “Maxus,” a Russian-speaking hacker who attempted to extort the company for $100,000 before dumping the data publicly) is the most-cited canonical example.
SSL implementation in the bubble era was uniformly bad. SSL 2.0 (1995) had known cryptographic weaknesses; SSL 3.0 (1996) was better but still ended up vulnerable to POODLE (2014). Many sites used 512-bit RSA keys that were technically factorable by the late 1990s (and definitely factorable now). Certificate authorities were less rigorous about validating who they issued certificates to; the entire PKI trust model was assembled on the assumption that the CAs were trustworthy, which turned out to be optimistic. Certificate pinning didn’t exist. Encrypted internet traffic was a minority of internet traffic until well after 2010.
DNS and routing were similarly under-secured. DNSSEC’s original RFCs (2535 in 1999) were a false start; the modern DNSSEC standards (RFCs 4033-4035) didn’t land until 2005, and deployment crawled for another decade. BGP route hijacking was theoretically possible in 1999 and remains practically possible in 2026; the RPKI work that addresses it is still rolling out. The basic architecture choice that authentication is layered on top of the network rather than into it dates from this era.
Default credentials, hardcoded passwords, and unauthenticated administrative interfaces were endemic. The convention that ships every device with admin/admin traces to the late-1990s consumer router market; the Mirai botnet in 2016 weaponized exactly this pattern at scale. The security shortcuts that the bubble’s speed-to-market culture normalized are why default-credential exploitation is still the dominant initial-access vector against IoT in 2026.
The first defensive responses emerged in the late 1990s and early 2000s. Network-based intrusion detection systems matured (Snort 1998). Web application firewalls appeared (the first commercial WAFs from Perfecto/Sanctum and KaVaDo around 1999-2000). The Gramm-Leach-Bliley Act (1999) imposed security requirements on US financial institutions. HIPAA (1996) did the same for healthcare. The regulatory baseline that protects consumers from the worst of dot-com-era security practices is the residue of the lawsuits and breaches that the bubble produced.
What the bubble teaches in 2026#
The bubble’s most useful lesson isn’t financial; it’s about how rapidly-deployed insecure technology becomes foundational for decades. The TCP/IP stacks shipped in late-1990s embedded devices are still in field-deployed industrial equipment that no firmware updates will ever reach. The early-generation e-commerce code (PHP 3 era, ASP Classic, Cold Fusion) is still running on plenty of small business sites that work fine and won’t be replaced. The PKI we have is the one the bubble era hashed out and we’ve been incrementally improving ever since.
The economic parallels with later bubbles (the housing-derivatives bubble of 2007-2008, the cryptocurrency bubbles of 2017 and 2021, the AI investment bubble that started in late 2022 and is ongoing as of 2026) are striking but not perfect. Each has its own version of the “burn rate as a metric of success” pattern; each has its own version of the “the rules have changed” rationalization; each has produced its own batch of survivors building durable companies and casualties whose names became cautionary tales. The specific technical issue that the dot-com era introduced (insecure web applications at internet scale) is the one that operators are still cleaning up, and will be for the foreseeable future.
The internet is bigger, more important, and more profitable than the bubble-era cheerleaders predicted. It’s also less trustworthy, more centralized, and more surveilled than the bubble-era libertarians hoped. Both of those things were true in 2002 and remain true in 2026.
Trivia#
The kind of facts that get a party conversation going, with the actual numbers this time:
- The term “dot-com” comes from the
.comtop-level domain that commercial websites used. The TLD itself dates from RFC 920 in October 1984, predating most of the bubble by a decade. - The most expensive domain name ever sold (as of 2026) was Voice.com, which Block.one bought for $30 million in 2019 to use for an EOS-blockchain social network. The platform launched in 2020 and was largely abandoned by 2023.
- Some of the most famous bubble-era failures include Webvan, Kozmo.com, eToys.com, Pets.com, Boo.com, and govWorks.com (the subject of the documentary Startup.com). Each became shorthand for a specific failure mode of the era.
- During the height of the bubble, companies with no earnings and no clear revenue path were valued at billions of dollars. Some IPOs achieved $1B+ market caps on the first day of trading despite never having posted a profitable quarter.
- The Pets.com sock puppet was operated and voiced by comedian Michael Ian Black. The mascot appeared in a Super Bowl ad in 2000 and outlived the company it was made to promote.
- The NASDAQ peaked at 5,132.52 on March 10, 2000 and bottomed at 1,114.11 on October 9, 2002 (intraday low 1,108.49 on October 10). The index didn’t reclaim its bubble-era peak until April 2015, fifteen years after the crash.
- Bubble-era startups used “burn rate” (how fast they were spending investor money) as a metric of success, in the inverted sense: a high burn rate signaled aggressive market capture. The metric came back into normal usage post-crash as a warning rather than a brag.
- The dot-com crash dragged down the telecom industry alongside it. WorldCom went bankrupt in July 2002 after an $11 billion accounting fraud was discovered; Global Crossing went bankrupt in January 2002. The dark-fiber overcapacity these companies had built became, ironically, the infrastructure for the next decade of internet growth.
- Some of the most successful internet companies (Google, Amazon, eBay, Priceline) weathered the bubble and emerged as dominant players in the decade that followed. Amazon’s stock dropped 95% from its 1999 peak before beginning the multi-decade recovery that produced the largest market capitalization in U.S. history.
- The dot-com era saw the emergence of online advertising as a real business model. DoubleClick (founded 1996, acquired by Google in 2007 for $3.1 billion) pioneered programmatic advertising and the ad-targeting infrastructure that powers most of the modern web.
- The first web banner ad ran on HotWired on October 27, 1994, an AT&T “You Will” campaign placement designed by Joe McCambley at Modem Media. The canonical click-through rate figure is 44%, wildly higher than what banner ads achieve today (typical CTRs in 2026 are under 0.5%).
- The bubble was where “unicorn” status (a private company valued over $1 billion) first became achievable without a profit or a viable business model. The term itself came later (Aileen Lee coined it in a 2013 TechCrunch post) but the phenomenon dates from the dot-com era.
- Total market-value destruction from the NASDAQ peak to the October 2002 bottom is commonly cited at $5 trillion across the broader tech sector. Individual companies lost 90%+ of their peak valuations; many lost 100%.
- “Vaporware” marketing (promising products that didn’t exist or weren’t ready) was rampant during the bubble. Companies announced products months or years in advance to capture mindshare and stock-price appreciation, with no working code or shipping plan.
- The term “brick and mortar” became popular during the bubble to distinguish traditional physical-world businesses from internet-based ones. The implication was usually that brick-and-mortar businesses were antiquated; many of them turned out to be more durable than the dot-com competitors that dismissed them.
- The Y2K bug fears of 1999-2000 inadvertently fueled the bubble. Businesses invested heavily in new computer systems and internet infrastructure to address potential millennium-bug issues, much of that spending flowing to internet-adjacent vendors and integrators.
- During the bubble, internet bandwidth was severely limited by dial-up connections (56K modems were typical; broadband was a minority of households until 2003). This led to the development of compressed image formats (PNG 1996), JPEG optimization techniques, and the “web-safe” 216-color palette that optimized for slow connections.
- The bubble produced 457 IPOs in 1999, more than the entire decade of the 1980s combined. The average first-day return for those IPOs was over 70%; some closed up 1000%+ on opening day.
- Bubble-era vocabulary that entered mainstream usage: “surfing the web,” “netizen,” “information superhighway,” “dot-com,” “B2B,” “B2C,” “click-through,” “stickiness,” “viral,” “first-mover advantage,” “burn rate.” Some have aged better than others.
- High-profile breaches during the era, like the CD Universe credit card theft in December 1999 (300,000 cards stolen by “Maxus,” a Russian-speaking hacker, $100,000 extortion attempt followed by public dumping when refused), led to the creation of the first commercial web application firewalls (Perfecto/Sanctum and KaVaDo, both 1999-2000).
- Many bubble-era companies offered services for free, planning to monetize through advertising or premium features later. This “free now, monetize later” model became the foundation for modern social media and SaaS companies; whether it works financially depends entirely on the specific economics, which the bubble’s first generation often got wrong.
- The bubble created a Silicon Valley party culture that became its own cultural phenomenon. Lavish launch parties, $50,000 product-launch events, and the “internet time” mentality that celebrated excess produced legendary excess. Many founders of that era talk about it now the way recovering alcoholics talk about their drinking years.
- The era saw intense patent litigation as companies raced to patent basic internet concepts: Amazon’s 1-Click purchase patent (U.S. Patent 5,960,411, granted September 1999), British Telecom’s hyperlink patent claim, and various overly-broad business-method patents that produced expensive litigation for years.
- Geocities, the first big consumer web-hosting platform, hosted around 38 million user-created pages at its peak. Yahoo announced the $3.57 billion stock acquisition on January 28, 1999 (closed May 28, 1999) and shut Geocities down October 26, 2009. The Archive Team rescued much of the content the day before the shutdown.
- The bubble produced comically named companies. Pets.com, Boo.com, Kozmo.com, eToys.com, Drkoop.com, Beenz.com, Flooz.com, govWorks.com. Names that favored .com additions or portmanteaus often at the expense of brand coherence. The naming pattern’s still recognizable two decades later.
- The famous Greenspan “irrational exuberance” speech was given on December 5, 1996. The bubble peaked more than three years later. Greenspan’s question was correct; his timing was atrocious.
Closing#
The dot-com era didn’t invent the internet, but it commercialized it, productized it, and (briefly) financialized it well past anything fundamentals could support. The crash was painful and overdue. The companies that survived built the foundations of the modern internet economy; the protocols and architectures that crystallized during the era still define what running a web application looks like; the security shortcuts that the era normalized are still the operator’s problem.
The bubble’s most genuinely strange feature, in retrospect, is how short it was. Five years from Netscape’s IPO to the NASDAQ peak, two and a half years from the peak to the bottom, and the entire arc fit inside a single generation’s working memory. People who lived through it on Sand Hill Road and in Silicon Alley still remember the launch parties and the layoffs in the same week. Most of them are still in the industry; some of them are still writing the cautionary tweets every time a new technology starts attracting similar investment patterns. The next bubble probably won’t be the dot-com bubble in costume, but it will rhyme.
