How Google went from Stanford research project to the best business of all time

Executive overview

PageRank solved a problem no existing search engine recognised: relevance through authority, not keyword frequency. While Yahoo built a curated directory and portals maximised time-on-site, Larry Page and Sergey Brin built a system that sent users away as fast as possible — and that became their advantage.

Google's dominance rests on two compounding insights: a ranking algorithm that got better with scale, and an auction-based ad model where revenue per search increases as the advertiser pool grows. Every other structural advantage — cheap infrastructure, aggressive distribution deals, AdSense — flows from those two foundations.

The defining insight: search is a market where increasing scale raises both sides of the equation — lower unit costs and higher revenue per query — making it worth spending nearly any amount to win distribution first.

Origins: PageRank and the crawler

• Larry Page's PhD idea was not search — it was a web annotation system; ranking annotations required a way to establish authority
• The breakthrough: treat hyperlinks as academic citations; anchor text describes the target better than the target describes itself
• PageRank is named for Larry Page, not web pages — though both apply
• Backrub (the prototype) was built at Stanford in 1996–97; Scott Hassan rewrote it in Python; the project borrowed computers from the loading dock of other research groups
• Google could only be built at that moment: the web was just small enough to crawl entirely as a research project; one or two years later it would have been prohibitively expensive
• They tried to sell the technology to Excite, Lycos, Infoseek, and Yahoo for $1 million — all rejected it; Excite's CEO killed a near-done deal because better search would reduce time-on-site and destroy CPM revenue

Infrastructure: commodity hardware as a strategic weapon

• Urs Hölzle (employee #8) and Jeff Dean joined from academia and DEC and rebuilt everything from scratch
• Rather than expensive enterprise servers, Google ran on commodity hardware with 10%+ annual failure rates — software was designed to tolerate failures through replication
• The Google File System broke the index into 64 MB chunks distributed across thousands of machines; a master server tracked chunk locations, enabling parallel queries at scale
• Data centre space was leased by square footage (not power); engineers packed corkboard-mounted motherboards into racks at extreme density
• This architecture eventually made Google the world's largest de facto computer manufacturer as a private company, and produced an ~87% gross margin on search advertising

The business model: from zero to $1.5 billion

• The Series A pitch (Sequoia and Kleiner, $25M at $100M post) proposed enterprise search licenses as the main revenue line — it never worked
• Omid Kordestani joined as CRO; initial ads were hand-sold on Madison Avenue, CPM-based, and inserted by fax
• Jeff Dean validated intent-based advertising using Amazon affiliate links: the test showed high click-through and high downstream conversion, proving the model
• The Yahoo portal deal (June 2000) doubled daily search volume to 14M queries on day one and paid $7.2M in the first full year — bridging Google through the dot-com winter
• GoTo/Overture (founded 1998) pioneered self-serve keyword auctions and cost-per-click — doing $100M in revenue in year one while Google was still fax-selling CPM ads
• AdWords V2 (fully launched 2002) borrowed Overture's auction and CPC model, then added ad rank — a quality score incorporating click-through rate that aligned advertiser, user, and Google incentives simultaneously; it is mathematically the expected-value-maximising formula for Google's own revenue
• The second-price (Vickrey) auction stored latent value by not gouging winners, reducing churn and bid-gaming
• The AOL deal (summer 2002): Google guaranteed AOL $100M in revenue it didn't yet have, betting the company; AOL made $35M in the first half-year and $200M in 2003; revenue went from $86M (2001) to $440M (2002) to $1.5B (2003)

Distribution: why organic growth was never enough

• More advertisers raise auction prices; higher prices fund more aggressive distribution; more users attract more advertisers — the flywheel compounds
• Google Toolbar (December 2000) turned a user worth ~$2/year into one worth $10+/year; Google paid Adobe, RealNetworks, WinZip, and Dell to bundle it with their installers
• Toolbar also included a pop-up blocker — a utility feature that drove installs independent of search loyalty
• Firefox default search deal became Mozilla's primary revenue source for decades
• Google Earth (acquired 2004) was bundled with Toolbar rather than monetised with ads — more than paying for the acquisition
• AdSense (February 2003, built by Jeff Dean in six weeks): extended monetisation beyond search to publisher pages; reached $1M/day in revenue within months; later became the template for the YouTube creator model
• Sundar Pichai joined in 2004 to run the desktop applications team that owned Toolbar

Culture, the IPO, and what came next

• Eric Schmidt hired in March 2001 after a 16-month search (Larry and Sergey's stated preferred candidate was Steve Jobs); operated as a true triumvirate with Larry as president of products and Sergey as president of technology
• "Googliness" — campus culture, 20% time, healthy disregard for the impossible — attracted top talent even during the dot-com boom when better-paying options existed
• The mission "organise the world's information and make it universally accessible and useful" appeared in the first press release in 1999 and proved scalable to every subsequent product
• The 2004 IPO introduced the dual-class share structure — later adopted by Facebook, Shopify, Airbnb, and virtually every major tech IPO since; Google was the first tech company to do this
• The Dutch auction pricing mechanism was not repeated by any subsequent company — the stock priced at $85, popped 18% on day one, and was nearly 5X within 16 months
• Gmail launched April 1, 2004; its prototype (Paul Buchheit's Unix mail search tool) was also the original proof-of-concept that led to AdSense