The Class Of: The Google Class of 2004

The Million Dollar Question: When Google went public in August 2004, how long did its rank-and-file employees have to wait before they could sell any stock at all?
A) 15 days B) 90 days C) 180 days D) A full year

Read on for the answer.

In late November 2004, three months after the most closely watched public offering of the decade, Google wrote to more than a thousand of its own employees with an unusual proposition. It would like to buy their stock back.

The offer was not a favour and it was not a raid. It was a legal remedy, and Google had no choice but to make it. The rescission offer, circulated that month, offered to repurchase 23,435,945 shares of Google common stock from 1,320 current and former employees and consultants, plus unexercised options over a further 5,215,379 shares from 282 more. For the shares, the price would be exactly what each person had paid — somewhere between 30 cents and $80.00 a share, plus statutory interest. For the unexercised options, it would be twenty per cent of the exercise price.

On the last trading day before the offer was dated, Google closed at $165.10.

Nobody accepted. Not one holder in the entire group. The offer expired on 30 December 2004 with, in the company’s own later phrasing in its annual report, none of the holders having accepted it.

That is a strange way for a wealth event to begin, and it is the right way into this one. The Google class of 2004 is the largest cohort in this series and, in aggregate, by far the richest. It is also the one where the arithmetic of who ended up wealthy has the least to do with hiring dates and the most to do with a question every member faced repeatedly for a decade afterwards: when do you let go? The rescission offer was the first time they were asked, and the answer was unanimous. It would not stay that way.

The denominator, for once, is printed

Two earlier pieces in this series turned on a count. The PayPal Mafia rests on a legend built from thirteen photographed men out of 776 who actually worked there. The WhatsApp 55 has the opposite problem: the famous number is real but comes from journalism rather than any filing, because the company was private and never disclosed a headcount.

Google is the easy case. It was going public, so it had to say.

The final IPO prospectus states that at 30 June 2004 the company had 2,292 employees — 705 in research and development, 1,141 in sales and marketing, 446 in general and administrative roles. The prospectus is unusually pointed about it, because Google had to reprint and correct a Playboy interview with the founders that the SEC had queried: where the article said the company had around 1,000 people, the correction notes drily that “currently, we have approximately 2,292 employees.” By 31 December 2004 the number was 3,021.

So the cohort is knowable, and it is large: roughly 2,300 people, against 776 at PayPal and 55 at WhatsApp. Alphabet’s headcount at 30 June 2026 was 198,933, meaning the entire class of 2004 would today be about one per cent of the company.

The offering itself was small relative to the company. Google sold 14,142,135 Class A shares — the digits of the square root of two, one of several arithmetic jokes in the paperwork — and selling stockholders sold another 5,462,917, for a total of 19,605,052 shares at $85.00. After the offering, 271,219,643 shares were outstanding, which put the whole company at roughly $23.1 billion. The stock closed its first day at $100.34, up about 18 per cent.

What the class was holding

Here is the part almost never quantified, and most of it is sitting on page 2 of the prospectus.

At 30 June 2004 Google had outstanding options over 6,276,573 shares of Class A stock at a weighted average exercise price of $9.42, and 10,456,084 shares of Class B stock at a weighted average exercise price of $2.68. That is 16,732,657 options in total — a little over 6 per cent of the company — against about $87 million of aggregate exercise cost. At the $85 offering price they were worth roughly $1.42 billion gross, or $1.33 billion net of what it would cost to exercise them.

But options were only half of it, and this is where most write-ups of the Google IPO stop too early. Employees had been exercising for years, so a large block of stock was already owned outright. The rescission offer alone accounts for 23,435,945 such shares — worth almost $2.0 billion at the offering price, more than the entire option pool — and that block is a floor rather than a total, because the rescission offer only reached people who had received grants between September 2001 and July 2004 and who lived, or had lived, in nineteen named states.

Put the two together and the class of 2004 was sitting on at least $3.3 billion of paper on the day the stock started trading, across about 2,292 people. That is an average of roughly $1.45 million a head.

The average is real arithmetic and also nearly useless, for the reason averages always are in these pieces: the distribution behind it was a pyramid. A 1999 hire held a grant an order of magnitude larger than a 2003 hire’s, and the founders’ stock is not in these numbers at all. The prospectus flags that many senior grants were already fully vested at the offering — a risk factor, in Google’s telling, because “employees may be more likely to leave us after their initial option grant fully vests.”

The most-repeated number about Google wealth comes from outside the filings, and it needs a date attached. In November 2007, The New York Times reported, in a story built around Bonnie Brown — hired in 1999 as Google’s in-house masseuse at $450 a week plus options, retired five years later a multi-millionaire — that roughly 1,000 Google employees held stock and options worth more than $5 million each. That is a well-sourced press estimate, not a disclosure, and it describes Google’s workforce in late 2007, by then close to 16,000 people with the stock near $660. It is not a statement about the 2,292.

The rank and file got out first

Now the mechanism that makes this cohort different, and the answer to the question at the top.

The standard shape of an IPO is that insiders sign a 180-day lock-up agreement with the underwriters, and for six months nobody sells. Google did not do that. The prospectus is explicit: “None of our officers, directors, employees or stockholders have entered into contractual lock-up agreements with the underwriters in connection with this offering.” Google itself agreed not to issue stock for 180 days, but its people signed nothing with Morgan Stanley or Credit Suisse First Boston. Their selling restrictions were contracts with Google, which meant Google decided when they lapsed, and reserved the right in its “sole discretion” to let anyone out early.

And there was not one schedule. There were three, in the prospectus’s “Selling Restriction Agreements” section, and they are ordered in a way almost nobody would guess:

• Employees and other holders: 5 per cent of their holdings released 15 days after the prospectus, another 10 per cent at 90 days, the remainder at 180 days.
• Executive officers: nothing until 90 days, when 5 per cent was released — 10 per cent for those with no vested holdings — and the remainder at 180 days.
• Parties to the Investor Rights Agreement, meaning the pre-IPO venture investors: one third at 90 days, one third at 120, one third at 150.

The answer to the Million Dollar Question is A. A Google employee could sell a slice of their vested stock fifteen days after the prospectus — a full seventy-five days before any executive officer could sell a share, and before the venture investors could touch anything. That inversion is the unusual part. In most offerings the hierarchy of the cap table is also the hierarchy of the exit; at Google the junior staff were let out first and the executives were held longest.

Google’s own summary of the arrangement is the flattest sentence in the document and the most consequential: the agreements “will allow significantly more shares to become freely tradeable soon after completion of the offering than is typical of initial public offerings.”

The price behaved accordingly. Between 19 August and 22 November 2004, Google’s closing price ranged from $100.01 to $196.03 — the stock nearly doubled through exactly the window in which its own staff were being progressively released to sell into it. And that is the thing that separates this cohort from the others. Every other group in this series had the sell-or-hold decision deferred for them by a contract. Google’s did not. From two weeks after the offering, an employee with vested stock was choosing, personally and repeatedly, between certainty and compounding, in a rising market, with no one to blame either way.

The stock they were never supposed to have

Which returns us to the rescission offer, and to the most under-told fact about this cohort: a large part of what it had been paid in was issued in breach of the securities laws.

On 13 January 2005, the SEC issued a cease-and-desist order against Google and its general counsel, David Drummond. The findings are unusually readable. Rule 701 lets a private company hand out stock options to employees without registering them — but a company issuing more than $5 million of options in any twelve months must give recipients real financial statements. Google did not want to. It regarded publishing its financials as “strategically disadvantageous,” and in the Commission’s account Drummond concluded that other exemptions covered the grants, and that if that analysis proved wrong, “Google could make an offer of rescission to the option holders.”

The analysis proved wrong. For the twelve months to 31 December 2003 Google issued approximately $49 million of stock options, and a further $33 million in the first four months of 2004 — against a $5 million disclosure threshold, with none of the required disclosures. Over 2002 to 2004 the total exceeded $80 million.

The order is careful about what Drummond knew and when, and it is worth being careful too. At a January 2003 board meeting, the Commission found, he did not report that the grants “might cause” Google to breach the threshold — and it expressly finds that the breach came about “contrary to Drummond’s expectations.” By the June 2003 meeting the finding is harder: he advised the board to adopt two further option plans without telling it that the grants would exceed the threshold, or that the fallback exemptions might not apply. Google and Drummond settled without admitting or denying the findings. The Commission credited their cooperation and imposed no penalty beyond an order to stop.

So the rescission offer was the remedy Drummond had planned for in advance. And by the time it arrived it was worthless, because rescission returns the purchase price, and the purchase price was 30 cents to $80 against a market price of $165. Google’s obligation was to make the offer, not to make it attractive; the November prospectus put the maximum possible cost at $28.3 million if every single holder accepted, and the SEC’s order notes that a rescission offer does not cure a Section 5 violation anyway. Zero people took it.

It is the cleanest illustration in the series of a rule the site keeps running into: employee equity is a security, and securities law does not care that the recipient is a colleague. The class of 2004 had been paid in paper for years, in instruments whose paperwork the company’s own general counsel had misjudged, while being told nothing about the company’s finances — and it made them rich anyway.

The price of selling

Now put a number on letting go.

Google’s stock has split twice. In April 2014 the company distributed one non-voting Class C share for each Class A and Class B share, which functioned as a two-for-one split; in July 2022 it did a straight twenty-for-one. A single share bought at $85 in the offering is therefore forty shares today — twenty Class A and twenty Class C. At the 4 September 2026 close of $338.46 for Class A and about $335.72 for Class C, that original $85 share is worth roughly $13,480: a return of about 158 times over twenty-two years, or close to 26 per cent a year compounded, before counting the dividend Alphabet began paying in 2024. Alphabet’s market capitalisation is around $4.1 trillion, against the $23.1 billion the offering implied.

Apply that multiple to what the cohort held and the counterfactual becomes uncomfortable. The option pool alone corresponds to about 669 million shares in today’s terms — roughly $226 billion. Add the already-exercised shares in the rescission block and the figure passes $500 billion.

That is not what the cohort has. It is what the cohort would have had if every option had been exercised and every share held for twenty-two years, and it is a deliberately unreal number: the options carried ten-year terms and had to be exercised by around 2013 or 2014 at the latest, exercising triggered a tax bill in the year of exercise, and no sane person concentrates their entire net worth in one employer’s stock for two decades. The gap between half a trillion dollars and whatever the real figure is represents the aggregate price this cohort paid for diversification, liquidity, houses, second acts, and sleep. Most of them would tell you it was worth it, and for most of them it was.

But the shape of the outcome is worth stating plainly. This cohort’s wealth was determined far more by holding than by hiring. The famous variable in these stories is the employee number. Here the variables that mattered were the exercise decision, the difference between an incentive and a non-qualified option at tax time, and the nerve to sit on an undiversified position through 2008, when Google fell by more than half. Very few people do that with money they have already been told is theirs. It is entirely possible — as a hypothetical, not a documented case — for a modest 2003 grant held to the present to have beaten a grant ten times its size sold in 2005, and that possibility is the whole story of this cohort in one sentence.

One footnote, because it is usually cited the other way. In December 2006 Google announced a Transferable Stock Option programme, letting non-executive employees sell vested options outright to banks through an online auction run by Morgan Stanley — the first ongoing, competitively bid scheme of its kind. It is often described as Google giving its people a clever new way out. It did not apply to this cohort: the press release is explicit that only options issued after the IPO were eligible, which excluded every pre-IPO grant the class of 2004 was holding.

Where the class went

Follow the cohort forward and it splits three ways, and only one of the three is famous.

The ones who stayed. The least written about. Susan Wojcicki, who rented her Menlo Park garage to Page and Brin in 1998 and became employee number sixteen, ran YouTube until 2023 and died in 2024. Jeff Dean, Urs Hölzle, Salar Kamangar and Sundar Pichai — who joined in April 2004, four months before the offering, and now runs the company — stayed for two decades and rode a stock compounding at around 26 per cent a year. Staying was, in pure return terms, the winning strategy, and it is invisible in every retelling because nothing happened to these people that reads as a story.

The ones who left and built. The visible group, concentrated between 2005 and 2008, once the initial grants had fully vested — exactly the risk the prospectus had warned about. Aydin Senkut, hired in 1999 as Google’s first product manager, left in 2005 and founded Felicis Ventures the following year, one of the first of the “super angel” funds. Paul Buchheit, the author of Gmail, left in 2006; Bret Taylor and Jim Norris followed in mid-2007, and in October that year the three co-founded FriendFeed with Sanjeev Singh, which Facebook bought in 2009. Taylor went on to be Facebook’s CTO, co-CEO of Salesforce, chairman of OpenAI’s board and founder of Sierra; Buchheit became a partner at Y Combinator. Chris Sacca, who joined in late 2003, left in December 2007 and started Lowercase Capital, one of the best-returning seed funds ever on the strength of Twitter, Uber and Instagram. Sheryl Sandberg, hired in 2001 to build the ad business, left for Facebook in 2008.

Our reading of that list is that it describes a conversion rather than a diaspora. The Google cohort produced comparatively few large operating companies of its own and a striking number of investors — because its members left with seven- and eight-figure sums at mid-career, which is roughly the amount of money that turns a former employee into a fund. PayPal’s alumni, richer in ambition than in dollars, had to build. Google’s could write cheques instead, and many did.

The ones who simply left. The largest untold group. Two thousand people times a decade of ordinary attrition means most of this cohort was gone before the stock did most of its work. They sold into the early tranches, took a sum that was life-changing without being life-altering, bought houses in a market about to peak, and are on nobody’s list. Bonnie Brown, who did far better than that, is the only one most people can name — remembered precisely because a masseuse retiring a multi-millionaire was the exception worth a headline.

One correction belongs here, because it is the most common error made about this cohort. Kevin Systrom and Ben Silbermann were not in it. Systrom joined Google in 2006 as an associate product marketing manager, moved to corporate development and left in January 2009; Silbermann worked in the online advertising group from December 2006 to November 2008. Both arrived more than two years after the offering, with post-IPO grants at post-IPO prices, and neither took a windfall out of Google. Instagram and Pinterest are Google alumni companies. They are not class-of-2004 outcomes.

What people get wrong

That the IPO made the cohort’s money. It made at least $3.3 billion of paper across roughly 2,300 people, which is real but is not the story. The stock has multiplied about 158 times since. The offering was the starting gun on a twenty-two-year holding decision, and that decision, not the IPO, sorted the outcomes.

That employees were locked up for six months. They were not locked up by the underwriters at all — the prospectus says so twice. Their restrictions were contracts with Google, releasable by Google at its own discretion, and for rank-and-file staff they began lapsing fifteen days after the prospectus, seventy-five days before any executive officer could sell. Google said in the document itself that the arrangement would free far more stock, far sooner, than a normal IPO. It is the most distinctive structural fact about this cohort and it is almost never mentioned.

That the option pool is the measure of what employees held. It is barely half. The 23.4 million shares in the rescission block were already owned outright and were worth more at $85 than every outstanding option combined. Any per-employee figure built on options alone understates the position by more than a factor of two.

That “1,000 Google millionaires” describes the class of 2004. It is a 2007 press estimate of how many people then at Google — a workforce of roughly 16,000, with the stock near $660 — held more than $5 million. No filing contains such a count. The figure is plausible and well sourced; it is not about the 2,292, and a good share of it belongs to people hired after the offering.

That the rescission offer was Google being generous. It was Google being caught. The SEC found the company issued more than $80 million of options between 2002 and 2004 without the disclosures Rule 701 required, and that its general counsel had considered the rescission offer in advance as the fallback if his exemption analysis failed. The offer was mandatory, priced at what employees had paid, and made when the stock was double the top of that range.

That the transferable-option programme was this cohort’s exit. Only options granted after the IPO were eligible. Every pre-IPO grant — the grants this entire piece is about — was excluded.

That the famous ex-Google founders are class-of-2004 alumni. Systrom, Silbermann and most of the names on the “Xoogler founder” lists joined after the offering. The 2004 cohort’s characteristic second act was not founding Instagram. It was raising a fund.

Bottom line

Three cohorts, three mechanisms, and Google isolates the variable the others hide. PayPal tested what a small windfall does: single-digit millions at a young age, which buys the runway to build and not much else, and produced a generation of founders. WhatsApp tested what an enormous windfall does across a tiny group: nine figures at mid-career, which mostly buys disappearance. Google tested something neither could, because it is the only one of the three where the asset kept compounding for two decades and the door was open the whole time.

What it shows is that this cohort’s fortunes were decided by an instrument, but not the one people expect. Not the option grant — the option to sell. Google unlocked its rank and file in stages beginning two weeks after the offering, ahead of its own executives, and never bound any of them to an underwriter. That is a more generous and more honest arrangement than the industry standard, and it also meant these people were asked, year after year, to choose between a sum they could hold in their hands and a multiple they could not yet imagine. Most chose the sum. The stock went up 158 times.

The rescission letter is the whole thing in miniature. In November 2004 Google offered these people their purchase price back, and every one of them said no, because refusing was obviously correct — the stock was at $165 and the offer topped out at $80. It was the easiest decision any of them would be asked to make about Google stock, and it was the last one that was easy.

---

Methods and sources. Headcount, share counts, option counts, exercise prices and the selling-restriction schedules are from Google’s final IPO prospectus of 19 August 2004: 2,292 employees at 30 June 2004 (705 R&D, 1,141 sales and marketing, 446 general and administrative); 19,605,052 Class A shares sold at $85.00, of which 14,142,135 by Google and 5,462,917 by selling stockholders, plus an over-allotment of 2,929,626; 271,219,643 shares outstanding after the offering; options over 6,276,573 Class A shares at a weighted average exercise price of $9.42 and 10,456,084 Class B shares at $2.68, both as at 30 June 2004 rather than at the offering date; the statement, made twice, that no officers, directors, employees or stockholders signed lock-up agreements with the underwriters, alongside Google’s own 180-day issuer lock-up; the “sole discretion” release right; and the three separate selling-restriction schedules — 5%/10%/remainder at 15, 90 and 180 days for employees and other holders; 5% (10% for those with no vested holdings) at 90 days and the remainder at 180 for executive officers; and thirds at 90, 120 and 150 days for parties to the Investor Rights Agreement. The prospectus also notes employees could sell only vested shares. The aggregate table of shares becoming eligible on those dates combines all three groups and is not an employee schedule; this piece uses the underlying agreements instead. The 3,021 figure at 31 December 2004 is from Google’s 2004 Form 10-K, which also records that at the rescission offer’s expiry on 30 December 2004 no holders had accepted. Rescission terms — 23,435,945 shares held by 1,320 persons, unexercised options over 5,215,379 shares held by 282 persons, a repurchase price of $0.30 to $80.00 per share for shares and 20% of the exercise price for options, a maximum aggregate cost of $28.3 million, eligibility limited to grants made between September 2001 and July 2004 to residents of nineteen named states, and the $100.01–$196.03 closing range through 22 November 2004 with $165.10 the last close before the offer was dated — are from the rescission prospectus of November 2004. An earlier August filing put the total group at 1,406 persons and the maximum cost at $25.9 million; the figures moved between filings and this piece uses the November numbers throughout. The $3.3 billion and $1.45 million-per-head figures add the rescission share block to the net option value at $85 and are described as a floor, because the rescission population excludes pre-September-2001 grants and holders outside those nineteen states; no filing gives a complete count of employee-held shares. The Rule 701 findings, the $49 million and $33 million issuance figures, the $80 million total for 2002–2004, the “strategically disadvantageous” characterisation and the distinction between the January and June 2003 board meetings are quoted from the SEC’s cease-and-desist order of 13 January 2005, which was settled without admission or denial, credited the respondents’ cooperation, and imposed no monetary penalty. The transferable-option programme, its restriction to post-IPO grants and its exclusion of the executive management group are from Google’s announcement of 12 December 2006; the bidding institutions are named in the bidding rules agreement of April 2007. The “roughly 1,000 employees with more than $5 million” figure and Bonnie Brown’s story are from Katie Hafner’s New York Times report of 12 November 2007, summarised the same day by TechCrunch; the NYT piece is paywalled. They are a press estimate about Google’s 2007 workforce, not a disclosure and not a statement about the 2004 cohort. Split arithmetic treats the April 2014 Class C distribution as a two-for-one and the July 2022 split as twenty-for-one, giving twenty Class A and twenty Class C shares per 2004 share; the ~$13,480, 158x and 26%-a-year figures use the 4 September 2026 closes of $338.46 for Class A and about $335.72 for Class C, exclude the dividend Alphabet began paying in 2024, and will drift with the price. The $226 billion and “over $500 billion” figures are explicit counterfactuals — they assume universal exercise and permanent holding, which ten-year option terms and ordinary income tax made impossible — and are offered as a measure of what selling cost in aggregate, not as anyone’s wealth. Individual career dates are from public biographies and contemporaneous reporting rather than filings. No source discloses what the median class-of-2004 employee ultimately realised, and this piece does not estimate one. This draft was fact-checked line by line before publication, and the check changed two central arguments: an earlier version presented the prospectus’s aggregate five-date table as an employee release schedule when two of its rows cover venture investors only — the underlying agreements show a three-step employee schedule that in fact released rank-and-file staff seventy-five days before executive officers, a sharper finding than the one it replaced — and an earlier version measured the cohort’s paper wealth from the option pool alone, omitting the 23.4 million shares employees already owned outright and understating the total by more than half. The same pass corrected the option repurchase price, the date of the $165.10 close, the date of the $338.46 close, the claim that the SEC compelled the Playboy reprint, the eligibility of the transferable-option programme, the founding date of FriendFeed, and a passage that had placed Bonnie Brown among those who left with little.

Related reading: The Class Of: The PayPal Mafia · The Class Of: The WhatsApp 55 · Equity Compensation: RSUs, ISOs, and the Tech Wealth Engine · Sudden Wealth: Liquidity Events, Lottery Winners, Athletes, and Inheritance Shocks · Tech Wealth: How Founders and Investors Live Differently · Venture Capital: How the Money Behind Startups Actually Works