TL;DR. The Jeff Bezos story is usually told as a collection of anecdotes and quotes that are hard to trace. The verifiable version is more useful. Amazon was incorporated in July 1994 and started selling on its website in July 1995. Its 1997 annual report shows $147.8 million in net sales and a $27.6 million net loss. The company posted its first annual operating profit in 2002 and its first annual net profit, $35.3 million, in 2003, after about $3.0 billion of cumulative net losses. In its fiscal 2025 annual report, Amazon reported $716.9 billion in net sales, $80.0 billion in operating income and about 1,576,000 employees. Amazon Web Services (AWS) generated 18.0 percent of sales but 57.0 percent of operating income. Bezos handed the CEO role to Andy Jassy on July 5, 2021. The decision rules in his shareholder letters are the part a reader can actually reuse: separate irreversible “Type 1” decisions from reversible “Type 2” ones (2015 letter), decide most things with “somewhere around 70%” of the information you want, and “disagree and commit” (2016 letter). The CEO move: classify your next career decision as a one-way or two-way door before you spend time on it. The student move: write down what you expect, then review it, as the 1997 letter promised to “learn from both our successes and our failures.”
This piece connects to three CEOtudent tools: the regret minimization framework for big life decisions, kill criteria that decide in advance when to quit and a decision journal template for improving judgment.
Why was this article rewritten, and what changed?
The 2020 version was a sequence of captions and images: childhood anecdotes, a “richest person in the world” claim, unsourced motivational quotes and a growth story without numbers. Every claim was rechecked against Amazon’s 10-K filings on SEC EDGAR, the shareholder letters Bezos signed, Amazon’s 8-K filings on the 2021 CEO transition, and a 2001 interview transcript.
Removed: three quotes not found in any primary source (on “evolving”, “doing hard things” and “we are our choices”), the net worth claim, childhood anecdotes and broken images. Kept, now sourced to Bezos’s recorded words: the 2,300 percent web growth figure, the 30 percent odds and the regret-based reasoning. Added: a 1997-2025 ledger, a profit analysis, a decision test and a limits section.
What are the verified facts of Jeff Bezos’s path?
The table lists only dates that appear in a filing, a signed letter or the interview transcript.
| Date | Event | Source |
|---|---|---|
| January 12, 1964 | Born in Albuquerque, New Mexico | Academy of Achievement profile |
| Before 1994 | Princeton degree in computer science and electrical engineering; senior vice president at D. E. Shaw, a Wall Street firm | Academy of Achievement profile |
| Spring 1994 | Reads that web usage is “growing at 2,300 percent a year” (his recollection) | Bezos interview, May 4, 2001 |
| July 5, 1994 | Inception date of Amazon.com, incorporated in Washington State | Amazon 10-K, fiscal 1997 |
| July 1995 | First sales on the website | Amazon 10-K, fiscal 1997 |
| May 1997 | Initial public offering | Amazon 10-K, fiscal 1997 |
| 2002 | First annual operating profit ($64.1 million) | Amazon 10-K, fiscal 2002 |
| 2003 | First annual net profit ($35.3 million) | Amazon 10-K, fiscal 2003 |
| February 2, 2021 | Amazon announces Bezos will become Executive Chair in the third quarter | Amazon 8-K, February 2, 2021 |
| July 5, 2021 | Andy Jassy becomes President and CEO | Amazon 8-K filed July 2, 2021 |
Source: compiled from the documents listed in the Sources section.
The 2021 filing adds one detail that matters for the leadership lesson: Jassy, who joined Amazon in 1997, had led AWS “since its inception”, first as a senior vice president from April 2006 and as its CEO from April 2016. The successor was the person who built the most profitable part of the business.
How did Bezos make the 1994 decision, in his own words?
The best primary record of the founding decision is a long interview Bezos gave in San Antonio on May 4, 2001, published as a transcript by the Academy of Achievement. Three points are in his words there.
First, the trigger was a growth rate. He describes the “wake-up call” as a statistic that web usage in spring 1994 was growing at 2,300 percent a year. CEOtudent could not locate the original source of that statistic, so treat it as his recollection, not as a measured fact.
Second, he priced the risk before he took it. He told his parents, who invested, that there was a “70 percent chance” they would lose their whole investment. He set this against his estimate that only about 10 percent of start-ups succeed: “Here I was, giving myself a 30 percent chance.”
Third, he used what he called a “regret minimization framework”. When he told his boss, the boss argued it was a better idea for someone without a good job and convinced him to think about it for 48 hours. Bezos imagined himself at 80 looking back, and concluded he would not regret trying and failing, but might regret never having tried. He added a general claim: “Most regrets, by the way, are acts of omission and not commission.” He also named the short-term cost he ignored: leaving mid-year meant walking away from his annual bonus.
Two cautions. This is a founder describing his reasoning seven years later: evidence of method, not proof that the method causes good outcomes. And his downside was limited by a marketable degree and Wall Street experience. Our regret minimization framework guide adds the downside check the original story leaves out.
How long did Amazon take to make money?
Much of the popular story skips this part. Amazon’s own filings show nine consecutive loss periods before its first annual net profit.
| Year | Net sales (USD million) | Operating income or loss (USD million) | Net income or loss (USD million) |
|---|---|---|---|
| 1995 | 0.5 | -0.3 | -0.3 |
| 1996 | 15.7 | -6.0 | -5.8 |
| 1997 | 147.8 | -29.2 | -27.6 |
| 1998 | 609.8 | -109.1 | -124.5 |
| 1999 | 1,639.8 | -605.8 | -720.0 |
| 2000 | 2,762.0 | -863.9 | -1,411.3 |
| 2001 | 3,122.4 | -412.3 | -567.3 |
| 2002 | 3,932.9 | 64.1 | -149.1 |
| 2003 | 5,263.7 | 270.6 | 35.3 |
Source: Amazon 10-K filings for fiscal 1997, 2002 and 2003, as originally reported (rounded from thousands). The partial period from inception to December 31, 1994 showed a $52,000 net loss.
CEOtudent analysis of Amazon 10-K filings. Summing every net loss from 1994 to 2002 gives about $3.01 billion. The worst single year was 2000, with a $1.41 billion net loss. The first annual operating profit came eight years after incorporation and the first net profit nine years after. Even in 2003 the net margin was 0.7 percent. Sales grew at a compound annual rate of 81.4 percent from 1997 to 2003, so the losses came with very fast growth.
This is what “discipline” means in the Amazon record: a stated willingness to accept long losses in exchange for scale. The 1997 letter said this openly: Amazon would make investment decisions “in light of long-term market leadership considerations rather than short-term profitability considerations or short-term Wall Street reactions.” It also said that if forced to choose between the look of its accounting and “maximizing the present value of future cash flows, we’ll take the cash flows.”
How big did Amazon get, and where does the profit come from now?
The next table puts four annual reports side by side.
| Indicator | FY1997 | FY2003 | FY2015 | FY2025 |
|---|---|---|---|---|
| Net sales (USD million) | 147.8 | 5,263.7 | 107,006 | 716,924 |
| Operating income (USD million) | -29.2 | 270.6 | 2,233 | 79,975 |
| Net income (USD million) | -27.6 | 35.3 | 596 | 77,670 |
| AWS net sales (USD million) | n/a | n/a | 7,880 | 128,725 |
| AWS segment operating income (USD million) | n/a | n/a | 1,863 | 45,606 |
| Employees (full-time and part-time, year-end) | 614 | about 7,800 | about 230,800 | about 1,576,000 |
Source: Amazon 10-K filings for fiscal 1997, 2003, 2015 and 2025. In 2015, segment operating income excluded stock-based compensation; total segment operating income was $4,523 million against consolidated income from operations of $2,233 million. In 2025 the three segments sum to the consolidated $79,975 million.
CEOtudent analysis of Amazon 10-K filings.
| Derived measure | Value |
|---|---|
| Net sales growth, 1997 to 2025 | about 4,852 times |
| Compound annual sales growth, 1997 to 2025 (28 years) | 35.4% |
| Compound annual sales growth, 2015 to 2025 | 20.9% |
| Compound annual AWS sales growth, 2015 to 2025 | 32.2% |
| Operating margin, 1997 / 2003 / 2025 | -19.8% / 5.1% / 11.2% |
| AWS share of net sales, 2015 / 2025 | 7.4% / 18.0% |
| AWS share of segment operating income, 2015 | 41.2% |
| AWS share of operating income, 2023 / 2024 / 2025 | 66.8% / 58.1% / 57.0% |
| AWS operating margin, 2025 | 35.4% |
| Operating margin of the non-AWS segments combined, 2025 | 5.8% |
| Net sales per employee, 1997 / 2025 | $240,600 / $454,900 |
Three readings follow. First, the business the 1997 letter described, an online bookstore, is not where most of the profit comes from today. AWS makes 57.0 percent of operating income on 18.0 percent of sales, with a 35.4 percent margin against 5.8 percent for retail and advertising combined. Second, the share has moved: it was 66.8 percent in 2023, when the international segment lost $2.7 billion, and fell as retail profits recovered. Third, the 2015 and 2025 segment figures use different definitions, so the 41.2 percent and 57.0 percent are direction indicators, not a precise trend.
The lesson is not “build a cloud business.” It is that Amazon’s most valuable move was a second engine the original plan never mentioned, run by a long-tenured insider who later became CEO.
What did the shareholder letters actually say?
Popular summaries paraphrase these letters loosely. Here are the passages, quoted briefly and verbatim.
1997: “Day 1” and long-term thinking. The first letter, which Amazon attached again to its 2015 letter, opens by calling the moment “Day 1 for the Internet.” It lists principles that include focusing “relentlessly on our customers”, making “bold rather than timid investment decisions”, and a learning commitment: “We will continue to learn from both our successes and our failures.” It ends the list with ownership: each employee “must think like, and therefore must actually be, an owner.” It also admits the cost of the culture: “You can work long, hard, or smart, but at Amazon.com you can’t choose two out of three.”
2015: Type 1 and Type 2 decisions. Bezos writes that some decisions are “consequential and irreversible or nearly irreversible”, one-way doors that must be made “methodically, carefully, slowly.” Most are not: they are reversible two-way doors, and “Type 2 decisions can and should be made quickly by high judgment individuals or small groups.” The risk he names is large organizations using the heavy process for everything, which leads to “slowness, unthoughtful risk aversion, failure to experiment sufficiently.” The same letter calls failure and invention “inseparable twins.”
2016: Day 2 and high-velocity decisions. The letter defines the threat bluntly: “Day 2 is stasis. Followed by irrelevance.” Its “starter pack” for staying in Day 1 is customer obsession, a skeptical view of proxies, the eager adoption of external trends, and high-velocity decision making. Under the last heading it gives two rules. Most decisions “should probably be made with somewhere around 70% of the information you wish you had. If you wait for 90%, in most cases, you’re probably being slow.” And: “use the phrase ‘disagree and commit.’” It adds that “being wrong may be less costly than you think, whereas being slow is going to be expensive for sure”, provided you are good at course correction.
How can you use these rules for your own career decisions?
The letters were written for a company. The following translation to one person’s career is ours.
CEOtudent editorial framework: Type 1 or Type 2? A personal classification table.
| Decision | Usually | Why | Process to use |
|---|---|---|---|
| Trying a new tool, course module or side project for 30 days | Type 2 | Cheap to reverse; you learn either way | Decide in a day with about 70% of the information |
| Asking for a stretch assignment inside your current job | Type 2 | Reversible; little downside if it fails | Decide quickly, set a review date |
| Changing team or role inside the same employer | Mostly Type 2 | Often reversible within 6 to 12 months | Short pre-mortem, then commit |
| Starting a degree or long certification | Mixed | Time and money are sunk, but you can stop | Set kill criteria before enrolling |
| Leaving a stable job to start a company | Type 1 | Income, savings and reputation are hard to recover | Slow: written plan, downside budget, regret test |
| Moving country, taking on large debt, signing a long contract | Type 1 | Costly or impossible to undo | Slow: consultation, a pre-mortem, cooling-off period |
CEOtudent editorial framework: the five-question decision test. Answer before you spend more than an hour on any decision.
- Door test. If this goes wrong, can I get back to roughly where I am within a year at a cost I can afford? Yes means Type 2. No means Type 1.
- Information test. For a Type 2 decision, do I already have about 70 percent of what I wish I knew? If yes, decide now. Waiting for 90 percent is the slowness the 2016 letter warns about.
- Downside budget. For a Type 1 decision, have I written down the maximum loss I accept, in money and months? Bezos priced his own odds at 30 percent before starting.
- Regret test. Projected to age 80, which choice would I regret more, the omission or the attempt? Use this only after the downside budget, not instead of it.
- Commit and review. Have I written my expected outcome and a review date in a decision journal? If others disagree, can I state my view, then disagree and commit?
Scoring. For a Type 2 decision, questions 1, 2 and 5 are enough; decide within a week. For a Type 1 decision, all five must be answered in writing, and at least one other person should read the answers.
Where does the “student” side of the story show up?
The better evidence of learning is in what Amazon wrote about itself.
- Learning was written into the operating principles. The 1997 letter promises to measure programs “analytically”, to drop those that do not earn acceptable returns, and to learn from successes and failures. That is a feedback loop, not a personality trait.
- The company admitted what it did not know. The same letter says: “We now know vastly more about online commerce than when Amazon.com was founded, but we still have so much to learn.”
- External trends counted as a learning duty. The 2016 letter lists “the eager adoption of external trends” among the four Day 1 defenses.
- Course correction beats perfect foresight. The 70 percent rule only works if you are good at “quickly recognizing and correcting bad decisions.” For an individual this means a regular review, such as an annual personal review with a CEO-style operating rhythm.
For a reader in 2026 the student move is concrete: pick one external trend affecting your field, run a Type 2 experiment with it this month, and log what you expected and what happened.
What does this story not prove?
- Survivorship bias. Amazon is one of very few companies that survived nine loss periods and the 2000 crash. Many firms that also invested for the long term failed. Bezos’s 2015 letter makes the same point in a footnote about its own decision argument: “there is undoubtedly some survivorship bias.” Use the rules as decision hygiene, not as a promise of results.
- Starting conditions. By his own account, Bezos left a senior Wall Street role, and his parents could invest a few hundred thousand dollars. Most people facing a Type 1 decision have a smaller safety net, which makes the downside budget more important, not less.
- Antitrust scrutiny. In September 2023 the U.S. Federal Trade Commission and 17 state attorneys general sued Amazon in federal court in Seattle, alleging that it unlawfully maintains monopoly power. Amazon’s 10-K says it disputes the allegations. Separately, Amazon recorded a $2.5 billion charge in the third quarter of 2025 for the settlement of a lawsuit with the FTC, equal to 3.1 percent of its 2025 operating income.
- Labor questions. Amazon’s fiscal 2025 10-K lists “labor and employment (including the characterization of delivery drivers)” among the areas of claims and proceedings it faces. The 1997 letter’s line about not being able to choose two of “long, hard, or smart” describes a demanding culture in the company’s own words. “Disagree and commit” is a tool for teams with real voice; it can also be misused to silence dissent.
- Headcount. Employee counts include part-time staff, so sales per employee is a rough indicator.
Frequently asked questions
When did Amazon first make a profit?
Amazon reported its first annual operating profit, $64.1 million, in 2002, and its first annual net profit, $35.3 million, in 2003. Net losses from 1994 to 2002 add up to about $3.0 billion.
What is the difference between Type 1 and Type 2 decisions?
In the 2015 shareholder letter, Type 1 decisions are consequential and irreversible or nearly so, and should be made slowly. Type 2 decisions are reversible and should be made quickly by individuals or small groups.
What does “disagree and commit” mean?
From the 2016 letter: when there is no consensus, you state your view and then commit fully to the team’s direction, so the decision moves without everyone having to be persuaded. It applies to the boss as well.
What is the regret minimization framework?
Bezos’s name, in a 2001 interview, for projecting himself to age 80 and asking which choice he would regret. He concluded he would not regret trying and failing, but would regret not trying.
How much of Amazon’s profit comes from AWS?
In 2025, AWS reported $45.6 billion of the $80.0 billion operating income, 57.0 percent, on 18.0 percent of net sales.
When did Bezos stop being CEO?
Amazon announced the transition on February 2, 2021. Andy Jassy became President and CEO on July 5, 2021, and Bezos became Executive Chair.
Sources
- Amazon.com, Inc. 1997 Annual Report, Letter to Shareholders (SEC filing ARS, April 1998; reprinted with the 2015 Letter to Shareholders).
- Amazon.com, Inc. Form 10-K405 for fiscal year 1997, filed with the SEC on March 30, 1998.
- Amazon.com, Inc. Form 10-K for fiscal year 2002, filed with the SEC in February 2003.
- Amazon.com, Inc. Form 10-K for fiscal year 2003, filed with the SEC in February 2004.
- Amazon.com, Inc. Form 10-K for fiscal year 2015, filed with the SEC on January 29, 2016.
- Amazon.com, Inc. Form 10-K for fiscal year 2023, filed with the SEC on February 2, 2024 (legal proceedings).
- Amazon.com, Inc. Form 10-K for fiscal year 2025, filed with the SEC on February 6, 2026.
- Amazon.com, Inc. 2015 Letter to Shareholders, Jeffrey P. Bezos.
- Amazon.com, Inc. 2016 Letter to Shareholders, Jeffrey P. Bezos.
- Amazon.com, Inc. Form 8-K dated February 2, 2021, Item 5.02, and Exhibit 99.1 press release.
- Amazon.com, Inc. Form 8-K dated June 28, 2021 (filed July 2, 2021), Item 5.02.
- Academy of Achievement, interview with Jeff Bezos, San Antonio, Texas, May 4, 2001 (transcript), and Jeffrey P. Bezos profile.
- U.S. Federal Trade Commission, press release “FTC Sues Amazon for Illegally Maintaining Monopoly Power”, September 26, 2023.
Method note: the loss ledger totals, growth multiples, compound growth rates, margins, AWS shares and sales per employee are CEOtudent analyses of Amazon 10-K filings, recomputed with a reproducible script. The Type 1 or Type 2 classification table and the five-question decision test are CEOtudent editorial frameworks. Deliberately not used: the motivational quotes in the previous version that could not be traced to a primary source, net worth rankings, and U.S. Department of Labor safety citations, whose official releases could not be opened for verification.
This content was compiled with the support of AI following in-depth research, then written and prepared for publication by the CEOtudent editorial team.
This post is also available in:
){kind=link}
&description=&image=https://ceotudent.com/wp-content/uploads/2021/07/6rz9fr-1024x609.jpg){kind=link}