A company can finish a decade much bigger than it started and still have several years when revenue went backwards. I wanted to separate those two things: growth over a long period, and growth every single year.
So I took a sourced list of 100 large US companies and checked their annual revenue in SEC EDGAR's free XBRL data. The question sounds simple: how many increased revenue in each of their latest ten fiscal years?
Fourteen of the 78 companies with a full ten-year comparison window increased reported revenue every year: 17.9%. Of the original 100, I excluded 15 financial businesses and classified seven as having less than ten years of independent public history. All 85 non-financial companies now have an explicit classification; none remains unresolved.
That denominator matters. A recent listing or spin-off is not a failed growth streak. It simply has not supplied the same length of standalone reporting history.
The companies that passed
These fourteen companies had higher reported revenue in every one of the ten annual comparisons:
| Company | Ten-year revenue CAGR |
|---|---|
| Vertex Pharmaceuticals | 27.8% |
| Broadcom | 25.1% |
| Palo Alto Networks | 23.6% |
| Amazon | 21.0% |
| Netflix | 20.9% |
| Salesforce | 20.1% |
| Alphabet | 18.3% |
| Prologis | 14.9% |
| Microsoft | 13.8% |
| S&P Global | 11.2% |
| Costco | 9.8% |
| McKesson | 7.8% |
| Automatic Data Processing | 6.5% |
| Walmart | 4.0% |
The spread is what caught my attention. Walmart and Vertex both passed the same consistency test, but their growth rates were nowhere near each other. Walmart's annual revenue rose from $482.130 billion to $713.163 billion across its window. Vertex went from about $1.032 billion to $12.001 billion.
A yes-or-no streak hides those differences. It does not tell you whether the business grew slowly from a huge base or rapidly from a much smaller one. That is why I have shown the compound annual growth rate alongside the streak rather than ranking companies by the streak alone.
Very fast growth did not guarantee a perfect record
NVIDIA is a useful counterexample. In this extraction, its revenue rose from $5.010 billion in fiscal 2016 to $215.938 billion in fiscal 2026, an annualized increase of about 45.7%. That was substantially faster than any of the fourteen uninterrupted growers.
But revenue declined from $11.716 billion in fiscal 2019 to $10.918 billion in fiscal 2020. Its current streak is six annual increases, not ten.
Apple also failed the uninterrupted-growth test, despite ending its window much larger. Annual revenue fell in fiscal 2019 and again in fiscal 2023. Its latest three fiscal years showed increases. The distinction matters: “larger after ten years” and “larger every year” are different statements.
These are not rankings of business quality. A single difficult year can break a streak without settling anything about a company's future. Equally, ten increases do not establish that the next year will be another one.
How I counted
I used the first 100 companies in CompaniesMarketCap's US market-cap ranking, captured on 7 November 2026. This is a practical third-party universe, not an independently reconstructed index or a historical top-100 cohort.
I matched tickers to SEC identifiers using the SEC's company ticker file, then downloaded each company's Company Facts data. Requests identified the research contact and ran sequentially, below the SEC's ten-requests-per-second limit.
I kept US-dollar facts from Form 10-K filings with durations between 350 and 380 days, including annual comparative columns. Most are marked FY; I also admitted Q4-labelled facts when the form and duration established that they were annual, not quarterly. Importantly, I grouped observations by actual period-end date, not the API's fy field: a later filing can report an earlier year's revenue as a comparative figure.
For each fiscal period end, I preferred the newest-filed annual observation across the company's appropriate total-revenue tags. That lets an older SalesRevenueNet observation join a later contract-revenue observation without treating a tag rename as missing business history. I did not sum product and service components. Mastercard's gross contract revenue was excluded in favour of net revenue; NextEra used total operating revenue, not contract revenue alone.
I also joined predecessor SEC identifiers for Broadcom, Disney, Marvell and Exxon Mobil. A new holding company is not a new operating history. Deere had duplicated fiscal ends a few days apart; I consolidated those annual observations rather than inventing extra years. Its week-based year ends were not a missing year.
Ten annual increases require eleven revenue observations: a starting year and ten comparisons. Each eligible company therefore has its own latest eleven-observation fiscal window. Most end in calendar 2025; some end in fiscal 2026. CAGR is the last revenue divided by the first, raised to one tenth, minus one. Equal revenue breaks a streak.
What I checked, and what remains uncertain
I checked three latest annual figures against the original 10-K documents: Apple's fiscal 2026 net sales of $478.834 billion, Microsoft's fiscal 2026 revenue of $331.839 billion, and Amazon's calendar 2025 net sales of $716.924 billion. All three matched the extracted figures. This validates those particular observations, not every number in the dataset.
The seven short-history cases are Palantir, CrowdStrike, AppLovin, Uber, GE Vernova, Sandisk and SpaceX. For spin-offs, the classification refers to the current independent company, not the age of the underlying business. Pre-listing comparatives are retained in the audit file but do not manufacture a decade of standalone public history.
I excluded banks, insurers, brokers and asset managers, including mixed financial groups. Their interest income, premiums, investment gains and other revenue measures need a separate study. Payment networks and financial-information providers remained in the sample.
These are reported dollars, not organic or inflation-adjusted sales. Acquisitions, divestitures and accounting changes affect the series; newest-filed figures do not create a fully harmonized decade. Broadcom's history, for example, includes its Avago predecessor and substantial acquisitions. Today's largest companies also introduce selection bias. This is a measurement exercise, not investment advice.
I'm building StockPearl, a free, chart-first US fundamentals site using SEC EDGAR data; the beta is coming soon and the waitlist is at stockpearl.com. This small exercise is a reminder of why I want the underlying history visible: a strong endpoint can hide a very uneven path.
Top comments (0)