Intel: the rail that never fired

One company, its own filed accounts, and the method run on them in the open. This one is here for a specific reason: it is the clearest filed example of a guardrail that had never been false, in every year on record, right up until the filing in which it broke.

Nothing below is a view on Intel. The conditions in the third section belong to a hypothetical reader, not to ValueScreener.

The filed figures

Ten fiscal years of Intel 10-K filings are on record here, FY2016 to FY2025, with no gaps.

Margins in percent, free cash flow in billions of US dollars, current ratio as a ratio. Each row is rebuilt from the statements in the filing named on it. Read from GET /api/public/retrotest on 2026-09-16.

The two figures the argument rests on: gross margin 55.45 percent (FY2021 10-K, filed 2022-01-27 (opens in a new tab)) and, one filing later, 42.61 percent (FY2022 10-K, filed 2023-01-27 (opens in a new tab)).

What they say

The gross margin fell 12.84 points between two consecutive filings, FY2021 to FY2022, and that single step is larger than the total move across the five filings before it. From the FY2017 peak of 62.30 the line fell in seven consecutive filed years to 32.66 in FY2024, then rose 2.11 points to 34.77 in FY2025.

Everything else moved with it and none of it moved first. Operating margin went from 24.62 to 3.70 in the same pair of filings, a fall of 20.92 points. Free cash flow was positive in each of the six filed years to FY2021, adding to 86.03bn, and negative in each of the four since, adding to -44.30bn.

Placed against the sector it files alongside, the FY2016 gross margin of 61.01 sat between the Technology median of 48.19 and the 75th percentile of 68.73. The FY2025 figure of 34.77 sits between the 25th percentile of 29.33 and the median. The cohort is 631 companies, from GET /api/public/distribution/gross_margin with an as-of of 2026-09-15, read on 2026-09-16.

The reading a reader might arrive at: this is a change in the level at which the business produces, visible first and most clearly in the gross margin, and the cash statement agrees with the income statement about when it happened.

The guardrails a reader would write

gross_margin gte 50. Run backwards over the ten filed years, false in 4 of them: FY2022, FY2023, FY2024 and FY2025. The observed range is 32.66 to 62.30.

operating_margin gte 15. False in 4 of 10. FY2022, FY2023, FY2024, FY2025.

fcf gt 0. False in 4 of 10. FY2022, FY2023, FY2024, FY2025.

Read those three counts again, because the important thing about them is not the number. All three rails were false in exactly the same four filed years. A reader holding all three does not have three tests; they have one test written three ways, and they will be told the same news three times in the same week while believing they have corroboration.

The retro-test is what makes that visible before the fact rather than after it. Two conditions whose false years are the same set are measuring one thing about the company. Two conditions whose false years differ are measuring two.

current_ratio gte 1.5. False in 2 of 10, FY2019 and FY2024, neither of which is one of the four above. That is an independent rail: it broke once during the strongest stretch on this page and once during the weakest, which is what a figure that measures something else looks like.

There is a second lesson in the first three, and it is the reason this company is in the set. Six filed years, FY2016 through FY2021, every one of them above 50. A reader writing that rail at the FY2021 filing would have had a condition that was false in none of the years on record. It was false in the next filing, and in every filing since.

A retro-test that returns no breaks is a statement about the past. It is not a probability, it is not a safety margin, and the honest way to read "false in 0 of 10" is that nothing in the filed record tells you what this rail does next. That is what the fourth section of this format is for.

What would falsify this

The reading is that the gross margin change is a change in level rather than a point in a cycle, and that the cash statement is confirming rather than repeating it.

Two consecutive filed years with gross margin back above 50 would say it was a cycle. The FY2025 rise of 2.11 points is one year, and one year is what a cycle and a bottom look like from the same distance. A second consecutive rise of that size, with free cash flow positive in the same filing, is the shape that would settle it.

A filed year in which the gross margin rises while free cash flow stays negative would say the two are not one fact and that the second section conflated them. The claim above is that both moved for the same reason. That claim breaks the first time they move apart.

A restatement of any of the four years since FY2021 would break every count on this page at once, and this product would restate them: the figures here are rebuilt from whatever the company has most recently filed, so a corrected line changes the history rather than sitting beside it.

Note what is not on this list. The three rails firing again in FY2026 would not falsify anything written here; it is what the reading predicts. Only the first two items are about the reasoning.

What this leaves out

One company, ten filed fiscal years, and a method. Not a sample of anything, not a view on Intel, and not a recommendation to do anything at all.

Accounts filed before a company began tagging its statements in XBRL are not machine readable and are outside every figure on this page. Nothing here uses a share price, so nothing here is a valuation.

Free cash flow is stated in billions and rounded to two decimals from the filed figure. The four thresholds are lines chosen so the retro-test has something to answer, and none of them is a standard anybody published.

Past filings do not indicate future results, and this page is a worked example of why that sentence is not a formality: a condition false in none of six filed years was false in the seventh.