Autodesk ADSK

Reference table

1. The business, and both segment blocks tie exactly

Autodesk provides design and make software for architecture, engineering, construction, manufacturing, and media and entertainment, sold predominantly by subscription. Chief executive Andrew Anagnost; chief financial officer Janesh Moorjani (per the 8-K signature); approximately 14,300 employees; Nasdaq-listed since 1985-06-28; fiscal year ends 31 January.

seg_prod sums to revenue to the dollar:

seg_geo also sums to revenue to the dollar, with the home region present:

Both blocks total $7,206,000,000 against inc_a FY2026 revenue of $7,206,000,000. On a batch in which one geography block summed to 166.1% of revenue, another dropped a segment worth 22.6%, and two were empty arrays, Autodesk's are exact and complete — and that is reported as a finding.

The composition matters. Architecture, Engineering and Construction is now half the company and grew 22.0% — three points faster than the group and eight points faster than AutoCAD. That is the platform displacing the licence. EMEA grew 21.1%, faster than the Americas, which is unusual and worth watching. Media and Entertainment, at $332M and growing 5.4%, is now immaterial.

The record:

Revenue growth ACCELERATED from 9.8% to 17.5% across three years and the operating margin expanded 10.8 points from the fiscal 2022 trough. The fiscal 2021 EPS of $5.44 is distorted by a $661 million tax BENEFIT and is not comparable.

And the first quarter of fiscal 2027 is the sharpest data point in the file:

A 729 basis point operating-margin expansion in one year is not organic operating leverage alone — it is the year-earlier quarter carrying restructuring charges. Section 2 establishes that from the filing.

2. What the payload does not carry: two restructurings and an acquisition that closed yesterday

The restructurings

The 10-K filed 2026-03-03 names TWO separate plans and the payload contains no restructuring line item at all:

> "...restructuring plan (the '2026 Plan'), to support Autodesk's initiatives to optimize its go-to-market organization and, at the same time, to reallocate resources to Autodesk's strategic pr[iorities]..."

> "...restructuring plan (the 'January 2026 Plan') that marks the final phase of our sales and marketing optimization program..."

> "...restructuring plans initiated during the fourth fiscal quarter ended January 31, 2026 ('January 2026 Plan') and during the first fiscal quarter ended April 30, 2025 ('2026 Plan')..."

So: a plan begun in the quarter ended 2025-04-30 and a second plan begun in the quarter ended 2026-01-31, described as the final phase. That dating explains the comparison precisely: the April 2025 quarter — the base for the 60.1% operating income growth just reported — was the quarter in which the first plan was initiated, and the January 2026 quarter carried the second.

The consequence for this dive is stated rather than smoothed: the 28.0% first-quarter operating margin cannot be cleanly separated from the charges that depressed the comparison, because the payload isolates no restructuring expense and the extract does not carry the charge tables. The margin improvement is real in direction and unquantifiable in composition here, and one quarter is one quarter.

The acquisition, completed one day before this dive

From the 8-K filed 2026-08-03:

> "On August 3, 2026, Autodesk, Inc. completed the previously announced acquisition of MaintainX Inc. ('MaintainX'), pursuant to the Agreement and Plan of Merger, dated as of May 28, 2026, by and among the Company, Matterhorn Acquisition Corp., a Delaware corporation and a wholly-owned subsidiary of the Company, MaintainX and Shareholder Representative Services LLC..."

That is the entire disclosure. No consideration, no share count, no expected revenue contribution, no goodwill allocation.

And from the 8-K filed 2026-07-13, three weeks earlier:

> "On July 13, 2026, Autodesk, Inc. established an unsecured commercial paper program... the Company may issue, from time to time, unsecured commercial paper notes with varying maturities not in excess of 365 days from the date of issue... with the maximum aggregate face or principal amount of Notes outstanding at any one time not exceeding $2.0 billion... The Company expects to use the proceeds of the Notes for general corporate purposes, including to partially finance the transactions..."

Three things follow.

First, the acquisition is debt-financed at least in part, and the facility is $2.0 billion. Autodesk held approximately $114 million of NET CASH at 2026-01-31 on the corrected basis; drawing materially on a $2.0 billion facility moves it to net debt.

Second, the size is genuinely unknown to this dive. A $2.0 billion facility established "to partially finance" a transaction is an upper bound on the commercial-paper component and nothing more. No fair-value anchor in this dive embeds any MaintainX contribution, dilution or leverage, and that limitation is stated.

Third, the timing is unusually tight for a reader. The acquisition closed on 2026-08-03 and this dive is dated 2026-08-04. The 2026-08-27 earnings report, 23 days away, will be the first document to size it.

For context on the acquisition cadence: acquisitionsNet reads $96M (FY2023), $70M (FY2024), $825M (FY2025) and $0 (FY2026). The fiscal 2025 figure of $825 million shows the company is willing to spend, and MaintainX follows a year in which it did not.

3. Stock-based compensation, and why every cash-flow figure has two versions

This is the most important analytical point on the name and it is arithmetic rather than opinion.

Stock-based compensation of $788 million is 10.9% of revenue — the highest ratio in this batch — and stockBasedCompensationToRevenueTTM of 9.50% confirms the trailing figure.

The consequences, stated three ways:

• Free cash flow. $2,409M gives a 4.81% yield and a 33.4% margin; $1,621M gives a 3.24% yield and a 22.5% margin. The knowledge base's one claim requires a 30-40% margin — Autodesk passes on the first measure and fails on the second.
• Earnings. Trailing GAAP diluted EPS is approximately $6.86 (summing the four quarters in inc_q) against trailing adjusted EPS of approximately $11.13 (summing earn_cal) — a 62% gap. At $237.19 that is 34.6x GAAP and 21.3x adjusted.
• Forward multiples. All consensus estimates are adjusted, so 18.8x FY2027E and 16.7x FY2028E exclude roughly $3.70 per share of pre-tax stock compensation. Adding back an after-tax equivalent of roughly $2.75 gives an economic FY2027 figure near $9.85 and a multiple near 24.1x.

This dive uses the adjusted basis for comparability with the estimates and states the economic figure alongside it every time. Neither treatment is dishonest; presenting only one is.

One further cash-flow observation. Operating cash flow is volatile: $2,071M, $1,313M, $1,607M, $2,452M across four fiscal years — a 37% fall followed by a 22% rise and an 53% rise — while net income moved $823M, $906M, $1,112M, $1,124M. incomeQualityTTM of 1.90 reflects a deferred-revenue and billings cycle that does not track earnings, and the fiscal 2026 figure is the high end of a wide band rather than a new baseline.

4. Balance sheet — net cash, about to change

The vendor's totalDebt adds $251M of lease obligations to filed debt — a 9.2% inclusion, the mildest in this batch — and netDebt of $485M subtracts cash only, omitting $348M of short-term investments. On the filed basis Autodesk held NET CASH of approximately $114 million at fiscal year end. enterpriseValueTTM of $50,134.5M should be approximately $49,967.5M.

That position is about to change. A $2.0 billion commercial paper programme established 2026-07-13 to part-finance an acquisition that closed 2026-08-03 will show up as debt in the next balance sheet, and this dive cannot size it.

Tangible book value is NEGATIVE $7.59 per share — goodwill and intangibles of $4,762M against equity of $3,045M — and tangibleAssetValueTTM reads MINUS $1,601M. Total equity is positive, so the equity ratios are usable and are not rejected, unlike three other names in this batch. For an acquisitive software company, negative tangible book is normal and is reported rather than treated as a defect.

Capital return is buyback-only: $1,402M in fiscal 2026, 2.80% of market capitalisation, against $852M and $795M in the two prior years. commonDividendsPaid reads $0 in all four years — Autodesk has never paid a dividend. Total shareholder yield is 2.80%, entirely repurchase.

5. Valuation

At $237.19:

Coverage is good on the anchor years — 19 and 20 analysts on FY2027 and FY2028 EPS — and the FY2027 range of 1.4% is the second-tightest in this batch. FY2029 rests on 10 and FY2030 on 4.

est.ebitdaAvg and est.ebitAvg are rejected for the fixed-ratio fabrication signature. ebitdaAvg is exactly 22.99% of revenueAvg and ebitAvg exactly 20.08% in FY2025 through FY2031. The implied EBITDA margin of 22.99% is 5.9 points BELOW the reported ebitdaMarginTTM of 28.88%, so the row is not merely formulaic but wrong in level and in the conservative direction.

And the FY2031 row is ZEROED. It reports netIncomeAvg of 0, epsAvg of 0, epsLow of 0 and epsHigh of 0, while carrying a revenue estimate of $12,075M from 7 analysts and an EPS analyst count of 5. A row with five EPS analysts and a zero EPS estimate is a null, not a forecast, and it is rejected.

One further estimate note: est FY2026 epsAvg of $10.23605 against a delivered adjusted figure of approximately $10.99 (summing the four fiscal-2026 quarterly actuals in earn_cal: $2.29 + $2.62 + $2.67 + $2.85 = $10.43, plus timing differences) — the array runs modestly below delivered results, consistent with five consecutive beats averaging 6.6%.

5a. What today's price assumes (the inversion)

At $237.19 — 18.8x FY2027 adjusted consensus, 34.6x trailing GAAP — the price embeds:

• Adjusted EPS reaches $12.601 in fiscal 2027 and $14.200 in fiscal 2028. (Consensus; 19 and 20 analysts.) The first quarter delivered $2.99, so the year needs $9.61 across three quarters against a consensus of $3.12 for the second — comfortably on track, and the last five prints beat by an average of 6.6%.
• The first-quarter 28.0% operating margin is durable. (Our derivation.) This is the most fragile assumption in the price, because it followed two restructuring plans and compares against a quarter in which the first plan was initiated. A reversion to the fiscal 2026 full-year 24.90% would cut roughly $250 million of annualised operating income.
• Stock-based compensation stays near $788 million. (Our number.) At 10.9% of revenue it is the difference between a 4.81% and a 3.24% free cash flow yield, and between 18.8x and roughly 24x forward earnings. The price on the adjusted basis assumes the market continues to ignore it; on the economic basis the stock is not cheap.
• MaintainX does not impair the balance sheet. (Our number; NOT sizeable from this archive.) An unknown consideration, part-funded from a $2.0 billion facility, closing one day before this dive.
• The market keeps paying 13–23x forward adjusted earnings. (Our number.) The observed twelve-month range on the FY2028 estimate is exactly that: the 52-week low of $187.72 is 13.2x and the high of $326.79 is 23.0x. The price sits at 16.7x, in the lower third.

5b. The return bridge (why the multiple moves)

Expected return over the next twenty-four months decomposes as: adjusted EPS growth (+29.8% cumulative, from FY2027E $12.601 to FY2029E $16.357) + multiple drift (roughly HELD at 17-19x on the then-forward year) + shareholder yield (+2.80% per year, entirely buyback) ≈ +14% to +18% a year if the base lands.

Our base of $270 applies 19.0x FY2028E, a modest expansion from today's 16.7x, and it sits inside — not above — the 13.2x-to-23.0x band the market has paid on that estimate within the last twelve months.

At an unchanged 16.7x the fair value is $237 — the current price — so the entire 13.8% base upside is two and a third multiple turns. That is a larger multiple assumption than most of the bases in this batch and it is named as the fragile leg.

Note also that the shareholder yield here is a buyback funded by a company that pays no dividend and has just borrowed to acquire — so its continuation is not guaranteed at the fiscal 2026 rate of $1,402 million.

5c. Variant perception (where we differ, what would surprise)

• We are 13.0% below the street's $310.20 and the difference is entirely the multiple. Our EPS numbers are the consensus.
• We think the stock-based compensation adjustment is being applied too generously by the market and we quantify it. $788 million, 10.9% of revenue, is the highest ratio in this batch and it turns a 4.81% free cash flow yield into 3.24% and an 18.8x forward multiple into roughly 24x. Watchable number: stock compensation in the 2026-08-27 report against a $197 million quarterly run rate.
• We think the first-quarter margin is being extrapolated without its context. A 729 basis point year-on-year operating-margin expansion compares against a quarter in which the "2026 Plan" restructuring was initiated. The 10-K names two plans; the payload isolates neither; nobody should treat 28.0% as a new baseline on one observation.
• We are agnostic on MaintainX because we have to be. No consideration is disclosed anywhere in this archive. A dive that assumed accretion or dilution on a transaction it cannot size would be inventing a number, and this one does not.
• Positive surprise that would force a re-rate: a second quarter of operating margin above 27%; MaintainX disclosed at a price implying a reasonable multiple with a named revenue contribution; stock compensation falling below 10% of revenue; or Architecture, Engineering and Construction sustaining 22% growth.
• Negative surprise that would break the thesis: operating margin back below 24%; a MaintainX consideration above roughly $2 billion with minimal disclosed revenue; a third restructuring plan; stock compensation above 12% of revenue; or a break of the 50-day average at $215.65.

Synthos fair values

All three anchors are multiples of the FY2028 consensus adjusted EPS distribution (mean $14.2004, low $13.58387, high $14.7765, 20 analysts), cross-checked against the observed twelve-month multiple range of 13.2x–23.0x on that same estimate.

• Bear ~$179 — 13.2x the FY2028 consensus LOW of $13.584, which is precisely the multiple paid at the 52-week low. Cross-check: 4.6% below the 52-week low of $187.72; 14.2x FY2027E; a 2.75% SBC-adjusted free cash flow yield. The scenario: the first-quarter margin proves to be a restructuring artefact, MaintainX is expensive, stock compensation stays above 10% of revenue, and the market reprices toward the bottom of the band it set this year. −24.5%.
• Base ~$270 — 19.0x the FY2028 consensus MEAN of $14.200. Cross-check: 21.4x FY2027E; 16.5x FY2029E; 13.0% below the street's $310.20; 17.4% below the 52-week high. The scenario: revenue compounds at the consensus 10-14%, the operating margin settles between the fiscal 2026 24.9% and the first-quarter 28.0%, MaintainX is neutral, and the multiple recovers two and a third turns toward the middle of its own annual band. +13.8%.
• Bull ~$340 — 23.0x the FY2028 consensus HIGH of $14.777, precisely the multiple paid at the 52-week high. Cross-check: 4.0% above that high of $326.79; 27.0x FY2027E; 9.6% above the street's $310.20. The scenario: the 28.0% margin holds, Architecture, Engineering and Construction compounds above 20%, MaintainX adds a growth line, and a 91%-gross-margin franchise re-rates to where the market valued it a year ago. +43.3%.

Base is 13.8% above spot; asymmetry roughly 1.77:1 (24.5% down, 43.3% up), plus a 2.80% buyback yield. A base above 13% is genuinely attractive and a payoff ratio of 1.77:1 is below the 2:1 bar. That combination is a Watch with a named trigger, not a Buy, and Section 9 gives the price.

6. Knowledge base — one claim, two numerical tests, and it passes both

Raw hits: 1. After a case-sensitive entity re-run: 1. Discarded: 0. Homograph collisions: 0.

The sweep ran ADSK, Autodesk and AutoCAD plus free text on "autodesk" across all 52,021 distilled claims.

> 2024-02-08 · NEUTRAL · conviction 55 · horizon: principle · bill_gurley · speaker aaron_levie · role independent · skill 1.0

> "Durable compounders growing 10-20% at 30-40% free-cash-flow margins that dominate a vertical for decades (Adobe, Autodesk, Ansys, Cadence, Synopsys) are underappreciated great businesses."

> — entities: ANSS, CDNS, SNPS, ADBE, ADSK

That is the entire lane — and unusually for a lane of one, it is gradeable, because it specified two numerical criteria in advance.

On the conventional measure — operating cash flow minus capital expenditure over revenue — Autodesk passes both tests, and is the only claim in this batch to have specified two numbers in advance and cleared both. On the economic measure, deducting the $788 million of stock-based compensation the conventional measure adds back, the free cash flow margin is 22.5% and the claim fails its second test.

Both readings are reported and neither is preferred, because the claim did not specify which basis it meant and the difference is 10.9 points of revenue.

What the lane does NOT contain, and it is a long list. No view on the two restructuring plans. No view on the MaintainX acquisition. No view on whether Architecture, Engineering and Construction is a durable platform or a cyclical construction proxy. No view on stock-based compensation. And — most importantly for a design-software company in 2026 — NOT ONE CLAIM in 52,021 addressing whether generative design tooling expands or erodes the seat-based licence model. That is a genuine analytical void and it is the reason the exponential score is 5 rather than higher or lower: this dive has no basis for a view.

Attribution is unusually good for a single claim. It carries a NAMED speaker with speaker_role: independent, and the content is an investor's category observation rather than company guidance, so the management half-weighting rule does not engage and the claim is weighted at full. It is nonetheless twenty-nine months old and names Autodesk as one of five examples.

Conclusion. Breadth 1, claim count 1, net conviction a single claim, tested on its own terms, passing on one measure and failing on the other. All three horizon confidence ratings are set to Low, because a lane of one cannot function as a check on the filings.

7. Data integrity

Seven findings: two clean checks and five defects.

CLEAN 1 — seg_prod AND seg_geo both sum to revenue TO THE DOLLAR. Both total $7,206,000,000 against inc_a FY2026 revenue of $7,206,000,000, with Americas present at $3,178M (44.1%). Both the over-counting and home-market omission tests pass, on both blocks. On a batch containing a 166.1% geography over-count, a block missing a 22.6% segment, and two empty arrays, Autodesk's segment data is the best in the set.

CLEAN 2 — the capital-expenditure diagnostic is low but explicable and trips no fatal signature. capexToDepreciationTTM reads 0.263 and the direct fiscal 2026 computation is $43M / $195M = 0.221. Capital expenditure of $43 million is 0.6% of revenue, which is normal for a company whose product is code and whose infrastructure is largely leased or purchased as a service. Capex is not zero, not positive-signed, and although the ratio is below the 0.5 level that would be normal for an asset-owning business, the absolute figure is credible for software. The datum is accepted with the qualification below.

DEFECT 1 — investmentsInPropertyPlantAndEquipment and capitalExpenditure DISAGREE in fiscal 2024. The FY2024 row reports −$61,000,000 for the first and −$31,000,000 for the second — a $30 million, 97% divergence — while the two fields match exactly in FY2023, FY2025 and FY2026. One of them is wrong for that year, and the file gives no way to determine which. Free cash flow for FY2024 is therefore $1,282M or $1,252M depending on the field chosen; this dive uses the capitalExpenditure field for consistency with the other years and flags the discrepancy.

DEFECT 2 — the FY2031 estimate row is ZEROED. It reports netIncomeAvg of 0, epsAvg of 0, epsLow of 0 and epsHigh of 0, while simultaneously reporting a revenue estimate of $12,075M from 7 analysts and an EPS analyst count of 5. Five analysts cannot collectively forecast zero earnings for a company earning $1.1 billion. The row is a null and is rejected.

DEFECT 3 — est.ebitdaAvg and est.ebitAvg carry the fixed-ratio fabrication signature AND are wrong in level. ebitdaAvg is exactly 22.99% of revenueAvg and ebitAvg exactly 20.08% in FY2025 through FY2031. The implied EBITDA margin of 22.99% is 5.9 points BELOW the reported ebitdaMarginTTM of 28.88% — unusual, because this fabrication signature normally overstates. Rejected; all valuation uses epsAvg.

DEFECT 4 — totalDebt includes $251M of lease obligations and netDebt omits short-term investments. Filed long-term debt is $2,483M; totalDebt reads $2,734M, a 9.2% inclusion — the mildest in this batch. netDebt of $485M subtracts cash only, ignoring $348M of short-term investments. On the filed basis Autodesk held NET CASH of approximately $114 million, and enterpriseValueTTM of $50,134.5M should be approximately $49,967.5M.

DEFECT 5 — no restructuring line item exists in the payload for a company that ran TWO restructuring plans in fiscal 2026. The 10-K names the "2026 Plan" (initiated in the quarter ended 2025-04-30) and the "January 2026 Plan" (initiated in the quarter ended 2026-01-31, described as the final phase of the sales and marketing optimisation programme). Neither inc_a nor inc_q carries a restructuring expense line, so the 729 basis point year-on-year operating-margin expansion reported in the first quarter of fiscal 2027 cannot be decomposed into organic leverage and a favourable comparison. This is stated as a limitation on the analysis rather than a vendor error.

Also noted: inc_a reports interestExpense of ZERO for both FY2026 and FY2025 on a company with $2,483M of long-term debt; inc_q reports $0 or $2M per quarter, and interestCoverageRatioTTM of 998.5x is a consequence of a near-zero denominator rather than a meaningful ratio. Interest expense is genuinely low because much of the debt is old and low-coupon, but exactly zero for a full year is not credible and the coverage ratio is not used. tangibleBookValuePerShareTTM of MINUS $7.588 is CORRECT and is reported rather than rejected, because total equity is positive and negative tangible book is normal for an acquisitive software company. dividendYieldTTM and dividendPerShareTTM of zero are CORRECT — Autodesk has never paid a dividend. km_ttm and ratios_ttm are present and complete. quote.yearHigh/yearLow ($329.09/$185.50) against tech.hi52/lo52 ($326.79/$187.72), a 0.7% and 1.2% discrepancy; tech is used throughout.

Vendor composite rating — noted. B+ / 3 overall, with 5 out of 5 on both return on equity and return on assets and 4 on discounted cash flow, and 1 out of 5 on debt-to-equity and price-to-book. The discounted-cash-flow score of 4 rests on the free cash flow figure this dive has restated on an SBC-adjusted basis and should be read with that in view.

Non-equity tripwire — checked and passed. ADSK is common stock, Nasdaq-listed since 1985-06-28. Price of $237.19 is not par-like; beta is 1.318; there is no dividend; volume was 1.46M shares (roughly $346M of turnover); the 52-week band of $187.72 to $326.79 is a 74% range. This is common equity.

8. Technicals and insiders

• Price $237.19, up 1.06% on 1,460,348 shares. Opened $231.19, ranged $229.70 to $237.72, closed within 0.2% of the day high.
• −27.3% from the 52-week high of $326.79; +26.5% above the 52-week low of $187.72. Position within the annual range: 35th percentile.
• Between the moving averages: +10.0% above a 50-day average of $215.65 but −6.3% below a 200-day average of $253.14. A stock that has rallied 10% off its short-term trend and is now pressing against its long-term one from beneath.
• RSI 67.72 — the second-highest in this batch. MACD +7.09, strongly positive.
• Maximum drawdown from peak over the trailing year: −30.6%.
• Relative performance: 3-month −4.1% against SPY +7.6%; 6-month −7.1% against SPY +11.1%; 12-month −20.1% against SPY +24.3% and QQQ +30.8% — a 44.4-point deficit. The deficit is 44 points over twelve months, 18 over six and 12 over three: the underperformance is decelerating but has not stopped.
• Sentiment: 38 buy, 9 hold, 4 sell across 51 analysts — the most positive rating distribution in this batch at 74.5% buy; consensus target $310.20 (+30.8%), median $302.50, high $369, low $262 — 10.5% ABOVE spot. Not one of 51 analysts has a target below the price.

Insiders — one genuine open-market purchase

Seven of the eight transactions are annual director equity awards granted on 2026-06-17, the date of the annual meeting, at a price of zero. They are mechanical and signal-free.

The eighth is a genuine open-market PURCHASE. John T. Cahill bought 2,000 shares at $189.20 on 2026-06-23, taking his holding from 2,000 to 4,000 — a 100% increase. Approximately $378,000. The stock closed today at $237.19, so he is 25.4% ahead.

Two qualifications and one observation. $378,000 is a modest sum for a director of a $50 billion company, and one purchase is not a pattern. He now holds 4,000 shares — approximately $949,000 — which is a small position by the standards of this board, where another director holds 29,157. But it is a purchase, it is the only one in the file, and it is the second genuine open-market insider buy this batch has found across roughly ninety transactions. It is reported at that weight: real, small, and favourable.

What the file does not contain: any transaction by Andrew Anagnost, the chief executive, or by Janesh Moorjani, the chief financial officer, in either direction — and no open-market sale by anyone.

Governance, from the 8-K filed 2026-06-18, which is not in the payload. At the 2026 annual meeting on 2026-06-17, eleven directors were elected and a Certificate of Amendment providing for officer exculpation as permitted by Delaware law was filed with effect the same day. Andrew Anagnost received 171,597,422 votes for and only 741,011 against — 0.4% opposition, the strongest support of the eleven. Stacy J. Smith received 23,865,848 against — 13.9%, by far the highest — and John T. Cahill, the director who subsequently bought stock, received 9,170,566 against.

9. Verdict, kill-criteria and flip conditions

Watch.

The business is performing well and the price is genuinely more reasonable than it looks on the headline multiple. Fiscal 2026 revenue of $7,206 million grew 17.5%, the fastest in six years, at a 91.14% gross margin; the first quarter of fiscal 2027 grew 18.4% with operating income up 60.1% and the operating margin at 28.0%; Architecture, Engineering and Construction is now half the company and growing 22.0%; the balance sheet held net cash; both segment blocks tie to revenue to the dollar; and a director bought stock on the open market in June at a price 20% below today's. At 18.8x the FY2027 adjusted consensus and 16.7x FY2028's, with the stock 27.3% below its high, our base of $270 is 13.8% above spot.

Four reasons this is a Watch rather than a Buy.

First, the asymmetry is 1.77:1, below the 2:1 bar — and it is constructed against the multiple band the market has actually paid this year (13.2x to 23.0x on the FY2028 estimate), which is the honest construction rather than the flattering one.

Second, the base case is two and a third multiple turns. At an unchanged 16.7x the fair value is $237 — the current price. The 13.8% is entirely a re-rating assumption.

Third, stock-based compensation of $788 million — 10.9% of revenue, the highest in this batch — makes every attractive figure on this page have a less attractive twin. A 4.81% free cash flow yield is 3.24%. An 18.8x forward multiple is roughly 24x. A 33.4% free cash flow margin that passes the knowledge base's test is a 22.5% margin that fails it.

Fourth, two things are unquantifiable from this archive and both matter. The MaintainX acquisition closed 2026-08-03 — one day ago — with no consideration disclosed anywhere, part-funded by a $2.0 billion commercial paper facility. And the 28.0% first-quarter operating margin compares against a quarter in which a restructuring plan was initiated, with no restructuring line item in the payload to separate them.

Pre-registered KILL criteria — what would take this to Avoid:

• Operating margin back below 24% in any quarter, beneath the fiscal 2026 full-year level.
• MaintainX consideration above roughly $2 billion with no disclosed revenue contribution, or a goodwill allocation exceeding 90% of the price.
• A THIRD restructuring plan, after the "2026 Plan" and the "January 2026 Plan".
• Stock-based compensation above 12% of revenue, reversing a three-year decline from 13.1%.
• Revenue growth below 10% in any quarter, against 18.4% just delivered and 13.8% consensus.
• A break below the 50-day average of $215.65 on company-specific news.

Pre-registered UPGRADE conditions — what would take this to Buy — Tactical:

• A price below roughly $215, at which the base case is +25.6% and the asymmetry moves through 2:1 with no change to any number here. That is 9.4% below spot and essentially at the 50-day average. This is the cleanest path and it requires nothing from the company.
• A second consecutive quarter of operating margin above 27%, which would establish the first-quarter figure as a base rather than a comparison artefact.
• MaintainX disclosed at a price implying a defensible multiple, with a named revenue contribution — the value is in the disclosure, because it converts the largest unknown in this dive into something gradeable.
• Stock-based compensation below 10% of revenue, which would close part of the gap between the adjusted and economic pictures.

Where ADSK sits in the Synthos Framework Portfolio. No position. On the watch list at the $215 trigger, for a first tranche of 1.5% in the technology sleeve and a 3% target on confirmation of the margin. Logged as a tracked Synthos call (Watch) as of 2026-08-04 at $237.19, with the fair-value anchors, kill criteria and upgrade conditions all gradeable. The specific item to grade this dive on is the stock-compensation restatement: presenting a 4.81% free cash flow yield without its 3.24% twin would have made this look like a Buy, and the two numbers are the same company.

Single biggest risk: stock-based compensation, and the fact that it makes every headline figure on this page ambiguous. $788 million a year is 10.9% of revenue — the highest ratio in this batch — and it is the entire distance between trailing GAAP earnings of $6.86 and trailing adjusted earnings of $11.13, between a 4.81% free cash flow yield and a 3.24% one, and between an 18.8x forward multiple and roughly 24x. It is a genuine, recurring, economic cost paid in dilution rather than cash, and the buyback of $1,402 million in fiscal 2026 exists in large part to offset it. Netting the two, Autodesk returned roughly $614 million of genuine surplus to shareholders last year, 1.23% of market capitalisation — not the 2.80% the buyback line suggests. Nothing here says Autodesk is a poor business; the 91.14% gross margin and 17.5% revenue growth say the opposite. What it says is that the cheap-looking multiple is a convention, and a research process that quotes the convention without its correction is not doing the work.

Provenance & disclosures

• Traceability: 1 name-level knowledge-base claim names Autodesk out of 52,021 distilled claims. Raw hits 1; surviving a case-sensitive entity re-run 1; discarded 0; homograph collisions 0; breadth 1; net conviction a single claim, tested and split. The claim — bill_gurley, 2024-02-08, NEUTRAL, conviction 55, principle, speaker aaron_levie with speaker_role: independent, entities ANSS/CDNS/SNPS/ADBE/ADSK — is "Durable compounders growing 10-20% at 30-40% free-cash-flow margins that dominate a vertical for decades (Adobe, Autodesk, Ansys, Cadence, Synopsys) are underappreciated great businesses." IT IS GRADEABLE ON TWO NUMERICAL CRITERIA AND IT PASSES BOTH ON THE CONVENTIONAL MEASURE: Autodesk grew revenue 17.5% in fiscal 2026 (inside the 10-20% band) and generated free cash flow of $2,409M on $7,206M of revenue, a 33.4% margin (inside the 30-40% band). It is the only claim in this batch to specify two numbers in advance and clear both. ON THE ECONOMIC MEASURE IT FAILS: deducting the $788M of stock-based compensation the conventional measure adds back gives a 22.5% margin, below the claimed band. Both readings are stated and neither is preferred, because the claim did not specify a basis and the difference is 10.9 points of revenue. The claim carries a NAMED speaker with speaker_role: independent and is investor commentary rather than management guidance, so it is weighted at full — but it is twenty-nine months old and names Autodesk as one of five examples. WHAT THE LANE DOES NOT CONTAIN: no view on the two restructuring plans, no view on the MaintainX acquisition, no view on stock compensation, and — most consequentially — NOT ONE CLAIM in 52,021 addressing whether generative design tooling expands or erodes the seat-based licence model of a design-software company. That is a genuine analytical void and is why all three horizon confidences are Low.
• Data as-of: fundamentals — annual statements for FY2021-FY2026 from the 10-K filed 2026-03-03 (fiscal year ended 2026-01-31); quarterly income statement through 2026-04-30 from inc_q, cross-checked against the 10-Q filed 2026-05-29 · estimates 2026-08-04 · prices 2026-08-04, quote timestamp 1785873561 = 2026-08-04T19:59:21Z ($237.19, +1.06%; 50-DMA $215.65; 200-DMA $253.14; RSI 67.72; MACD +7.09) · knowledge-base claims 2026-08-04. Autodesk's fiscal year ends 31 January; fiscal 2026 ended 2026-01-31 and fiscal 2027 began 2026-02-01. All figures come from the Synthos vendor data file for ADSK or from the SEC filings in the ADSK archive.
• Filing archive contents: 10-K filed 2026-03-03 (fiscal 2026 — the source of the TWO restructuring plans, the "2026 Plan" initiated in the quarter ended 2025-04-30 and the "January 2026 Plan" initiated in the quarter ended 2026-01-31 and described as "the final phase of our sales and marketing optimization program"); 10-Q filed 2025-11-26 (October 2025 quarter); 10-Q filed 2026-05-29 (April 2026 quarter); 8-K filed 2026-06-18 (Items 5.03 and 5.07 — a Certificate of Amendment providing for officer exculpation under Delaware law, effective 2026-06-17, and the certified voting results of the 2026 annual meeting, at which Andrew Anagnost drew 741,011 votes against and Stacy J. Smith 23,865,848); 8-K filed 2026-07-13 (Item 8.01 — the establishment of a $2.0 billion unsecured commercial paper programme with maturities not exceeding 365 days, "to partially finance the transactions"); 8-K filed 2026-08-03 (Item 8.01 — the COMPLETION of the acquisition of MaintainX Inc. under a merger agreement dated 2026-05-28 via Matterhorn Acquisition Corp., ONE DAY before this dive, with NO consideration disclosed).
• Where the filings and the arithmetic corrected the vendor (detailed in Section 7): the MaintainX acquisition, completed 2026-08-03, and the $2.0 billion commercial paper programme established 2026-07-13 to part-finance it, NEITHER of which appears in any field of the payload — and neither is sized in this archive, so no fair-value anchor embeds any contribution, dilution or leverage from them; the TWO restructuring plans named in the 10-K, for which the payload carries NO line item in either inc_a or inc_q, meaning the 729 basis point year-on-year operating-margin expansion in the first quarter of fiscal 2027 cannot be decomposed; investmentsInPropertyPlantAndEquipment of −$61,000,000 against capitalExpenditure of −$31,000,000 in fiscal 2024 — a 97% divergence in one year while the fields match exactly in the other three; the FY2031 estimate row, reporting netIncomeAvg, epsAvg, epsLow and epsHigh all as ZERO while carrying a $12,075M revenue estimate from 7 analysts and an EPS analyst count of 5; est.ebitdaAvg/ebitAvg, carrying a fixed-ratio signature of exactly 22.99% and 20.08% of revenue across seven years, with the implied EBITDA margin 5.9 points BELOW the reported 28.88% — a rare case of the fabrication understating; totalDebt of $2,734M, which includes $251M of lease obligations (a 9.2% inclusion, the mildest in this batch) against filed long-term debt of $2,483M; netDebt of $485M, which omits $348M of short-term investments, against NET CASH of approximately $114M on the corrected basis and a corrected enterprise value of approximately $49,967.5M; and interestExpense of ZERO in inc_a for both FY2026 and FY2025 on a company with $2,483M of debt, producing an interestCoverageRatioTTM of 998.5x that is not used. Where vendor and filing AGREED — recorded, because these were directed tests: seg_prod AND seg_geo BOTH PASSED, each summing to $7,206,000,000 against FY2026 revenue to the dollar, with Americas present at $3,178M (44.1%) — clearing both the over-counting and the home-market omission tests on both blocks, the best segment data in this batch; the capital-expenditure figure of $43M is credible for a software company at 0.6% of revenue and trips no fatal corruption signature, with capexToDepreciationTTM of 0.263 consistent with a direct 0.221; the implied share count of 211,146,304 matches inc_q's 212,000,000 weighted diluted to 0.4%; dividendYieldTTM and dividendPerShareTTM of zero are CORRECT, not a defect; tangibleBookValuePerShareTTM of MINUS $7.588 is CORRECT and is reported rather than rejected, because total equity is positive and negative tangible book is normal for an acquisitive software company; and km_ttm and ratios_ttm are present and complete.
• Basis note — and it is the central analytical point of this dive. All consensus estimates are on an ADJUSTED basis that excludes stock-based compensation of $788 million, 10.9% of fiscal 2026 revenue and the highest ratio in this batch. Trailing GAAP diluted EPS of approximately $6.86 (summing the four quarters in inc_q) sits against trailing adjusted EPS of approximately $11.13 (summing earn_cal) — a 62% gap — giving 34.6x and 21.3x respectively at $237.19. Free cash flow is recomputed as filed operating cash flow minus capital expenditure, giving $2,409M, a 4.81% yield and a 33.4% margin; deducting stock-based compensation gives $1,621M, a 3.24% yield and a 22.5% margin. BOTH figures are stated wherever the measure appears, and the forward multiples of 18.8x and 16.7x are accompanied by an SBC-included estimate of roughly 24.1x for FY2027. Note also that the $1,402M fiscal 2026 buyback net of $788M of stock compensation is roughly $614M of genuine surplus returned — 1.23% of market capitalisation rather than the 2.80% the buyback line implies.
• Estimate coverage: 19 analysts on FY2027 adjusted EPS with a range of $12.530 to $12.701 — a 1.4% spread, the second-tightest in this batch — and 20 on FY2028, the anchor for all three fair values, with a range of $13.584 to $14.777 (8.4%). 10 on FY2029; 4 on FY2030 and a ZEROED FY2031 row, both excluded.
• Peer note: the vendor peer set — Cadence, CoreWeave, Datadog, Fair Isaac, Fortinet, Infosys, Motorola Solutions, Strategy, Synopsys and Workday — contains three of the four other companies named alongside Autodesk in the single knowledge-base claim (Cadence, Synopsys; Ansys and Adobe are absent), which is unusually apt, alongside a bitcoin treasury company and an IT services firm that are not comparable. No earnings or estimate data for any peer is present in this file, so no peer-multiple comparison is drawn.
• Fair-value caveat: the $179 / $270 / $340 anchors are multiples of the FY2028 consensus ADJUSTED EPS distribution — 13.2x the low of $13.58387, 19.0x the mean of $14.2004, and 23.0x the high of $14.7765 — and the bear and bull multiples are precisely those the market paid at the 52-week low of $187.72 and high of $326.79 on that same estimate, so the range is anchored to observed behaviour rather than to judgement. Stated arithmetic, not a discounted cash flow. The base assumes MULTIPLE EXPANSION from 16.7x to 19.0x and this is named as the fragile leg: at an unchanged 16.7x the fair value is $237 — the current price — and the entire 13.8% base upside disappears. No MaintainX contribution, dilution or leverage is embedded in any anchor, because the transaction is not sized in this archive.
• Timing: first-quarter fiscal 2027 results were released 2026-05-28, beating the adjusted consensus by 5.3% ($2.99 against $2.84) — the fifth consecutive beat, with the five averaging 6.6%. The MaintainX acquisition COMPLETED 2026-08-03, ONE DAY before this dive. The next print is 2026-08-27, 23 days away, against a consensus of $3.12 and $2,011,168K, and it will be the first report to include MaintainX and the first to show the effect of the commercial paper programme. The only open-market insider transaction in the file is a PURCHASE of 2,000 shares at $189.20 on 2026-06-23 by director John T. Cahill, doubling his holding; the remaining seven records are annual director equity awards granted 2026-06-17 at zero price. 2026-08-04 carried no company-specific SEC filing; the most recent is the 8-K of 2026-08-03.
• Accessibility note: no information in this dive is conveyed by colour. All emphasis is carried by bold text, table structure and explicit labelling.
• Not investment advice. Independent research, educational and informational only, never personalised. No recommendation to buy, sell or hold any security is made to any person.
• Version: 2026-08-04-full.