Accenture ACN

Reference table

1. What the business is, and how it actually reports

Identity. Accenture plc is incorporated in Ireland, headquartered at 1 Grand Canal Square, Dublin, and listed on the New York Stock Exchange. Chair and Chief Executive is Julie Spellman Sweet. The workforce was approximately 799,000 as of 31 May 2026 per the FQ3'26 10-Q, against 779,000 at 31 August 2025 and 791,000 at 31 May 2025 — the payload's fullTimeEmployees of 799,000 matches the filing exactly. The fiscal year ends 31 August, so "FY2025" means the twelve months to 31 August 2025 and the most recent reported quarter, FQ3'26, ended 31 May 2026.

Scale and client base. The 10-K discloses approximately 9,000 clients, of which 195 of the top 200 have been clients for ten or more years and 305 are "Diamond" relationships, the largest tier. The United States represented 45% of consolidated revenues in each of fiscal 2025, 2024 and 2023, and no other country reached 10%. Ireland, the domicile, is about 1%.

Three ways the company cuts its revenue — and the vendor confuses them. This matters, so it is set out precisely.

1. By geographic market — this is the reportable-segment basis. 10-K Note 16 states: "Our three reportable segments are our geographic markets: Americas, EMEA (Europe, Middle East and Africa) and Asia Pacific." The chief operating decision makers are the Chief Executive and Chief Financial Officer, who evaluate the segments on revenue and operating income. In the first quarter of fiscal 2025 the Latin America market unit moved from Growth Markets to North America; North America became "the Americas" and Growth Markets became "Asia Pacific," with prior periods reclassified.

2. By industry group — Health & Public Service; Communications, Media & Technology; Financial Services; Products; Resources. Used for commentary, not for segment reporting.

3. By type of work — consulting and managed services. Note that "outsourcing" is a retired label; every current filing says managed services.

The vendor's seg_geo is stale by two fiscal years and uses the retired names. Its most recent row is FY2023, showing Europe $21.285B, North America $30.296B and Growth Markets $12.531B. The vendor reports Europe / North America / Growth Markets through FY2023; the 10-K shows Americas / EMEA / Asia Pacific from FY2025 with prior periods restated; we use the 10-K basis. The FY2025 dollar split by geographic market cannot be reproduced here because the segment tables did not survive the filing text extraction (Section 8, item 10) — that is stated as a gap rather than estimated around.

The vendor's seg_prod mixes two different disaggregations across years. FY2025, FY2023, FY2022, FY2021 and FY2020 are type of work; FY2024, FY2019, FY2018 and FY2017 are industry group. Each year sums correctly to that year's revenue — FY2025's $35.107B consulting plus $34.566B managed services equals $69.673B exactly, and FY2024's five industry groups sum to $64.896B exactly — so there is no double-count. But it is not a time series and must not be read as one.

FY2025 type-of-work split, from the vendor and reconciled to reported revenue:

The direction of travel is clear even from this partial series: managed services has grown from 44.7% of revenue in FY2022 to 49.6% in FY2025. That mix shift is central to the whole argument. Managed services contracts are multi-year, largely non-cancellable in part, and convert to revenue slowly; consulting contracts are often under twelve months and terminable on 30 days' notice, per the 10-K. A company shifting toward managed services is a company whose revenue is becoming more predictable and less discretionary — and, per FQ3'26, growing faster (5% local currency against consulting's 1%).

2. The most important thing to understand about these numbers — bookings versus revenue, and the currency wedge between them

Accenture's reported dollar revenue growth is the least informative number the company publishes, and the number that matters most is going the other way.

Three facts have to be held together.

First, roughly 2.5 points of fiscal-2026 reported growth is the dollar, not the business. The FQ3'26 10-Q states it plainly: "The U.S. dollar weakened against various currencies during the three and nine months ended May 31, 2026 ... resulting in favorable currency translation and U.S. dollar revenue growth that was approximately 2.5% and 2.7% higher, respectively, than our revenue growth in local currency." The company further estimates that full fiscal-2026 dollar growth will be about 2 points above local-currency growth. That is management framing about a future period and is half-weighted accordingly, but the 2.5 and 2.7 point figures for periods already reported are audited-period disclosures, not guidance.

Second, new bookings — the leading indicator — are declining. Accenture defines new bookings as new contracts including those acquired, plus renewals, extensions and changes to existing contracts, and warns explicitly that there are no third-party standards governing the calculation and that bookings are not comparable to revenue. With that caveat stated, this is the single most important disclosure the company makes, because managed-services bookings convert to revenue over years and therefore tell you what the revenue line looks like after the current backlog is worked off.

FY2024 bookings are derived from the 10-K statement that FY2025 bookings of $80.6B were down 1%, and are labelled as derived. FQ1'26 bookings are not available — the first-quarter 10-Q is not in the filing set, and that is stated as a gap.

Read the last column downward. 1.25x, then 1.16x, then 1.03x. A book-to-bill of 1.03x means Accenture signed barely more new work in the quarter than it burned. That is the bear case, sourced entirely from the company's own filings, and it is why this dive is Tactical rather than Core.

Third, the counterweight is equally documented. Remaining performance obligations — the non-cancellable portion of signed contracts, which the company is careful to say is only a fraction of what it counts as bookings — went $30B (31 Aug 2024) → $34B (31 Aug 2025) → $38B (31 May 2026). That is roughly 13% growth in FY2025 and a further 12% in nine months, and it is rising faster than revenue. Bookings falling while RPO rises is exactly what a mix shift from short cancellable consulting work toward long non-cancellable managed-services work looks like. It does not resolve the argument. It does mean the argument is not yet a rout.

Therefore: this dive treats local-currency bookings, not dollar revenue, as the primary trend indicator, and treats the 2026-09-24 print as the decisive scheduled observation.

3. Growth, margin and the revenue-per-head test

Revenue. $44.327B (FY2020) → $50.533B (FY2021) → $61.594B (FY2022) → $64.112B (FY2023, +4.1%) → $64.896B (FY2024, +1.2%) → $69.673B (FY2025, +7.4%). The FY2024 trough and the FY2025 recovery are both real, and the 10-K confirms FY2025 was 7% in both dollars and local currency, i.e. a clean year with, in the company's words, "minimal currency translation impact."

Fiscal 2026 by quarter, from the payload's quarterly income statements and cross-checked to the 10-Qs:

The gap between GAAP and adjusted in FQ4'25 and FQ1'26 is business-optimisation cost, discussed below. The 10-Q confirms the reconciliation exactly: business optimisation costs of $250M net of tax reduced nine-month diluted EPS by $0.40, giving adjusted nine-month EPS of $10.67 against GAAP $10.27. The FQ1'26 operating-margin row carries a vendor defect that is worth flagging where it occurs rather than only in Section 8: the vendor reports FQ1'26 operating income of $3.152B, which is 16.82% of revenue; the 10-Q shows nine-month GAAP operating margin of 15.4% and adjusted of 15.9%, which is only reconcilable if the $308M business-optimisation charge is excluded from that quarter — $3.152B less $308M is $2.844B, or 15.17%, and the nine-month GAAP margin then computes to 15.34%, matching the filing. We use the GAAP figure and note that the same quarter's net income and EPS in the payload ARE on a GAAP basis, so the vendor's own quarterly statement is internally inconsistent.

Margin. GAAP operating margin was 14.7% in FY2025 against 14.8% in FY2024; adjusted, it was 15.6% against 15.5% — so on a clean basis margin expanded 10 basis points. FQ3'26 operating margin was 17.0% against 16.8%, another 20 basis points. Gross margin is doing none of that work: FQ3'26 gross margin was 32.8% against 32.9%, and the 10-Q attributes the decline to "higher non-payroll costs, including higher subcontractor costs, largely offset by lower payroll costs." The expansion is coming from below the gross-margin line — sales and marketing fell from 9.9% to 9.7% of revenue in the quarter and from 10.1% to 9.8% over nine months. This is a company holding price flat and taking cost out of the selling machine, not a company expanding delivery economics.

Business optimisation charges have now run three consecutive years and are becoming difficult to call one-off:

• FY2024: $438M for actions initiated in fiscal 2023 and completed in fiscal 2024, primarily severance.
• FY2025: $615M, comprising $344M associated with a "refreshed talent strategy" and $271M of asset impairments primarily related to the divestiture of two acquisitions in the Americas "no longer aligned with our strategic priorities."
• FQ1'26: $308M, primarily employee severance, completing a six-month programme.

The $271M impairment is worth dwelling on: it is a live, small-scale instance of exactly the mechanism that would express a demand shock in the accounts — bought capability written down. Against $22.536B of goodwill, it is trivial. As a precedent, it is not.

The revenue-per-head test — the direct mechanical examination of the automation thesis.

Trailing-twelve-month revenue is the sum of the four most recent quarterly income statements in the payload. Headcount figures are the 10-K and 10-Q disclosures at each date.

Revenue per employee is rising, not falling, and it is rising while headcount grows. Strip the roughly 2.7 points of currency and the local-currency figure is still up roughly 3%. Utilisation is at 93%, the highest in the disclosed series. Attrition has fallen to 14% from 16%.

That is not what task-automation compression looks like — yet. It is the strongest single piece of evidence for the bull, it comes from filings rather than from anybody's opinion, and it is the number this dive will grade itself against. The honest caveat is that headcount rose 20,000 between 31 August 2025 and 31 May 2026 after falling 12,000 in the preceding quarter, so the workforce is being actively managed and the ratio is partly a management output rather than a market signal.

4. Balance sheet and the enterprise-value rebuild

As reported in the FQ3'26 10-Q, at 31 May 2026:

• Cash and cash equivalents $10.2B (against $11.5B at 31 August 2025).
• Senior unsecured notes, $5.0B aggregate principal, issued by Accenture Capital Inc. and fully guaranteed by Accenture plc, maturing 2027 through 2034. Estimated fair value $4.9B.
• Commercial paper $100M, weighted-average effective rate 3.8%.
• Finance leases: immaterial. The 10-K states directly: "As of August 31, 2025 and 2024, we had no material finance leases."
• Undrawn borrowing facilities totalling $8.1B (a five-year facility maturing April 2031 and a 364-day facility maturing April 2027).
• Available share-repurchase authorisation $3,244M at 31 May 2026, from $59.1B cumulatively authorised since August 2001.

The corrected net position: $10.2B − $5.0B − $0.1B = approximately +$5.1B of NET CASH.

The vendor's defect, stated precisely. The payload's FY2025 balance sheet reports totalDebt of $8.183B, built as long-term debt $5.034B plus capitalLeaseObligations of $3.034B plus short-term debt $114.5M. The vendor reports $8.183B of total debt; the 10-K states there were no material finance leases and that the "total outstanding debt" summary line labelled as leases "primarily include[s] finance lease liabilities"; the $3.034B is therefore the OPERATING lease liability on office real estate, which the 10-K describes as "substantially all of our lease obligation ... for office real estate" with a 6.5-year weighted-average remaining term. We use the 10-K. Corrected FY2025 borrowings are $5.149B, and corrected FY2025 net cash is $6.34B, not the vendor's $3.296B.

The enterprise-value rebuild.

The correction is 3.2% — small in percentage terms, and materially smaller than the classic version of this defect, but it is real, it is entirely attributable to counting office leases as borrowings, and it runs in the direction of making the company look more levered than it is.

Two further balance-sheet facts that no balance sheet in this file reflects.

First, and most important: on 2026-07-10 — 24 days before this dive and 22 days after the FQ3 10-Q was filed — Accenture Capital Inc. closed the sale of $4.997B of senior notes, comprising $300M of floating-rate notes due 2029, $1.0B of 4.750% notes due 2029, $1.5B of 5.000% notes due 2031, $1.1B of 5.300% notes due 2033 and $1.1B of 5.600% notes due 2036, with estimated net proceeds of approximately $4.979B. That roughly doubles gross borrowings to about $10.0B. No balance sheet in the payload or in any filing on hand reflects it. If the proceeds are still in cash, the net position is unchanged at roughly $5.1B net cash and enterprise value is unchanged; if they are deployed into repurchases or acquisitions, enterprise value rises toward the full $104B market capitalisation. This dive uses the 31 May 2026 position as the anchor and flags the July issuance explicitly rather than pretending the balance sheet is current.

Second, the composition of the equity is unusual and is a risk in itself. Goodwill was $22.536B at 31 August 2025 (against $21.120B in FY2024 and $15.573B in FY2023 — a $5.5B step-up in FY2024 on $6.554B of acquisitions), and intangibles a further $2.411B. That is 34.5% of total assets in goodwill alone, and it is why book value per share is $53.93 while tangible book value per share is only $8.48. Accenture's stated strategy is to buy capability rather than build it, and the 10-K confirms goodwill is tested at the reportable-segment level with no impairment in FY2025 or FY2024 because "each reportable segment's estimated fair value substantially exceeded its carrying value." That is the current position. It is also the channel through which a durable demand shock would first appear in the accounts.

Coverage and returns. Interest coverage 40.9x. Debt to equity 0.263 (0.165 corrected for the lease misclassification). Return on equity 25.0%, return on invested capital 16.9%, return on capital employed 23.0%. Current ratio 1.34.

5. Cash flow and capital returns — the whole shareholder-return story

Trailing figures are derived from km_ttm and cross-check three ways: operatingCashFlowPerShareTTM of $21.532 on 611.94M shares gives $13.18B; freeCashFlowYieldTTM of 12.056% on a $104.36B market capitalisation gives $12.58B; evToFreeCashFlowTTM of 8.153 on the vendor's EV gives $12.58B. The 10-K figure of $8.3B "cash returned to shareholders" in FY2025, split $3.7B dividends and $4.6B share purchases, matches the cash-flow statement exactly.

This is an extraordinarily cash-generative business and the numbers deserve to be stated flatly. Capital expenditure is $600M on $69.7B of revenue — 0.86%. Free cash flow converts at 95% of operating cash flow and 1.6x reported net income (incomeQualityTTM 1.675). Trailing free cash flow of about $12.58B against a $104.36B market capitalisation is a 12.06% yield.

Capital return, honestly computed.

• Dividend. The board declared a quarterly dividend of $1.63 per Class A ordinary share on 17 June 2026, payable 14 August 2026 — an annualised $6.52, a 3.82% yield at $170.54. Dividends paid have grown $2.455B → $2.824B → $3.241B → $3.700B across FY2022-FY2025, roughly 14-15% a year. The payout is 47% of consensus FY2026 adjusted EPS and 29% of trailing free cash flow.
• Buyback. $4.619B in FY2025. On 2026-06-23 the company announced a $2B increase to the fiscal-2026 repurchase programme, bringing total expected repurchases to $7.5B — 7.2% of the market capitalisation. That is company guidance and is half-weighted and labelled as such; the enabling fact that is not guidance is that only $3,244M of authorisation remained at 31 May 2026, so a board increase was genuinely required, and the announcement came five days after the FQ3 print.
• Total shareholder yield. On FY2025 actuals, $8.319B on $104.36B is 7.97%. On the fiscal-2026 plan, $7.5B of buyback plus roughly $3.99B of dividends (611.94M shares at $6.52) is about $11.5B, or 11.0% — against trailing free cash flow of $12.58B, a 91% payout. That is the reason the July debt was raised, and it should be said plainly rather than implied.

The share-count arithmetic, and why it has quietly improved a great deal. Weighted-average diluted shares went 638.6M (FY2023) → 635.9M (FY2024) → 632.4M (FY2025) — a net reduction of only 0.55% a year despite $4.3-4.6B of annual gross repurchase, because $1.9-2.1B of annual stock compensation was reissuing most of what was bought, at a much higher share price. The quarterly series tells a different story: 626.0M (FQ1'26) → 622.6M (FQ2'26) → 615.6M (FQ3'26), a 1.7% reduction in two quarters, annualising above 3%. The 10-Q discloses the mechanism — 5,137,663 shares repurchased in FQ2'26 for $1,231M (an average of $239.60) and 5,754,179 in FQ3'26 for $1,147M (an average of $199.33). The same dollars now retire far more stock. That is the one unambiguous benefit of the de-rating, and at $170.54 it improves again. It is also worth saying without flinching that this year's repurchases have so far been executed at prices 17% to 41% above today's — the buyback has been a poor market-timer in fiscal 2026.

One scenario worth stating as arithmetic, labelled as such. If the $4.979B of July note proceeds were applied entirely to repurchase at roughly $170, that retires about 29.3M shares — 4.8% of the count, worth roughly $0.70 of annual EPS on a $14.68 FY2027 base — against incremental interest of about $250M a year at a blended ~5%, or roughly $0.31 a share after tax at the 24.3% effective rate. Debt-funded repurchase is accretive by roughly two to one at this price. The use of proceeds has not been disclosed and this is not a forecast.

6. Valuation — priced in or room?

At $170.54 (market cap $104.36B, corrected enterprise value $99.26B):

The estimate basis is established, not assumed. Consensus FY2026 EPS of $13.861 equals the 10-Q's adjusted nine-month figure of $10.67 plus the vendor's fourth-quarter estimate of $3.19 — exactly. The same test works on actuals: earn_cal reports FQ4'25 EPS of $3.03 against the income statement's GAAP $2.25, and FQ1'26 of $3.94 against GAAP $3.54, differences of $0.78 and $0.40 that reconcile precisely to the disclosed business-optimisation impacts. Therefore the estimates and the earn_cal actuals are on the adjusted, ex-business-optimisation basis, while the quarterly income statements are GAAP. Every forward multiple in this dive is on the adjusted basis and is labelled.

Two estimate rows are not used. FY2029E adjusted EPS of $18.44 on four analysts represents a 17.1% jump from FY2028's $15.749 on eleven, against 5-7% in every adjacent year; FY2030E of $18.92 is also on four. Neither is used, and the FY2027 row (18 analysts on EPS, 20 on revenue) is the primary forward anchor.

A low bar hiding in the FY2026 consensus. Nine-month revenue was $55.504B, so the $73.595B full-year consensus implies fourth-quarter revenue of about $18.09B against $17.596B a year earlier — just 2.8% in dollars, against 6.6% for the nine months. The independent earn_cal estimate of $18.055B agrees. Given the company's own estimate of roughly 2 points of full-year currency benefit, consensus is embedding local-currency growth of roughly 1% in the quarter that reports on 24 September. That is either stale, conservative, or a genuine forecast of sharp deceleration — and it is falsifiable in 51 days.

Peer context. The vendor peer set is mostly unusable: AMAT, KLAC, QCOM, TXN, APH and SONY are semiconductor and hardware names and PANW is security software, none of them comparable to a labour-based services firm. Only Infosys ($50.94B) and Leidos Holdings ($16.43B) are genuine comparables, and no peer multiples are supplied in the file, so no peer-multiple comparison is drawn. The genuinely informative cross-reference in this batch is Adobe, which closed at $257.64 with a market capitalisation of $102.37B against Accenture's $104.36B — the two are now near-identical in size, both rose modestly today (Adobe +2.51%, Accenture +2.88%) while the semiconductor complex rose 5-13%, and both are priced on the same one-line question about whether new capability destroys an incumbent's revenue model. Salesforce (+2.71%) makes it three. The market spent 2026-08-04 paying for the capital-equipment chain that builds the capability and grudgingly re-rating the incumbents it might displace.

The vendor's own quality model rates Accenture A- with an overall score of 4, with return on equity and return on assets both scored 5 and price-to-book scored 2.

6a. What today's price assumes (the inversion)

At $170.54 — 12.30x FY2026E and 11.62x FY2027E adjusted earnings, a 12.06% free-cash-flow yield and a 3.82% dividend — the price embeds roughly the following falsifiable claims.

• Revenue compounds only about 4-5% a year in dollars through FY2028, and roughly 2-3% in local currency. (Consensus-derived: $73.595B FY2026E → $76.606B FY2027E → $80.671B FY2028E, on 18-20 analysts; the local-currency translation uses the company's own estimate of about 2 points of currency benefit, half-weighted as management framing.) This is a mature-services growth rate, not a decline. The price is not discounting shrinkage.
• Adjusted EPS reaches about $15.75 by FY2028, i.e. +6.4% a year from FY2026's $13.861. (Consensus, 11 analysts on FY2028.) Since revenue grows 4-5%, that leaves roughly 1-2 points a year from margin and buyback combined — the price assumes no meaningful margin expansion, despite adjusted operating margin having risen 10 basis points in FY2025 and 20 in FQ3'26.
• The market never pays more than about 11-12x forward earnings again. (Derived: at the 52-week closing high of $288.54 the stock was 20.8x the current FY2026 estimate; it is now 12.30x.) The price assumes that compression is permanent rather than an overshoot.
• Adjusted operating margin holds near 15.5-16% and does not deflate. (Consensus-derived: FY2027 net income of $9.035B on revenue of $76.606B is an 11.8% net margin against 10.70% trailing.) If automation compresses billable hours faster than it reduces delivery cost, this is the line that breaks first.
• THE MOST FRAGILE ASSUMPTION: that the 3% local-currency bookings decline in FQ3'26 is a pause rather than the start of a trend. (Our reading of the filing series: book-to-bill 1.25x derived FY2024, 1.157x FY2025, 1.031x FQ3'26.) A genuine melting-ice-cube multiple for a business of this quality would be 8-9x FY2027E, which is $117-$132 — and the 52-week closing low of $124.44 is 8.5x. The market has already visited that price once in the last twelve months. The distance between today's $170.54 and that low is the distance between "mature" and "terminal," and it is 27%.

6b. The return bridge (why the multiple moves)

Expected return over 12-24 months decomposes as: adjusted EPS growth (+5.9% FY2026→FY2027, +7.3% FY2027→FY2028, consensus) + shareholder yield (3.82% dividend plus a net share-count reduction now running above 3% annualised) + multiple drift.

Start with the no-re-rating case, because it is the load-bearing one. If the forward multiple simply holds at today's 12.30x, then in twelve months the market is pricing FY2027's $14.679 and the price is $180.6, up 5.9%; add the dividend and the total is roughly 9.7%. Over two years, 12.30x FY2028's $15.749 is $193.7, up 13.6%, plus two years of dividend, for roughly 21%. You do not need the crowd to change its mind to earn high single digits here. That is the whole reason this is a Buy rather than a Watch.

Our base of $190 implies 12.9x FY2027E against 11.62x today — modest multiple EXPANSION of about 11%, and we flag it as the fragile leg. The justification is not growth re-acceleration; it is risk normalisation. A multiple moves from 11.6x back toward 13x when the market stops treating a structural question as an imminent one — concretely, when local-currency bookings stop declining. That is a single observable event, scheduled for 24 September and again in December. If it does not happen, the no-re-rating path still delivers roughly 10% a year, and that is the floor under the Tactical rating.

The bull case at $235 requires 16x FY2027E, which is a genuine re-rating toward the low end of the multiple range this business used to command, and it needs bookings growing again in local currency. That leg is fragile and is not required for the base case to work. The bear case at $132 is 9.0x FY2027E — the melting-ice-cube multiple, roughly where the stock actually traded at its 52-week low.

6c. Variant perception (where we differ, what would surprise)

• We differ from the street's target, downward. Consensus is $198.56, median $192, with a high of $282 and a low of $130 — a 2.2x spread, the widest disagreement in this batch and, in itself, the most honest single statistic about this name. Our base of $190 sits essentially at the median and 4.3% below the consensus. We think the consensus average is being pulled up by targets set before the bookings series rolled over, and we are not willing to underwrite multiple expansion beyond 13x while book-to-bill is 1.03x. Watchable number: the direction of target revisions after 2026-09-24.
• We differ from our own knowledge base's bullish lane, and say so. The only dated, testable bullish claim naming Accenture is from 2025-01-31 — "IT consulting firms are at record highs as enterprises turn to them to deploy AI" — made within weeks of the all-time high, and the stock has since fallen roughly 40%. We do not use it as support and we count it as a graded miss (Section 7).
• We differ from the bears on the mechanism, and this is the sharpest divergence. The bear case requires revenue per head to fall as automation displaces billable work. It has risen 5.7% over twelve months, to about $91,489 from about $86,577, while headcount grew and utilisation rose to 93% and attrition fell to 14%. Every one of those figures is from a 10-Q or 10-K, not from a presentation. If automation were already compressing the pyramid, the bench would be building and utilisation would be falling. It is not. Watchable numbers: revenue per employee and utilisation at each print — and note the honest caveat that headcount is actively managed, so the ratio is partly a management output.
• We differ on what the research-and-development line implies. The 10-K discloses R&D of $0.8B (FY2025), $1.2B (FY2024) and $1.3B (FY2023) — a 38% cut over two years at a firm whose thesis is technology leadership. The bull reading is that Accenture buys rather than builds ($6.554B of acquisitions in FY2024) and that its stated $3B multi-year investment in generative capability sits partly outside the R&D line. The bear reading is that a services firm cutting research while claiming leadership is telling you where its confidence actually is. We hold both and flag that the payload reports R&D as zero in every year before FY2025, which is simply wrong (Section 8).
• Positive surprise that would force a re-rate: local-currency bookings returning to growth on 2026-09-24, a fiscal-2027 revenue guide above 5% local currency, or disclosure that the $4.979B of July note proceeds went into repurchase near current prices. Negative surprise that would break the thesis: a second and third consecutive quarter of local-currency bookings decline, a FY2027 guide of 0-2% local currency, further deterioration in the U.S. federal business, or any goodwill impairment against the $22.536B balance.
• Where we have no edge, stated plainly: we cannot resolve the terminal question from this data, and neither can anybody else. The dive is sized and staged on that basis.

Synthos fair values

All three anchors are multiples of consensus FY2027 adjusted EPS of $14.679 (18 analysts), cross-checked against free-cash-flow yield and against the observed 52-week range.

• Bear ~$132 — 9.0x FY2027E. Cross-check: a 15.6% free-cash-flow yield on flat free cash flow, and 8.5x is where the stock actually traded at its 52-week closing low of $124.44. The scenario: local-currency bookings decline through fiscal 2027, book-to-bill settles below 1.0x, revenue goes flat in local currency, the federal drag widens and the market prices structural decline. −22.6%.
• Base ~$190 — 12.9x FY2027E, 13.7x FY2026E. Cross-checks: a 10.8% free-cash-flow yield on flat free cash flow, a 3.43% dividend yield at the current rate, and a corrected EV of $111.2B at 8.8x trailing EBITDA. Reachable with only 11% of multiple expansion, and the no-re-rating path independently delivers roughly $181 in twelve months and $194 in twenty-four. The scenario: bookings stabilise, revenue compounds 4-5% in dollars, margin holds, and the buyback runs at the announced pace. +11.4%, plus a 3.82% dividend.
• Bull ~$235 — 16.0x FY2027E. Cross-check: an 8.7% free-cash-flow yield and a 2.77% dividend yield, both still undemanding. The scenario: local-currency bookings inflect positive, the enterprise deployment cycle proves to be a multi-year revenue expansion rather than a substitution, and the market re-rates a 25%-return-on-equity, net-cash compounder back toward the low end of its historical range. +37.8%.

Base is 11.4% above spot; asymmetry roughly 1.7:1 to the upside (22.6% down, 37.8% up), with the downside case anchored to a price the market has already visited. That, plus a 3.82% dividend and a 7.2%-of-market-cap repurchase, is a Buy — but the staging is not optional.

7. Knowledge base — seven claims, thin, split, and one graded miss

Tracked knowledge-base claims naming Accenture: 7 raw hits. Used: 5. Discarded or downweighted: 2. Search terms were ACN and Accenture, run across all 51,928 distilled claims. Six sources.

Triage, stated before the content.

• Six hits are match: entity (the claim names the company in its entity list) and one is match: text (Accenture appears in prose about something else).
• Two of the six entity hits are the same document — compound_and_friends-CQCA0iLGOxY, dated 2025-01-31, appearing once with horizon "fact" and once with horizon "thesis" and near-identical text. Counting these as two claims overstates breadth. They are treated as one.
• Two entity hits name Accenture only as an example inside macro claims about employment (Raoul Pal, 2025-11-06; and the text match, Jordi Visser, 2025-10-05). They are supporting colour, not claims about the equity.
• Net: three claims genuinely about Accenture as a business or an investment, one about the client-side risk of outsourcing, and three about the labour market. For a $104B company that is a very thin lane, and thin is the finding.

The bullish side.

> 2025-01-31 · bullish · conviction 65 · match: entity · source compound_and_friends · entities Accenture, Cognizant, Gartner

> "IT consulting firms are at record highs as enterprises turn to them to deploy AI."

> 2026-04-28 · bullish · conviction 58 · match: entity · speaker Aaron Levie · channel a16z · entity ACN

> "Enterprises need massive change-management and systems integration for agents to work, making integrators like Accenture and Deloitte a multi-decade opportunity for existing and next-gen firms."

The bearish side.

> 2026-06-29 · bearish · conviction 55 · match: entity · speaker Avi Felman · attribution guest · channel anthony_pompliano_show · entity ACN

> "AI is disrupting consultants—Accenture went down 30% in a day; maybe short the consultant and buy biotech"

> 2023-01-26 · bearish · conviction 65 · match: entity · source odd_lots · entities Deloitte, Accenture

> "Outsourcing critical software to consultancies/vendors leaves firms without in-house experts to recover fast when outages hit; option value evaporates."

> 2025-11-06 · bearish · conviction 75 · match: entity · source raoul_pal · entities Amazon, Walmart, Accenture

> "Big employers freezing/cutting headcount to protect margins; Amazon has same employee count as 2022 despite GDP growth—connection between GDP and hiring broken."

The neutral text match.

> 2025-10-05 · neutral · conviction 74 · match: text · source jordi_visser · speaker_role independent · skill weight 2.0

> "AI is displacing tasks and reshaping every job; top private employers (Walmart, Amazon, Accenture, Google) are freezing/cutting hiring, and not adopting AI means falling behind at an unprecedented pace."

What this lane establishes, and what it fails to establish.

First — and this is the most useful thing in the section — the lane contains a graded miss, and we grade it. The 2025-01-31 claim that consulting firms "are at record highs as enterprises turn to them to deploy AI" was a fact claim about price at the moment of maximum optimism. Accenture's 52-week closing high is $288.54 and it closed today at $170.54. The observation was accurate as description and catastrophic as inference, and the difference between those two things is the entire discipline. It is not used as support anywhere in this dive.

Second, attribution discipline changes how two of these read. The 2026-04-28 bullish claim is attributed to speaker Aaron Levie, with a16z as the channel; Levie runs a listed enterprise-software company with a direct commercial interest in the enterprise-deployment narrative, which places him in the promoter class for this specific claim, and it is downweighted accordingly. Conversely, the 2026-06-29 bearish claim is attributed to guest Avi Felman, not to the channel host — under the house weighting rule the host's own voice is low-weight while guests are first-class, so this one is not discounted for channel. Its content is a trading quip at conviction 55, and its factual assertion that "Accenture went down 30% in a day" cannot be verified from our data, which contains no daily price series; what the payload does support is a 58.95% maximum drawdown from the six-year peak and repurchases executed at an average of $199.33 during the quarter ending 31 May against $170.54 today, both consistent with a sharp post-print decline. The magnitude is the speaker's, not ours.

Third, two of the claims are partly contradicted by the filings, which is exactly what a knowledge base is for. Both the Visser and Pal claims assert that Accenture is among the large employers freezing or cutting headcount. The filings show the opposite over the relevant window: the workforce grew from approximately 779,000 at 31 August 2025 to approximately 799,000 at 31 May 2026, an increase of 20,000, after falling 12,000 in the preceding quarter. The claims were directionally right about fiscal 2025 and wrong about fiscal 2026 to date. Visser carries the only above-baseline skill weight in the lane (2.0) and his claim is the more carefully framed of the two — it is about tasks and hiring in general, with Accenture as one example among four — so it is read as colour on the labour market rather than as a claim about this equity.

Why Low rather than Moderate. Six sources but seven claims; one duplicate; one text-only match; two macro claims where the company is an example; the single most specific bullish call badly wrong; nothing at all addressing the valuation at $170.54; and no claim from a source with a demonstrated single-name research process on this company. A thin, split, partly-discredited lane is a real finding about how little genuine independent conviction exists on this name, and it is reported as one rather than inflated. kb_net_conviction is mixed-negative and it is a mild negative input to the verdict, not a supporting one.

8. Data integrity — what we rejected from the vendor file and why

Fifteen findings. The share-class check demanded by the batch brief is item 1, and it passes.

1. Share count and dual-class structure — CHECKED AGAINST THE FILING COVER PAGE, VENDOR IS CORRECT. The FQ3'26 10-Q cover states: "The number of shares of the registrant's Class A ordinary shares, par value $0.0000225 per share, outstanding as of June 4, 2026 was 667,510,210 (which number includes 55,568,101 issued shares held by the registrant). The number of shares of the registrant's Class X ordinary shares ... was 300,673." Net of treasury, that is 611,942,109 Class A shares. At $170.54 that gives $104,360,607,268.86, against the vendor's marketCap of $104,360,607,269 — a match to the cent. The vendor did not make the share-class error; it correctly used Class A net of the 55.6M treasury shares rather than the gross 667.5M figure, which would have produced a spurious $113.8B. Two refinements are recorded rather than corrected: the 300,673 Class X shares carry a par value of $0.0000225 and no economic entitlement, so their exclusion is right; and the diluted EPS count of 615.6M in FQ3'26 additionally reflects Accenture Canada Holdings Inc. exchangeable shares convertible one-for-one, so a fully diluted market capitalisation would be about $105.0B, 0.6% higher. We use 611.94M and $104.36B.

2. totalDebt folds $3.034B of OPERATING lease liabilities into debt — REJECTED, and the EV rebuilt. The vendor reports FY2025 totalDebt of $8.183B, comprising long-term debt $5.034B, short-term debt $114.5M and capitalLeaseObligations $3.034B; the 10-K states "As of August 31, 2025 and 2024, we had no material finance leases" and describes the lease obligation as "substantially all ... for office real estate" with a 6.5-year weighted-average term; we use the 10-K. Corrected FY2025 borrowings $5.149B; corrected FY2025 net cash $6.34B against the vendor's $3.296B.

3. netDebt and enterpriseValueTTM understate net cash by $3.32B — corrected. The vendor's EV of $102.584B against a market capitalisation of $104.361B implies net cash of $1.777B. Using the FQ3'26 10-Q's disclosed $10.2B of cash and cash equivalents against $5.0B of senior notes and $100M of commercial paper, net cash is approximately $5.1B and corrected enterprise value is $99.26B. Corrected multiples: EV/EBITDA 7.83x (vendor 8.094x), EV/free cash flow 7.89x (vendor 8.153x), EV/revenue 1.358x (vendor 1.403x).

4. The payload's newest balance sheet is 11 months old, and a $5.0B debt issuance post-dates every balance sheet available anywhere in this file set. bal_a ends at 2025-08-31. The 8-K of 2026-07-10 discloses the closing of $4.997B of senior notes (a $300M floating tranche plus 4.750% 2029s, 5.000% 2031s, 5.300% 2033s and 5.600% 2036s) with net proceeds of approximately $4.979B, roughly doubling gross borrowings to about $10.0B. Neither the payload nor the 10-Q reflects it. Use of proceeds is not disclosed. Stated, not modelled.

5. seg_geo is stale by two fiscal years and uses retired segment names — REJECTED. The vendor's most recent geographic row is FY2023, split Europe / North America / Growth Markets; 10-K Note 16 states the three reportable segments are the geographic markets Americas, EMEA and Asia Pacific, following the Q1 FY2025 reclassification of the Latin America market unit, with prior periods restated; we use the 10-K basis. The FY2025 dollar split is not reproducible here (see item 10) and is stated as a gap.

6. seg_prod interleaves two different disaggregations and is not a time series — used only within years. FY2025, FY2023, FY2022, FY2021, FY2020 and FY2016 and earlier are type of work; FY2024, FY2019, FY2018 and FY2017 are industry group. Each year reconciles exactly to reported revenue (FY2025 $35.107B + $34.566B = $69.673B; FY2024 $19.554B + $13.841B + $11.610B + $10.837B + $9.054B = $64.896B), so there is no double-count — the defect is comparability, not arithmetic. The label "Outsourcing" is retired; current filings say managed services. And neither disaggregation is the reportable-segment basis, which is geographic.

7. inc_a reports research and development as $0 in every year before FY2025 — REJECTED as a series. The vendor shows R&D of $817.3M in FY2025 and zero in FY2024, FY2023, FY2022, FY2021 and FY2020. The 10-K states R&D was "$0.8 billion, $1.2 billion and $1.3 billion in fiscal 2025, 2024 and 2023, respectively"; we use the 10-K. The correction is analytically material — the true trend is a 38% reduction over two years, which the vendor's zeros conceal entirely, and it is discussed in Section 6c.

8. earn_cal EPS actuals are NON-GAAP while inc_q EPS is GAAP — the two must not be compared, and the estimate basis follows the former. earn_cal reports $3.03 for FQ4'25 against the income statement's $2.25, and $3.94 for FQ1'26 against $3.54. The differences of $0.78 and $0.40 reconcile exactly to the disclosed business-optimisation impacts ($615M in FQ4'25; $250M net of tax, stated in the 10-Q as $0.40 per share, in FQ1'26). Consensus FY2026 EPS of $13.861 equals adjusted nine-month EPS of $10.67 plus the $3.19 fourth-quarter estimate. All forward multiples in this dive are therefore on the adjusted basis, stated explicitly at each use.

9. est FY2029 and FY2030 rows rest on four analysts and contain a discontinuity — NOT USED. FY2028E $15.749 (11 analysts) to FY2029E $18.44 (4 analysts) is a 17.1% step against 5-7% in adjacent years. FY2027 is the forward anchor throughout.

10. The filing text extraction dropped every financial-statement and segment TABLE — a source limitation, not a vendor defect, but it constrains this dive. Balance sheets, income statements, cash-flow statements, the bookings tables, the segment tables and the geographic revenue tables are absent from the extracted text; only narrative prose and XBRL-tagged inline figures survived. Consequently: the FQ3'26 balance sheet is known only through the narrative disclosures quoted above ("Cash and cash equivalents was $10.2 billion"; "$5 billion aggregate principal"); the FY2025 revenue split by geographic market is unavailable; and quarterly bookings by consulting versus managed services — which the brief specifically asked for — could not be recovered. The consulting-versus-managed-services split is therefore reported only in the qualitative form the MD&A provides (FQ3'26 consulting +1% and managed services +5% in local currency), which is stated as a partial answer rather than presented as the disclosure itself.

11. est.ebitdaAvg and est.ebitAvg are inconsistent with actuals and are not used. FY2025E EBITDA of $12.107B against a reported operating income of $10.226B plus depreciation and amortisation of $1.705B — $11.931B — is close enough to be plausible, but FY2024E EBITDA of $11.308B sits below FY2023E of $12.160B despite higher revenue and income, which is not. Per the data contract these rows are corrupt in roughly 70% of files; epsAvg is used throughout and no EBITDA-based forward multiple appears in this dive. The single trailing EBITDA figure used in the EV rebuild (~$12.674B) is derived from ratios_ttm.ebitdaMarginTTM of 17.337% applied to trailing revenue and is labelled where used.

12. priceToBookRatioTTM (3.274) is internally inconsistent with bookValuePerShareTTM ($53.925). $170.54 ÷ $53.925 = 3.163x; the reported 3.274 implies $52.09, which is shareholdersEquityPerShareTTM. We use 3.16x and note the discrepancy. It is immaterial to any conclusion.

13. quote.yearHigh/yearLow ($291.09 / $118.15) disagree with tech.hi52/lo52 ($288.54 / $124.44) — both reported, closing basis used. The tech block is computed from six years of closing prices and the quote block carries intraday extremes, which fully explains the direction of both differences. This dive uses the closing basis ($288.54 / $124.44) throughout and states the intraday figures here. On the intraday basis the drawdown from high is 41.4% rather than 40.9% and the rally from low is 44.3% rather than 37.0%.

14. quote.timestamp is 1785873536 = 2026-08-04T19:58:56Z, i.e. 15:58:56 Eastern — two minutes BEFORE the closing bell. The price of $170.54 is treated as the close and any final two-minute move is not captured; tech.last of $170.51 is a third of a cent per dollar different. The reported volume of 3,914,114 is likewise a near-close figure, and no average-volume field is supplied, so no volume comparison is drawn.

15. inc_q operating income for FQ1'26 is on the ADJUSTED basis while the same quarter's net income and EPS are GAAP — internally inconsistent, corrected. The vendor reports FQ1'26 operating income of $3.152B (16.82% of revenue); the 10-Q states nine-month GAAP operating margin of 15.4% and adjusted of 15.9%. Subtracting the disclosed $308M business-optimisation charge gives $2.844B, a 15.17% GAAP margin, and a nine-month GAAP margin of 15.34% and adjusted of 15.89% — both matching the filing to a rounding place. We use the GAAP figure. The same defect does not appear in FQ4'25, where the quarterly operating income of $2.050B correctly includes the $615M charge and the four FY2025 quarters sum exactly to the annual $10.226B. This one quarter is therefore an isolated mis-mapping, not a systematic basis choice, which is precisely why it is dangerous — a reader computing sequential operating-margin trends from the payload alone would see FQ1'26 as the strongest quarter of the year when on a like-for-like basis it was the third strongest.

Also rejected as uninterpretable rather than wrong: priceToEarningsGrowthRatioTTM (−28.56) and priceToEarningsDilutedGrowthRatioTTM (−42.77), which are negative and therefore carry no information; grahamNetNetTTM (−$22.20) and netCurrentAssetValueTTM (−$6.854B), which are meaningless for an asset-light services firm; and daysOfInventoryOutstandingTTM / inventoryTurnoverTTM (both zero), correctly so for a business with no inventory.

Correctly reported, and confirmed against the filings: profile.fullTimeEmployees of 799,000 matches the 10-Q's "approximately 799,000 as of May 31, 2026" exactly; marketCap matches the cover-page share count net of treasury (item 1); seg_prod sums to reported revenue in every year checked; bal_a short-term investments of $5.9M are genuinely trivial, so there is no omitted-short-term-investments defect of the kind that afflicts other names; the FY2025 cash-flow figures reconcile to the 10-K's "$8.3 billion returned to shareholders, including dividends of $3.7 billion and share purchases of $4.6 billion"; and priceToEarningsDilutedRatioTTM of 13.62x is arithmetically exact on trailing GAAP diluted EPS of $12.52.

Non-equity tripwire — checked and passed. ACN is common equity: Class A ordinary shares of Accenture plc, par value $0.0000225, listed on the NYSE, trading at $170.54 with a beta of 1.122, a 52-week range of $124.44 to $288.54 (a 2.3x band), a growing rather than fixed dividend ($1.63 per quarter, raised over time from a level that produced $2.455B of total payments in FY2022 to $3.700B in FY2025), and volume near 3.9M shares. This is not a preferred share, a baby bond or a tracking unit inheriting parent fundamentals.

9. Technicals — a wreck that has already bounced hard

• Price $170.54, −40.9% below the 52-week closing high of $288.54, +37.0% above the 52-week closing low of $124.44. Maximum drawdown from the six-year peak: −58.95%.
• Above the 50-day moving average of $153.95 by 10.8%; BELOW a falling 200-day moving average of $210.28 by 18.9%. The two averages are $56 apart, which is the visual definition of an unrepaired downtrend with a sharp counter-rally inside it.
• RSI14 74.76 — overbought. MACD +5.63 and positive.
• Relative performance: 12-month −33.2% against SPY +24.3% and QQQ +30.8% — deficits of 57 and 64 points, the worst in this batch by a distance. 6-month −36.1% against SPY +11.1%, a 47-point deficit. 3-month −5.3% against SPY +7.6% — a 13-point deficit. The compression of that gap from 57 points to 13 is the single most encouraging number on the chart.
• Sentiment does not corroborate the price action. The covering street shows 35 buy, 17 hold and 1 sell with a consensus target of $198.56 — a Buy consensus on a stock down 33%. That is unusual and it means the street has not capitulated. The dispersion is where the information is: a high of $282 against a low of $130, a 2.2x spread.

Today's move and what it does to the entry

Accenture did not gap up today. It gapped down and was bought, and the distinction matters.

The mechanics: previous close $165.76; open $160.64, a 3.1% gap down; day low $160.40; day high $171.09; close $170.54, +2.88% and within 55 cents of the high. That is a 6.7% intraday range resolved decisively upward, and it is a materially more constructive pattern than a name that opened at its high and drifted.

It is also a different trade from the rest of the batch. The semiconductor complex had a violent up-day — Marvell +12.81%, Intel +10.92%, Lam Research +7.85%, Micron +7.62%, AMD +7.00%, KLA +6.95%, Applied Materials +5.48%, ASML +4.22%. Accenture rose 2.88%, Adobe 2.51%, Salesforce 2.71%. The market spent the session paying enthusiastically for the equipment chain that builds new computational capability and only grudgingly re-rating the incumbent service and software businesses that capability might displace. Accenture is in the second cohort, not the first. Nobody is being asked to chase a one-day gap in this name.

So what is the honest entry problem? It is not today. It is the six weeks before today. The stock is 37.0% above its 52-week low and carries an RSI of 74.76. Buying a full position into a 74.8 RSI reading after a 37% rally is a poor expected-value decision regardless of how cheap the multiple is, and it is not made better by the fact that the multiple is 11.6x.

The practical instruction, stated plainly.

1. Take a third to a half of the intended position now. The valuation, the 12.06% free-cash-flow yield, the 3.82% dividend and the $7.5B repurchase are all real today, and there is a genuine risk that a bookings inflection on 24 September is met with a gap the other way.

2. Complete on either of two conditions: a retest of the $150-158 zone around the 50-day moving average, which would be a 7-12% improvement in entry, or confirmation on 2026-09-24 that local-currency bookings have stopped declining, which would justify paying up.

3. Do not average down below $140 without re-reading the kill criteria — that price is 9.5x FY2027E and would mean the market is re-pricing toward the structural case.

The verdict is Buy — Tactical rather than Buy — Core specifically because of this paragraph. The business quality and the valuation would support a Core rating; the RSI, the falling 200-day average, the shrinking order book and the 51-day gap to the decisive disclosure do not.

10. Insiders — one officer trimming, no buying at all

The reading, stated carefully. All eight transactions are in Class A ordinary shares, consistent with item 1 of Section 8. Three are awards, of which the Chief Executive's is 285 shares and the Chief Accounting Officer's 778 — routine, and carrying no signal in either direction.

The five sales are one officer, on one day, in a rising price ladder — $160.86, $162.29, $163.27, $164.44, $165.16 — which is the signature of a programmatic execution rather than a discretionary decision to exit. It took Joel Unruch's direct holding from 28,233 shares to 17,735, a 37% reduction, and realised roughly $1.71M at prices between $160.86 and $165.16, i.e. 3-6% below today's close.

The negative worth stating is what is absent. Across a twelve-month period in which the stock fell 33% and traded as low as $124.44, not one insider made an open-market purchase. The Chief Executive's direct holding is 16,520 shares, roughly $2.8M at today's price, on a $104B company — that number excludes unvested awards and any indirect holdings not captured in this file, so it should not be over-read, but it is not the picture of a management team buying its own de-rating. Compare Markel, also covered in this batch, where every insider transaction was an acquisition and directors accumulated into the drawdown. Accenture's insiders did the opposite of nothing: they sold a little and bought none. A modest negative, and it is scored as one.

11. Verdict, kill criteria and flip conditions

Buy — Tactical.

The case in one paragraph. Accenture generates about $12.58B of free cash flow a year on $600M of capital expenditure, earns a 25.0% return on equity, holds roughly $5.1B of net cash once the vendor's misclassified office leases are removed, and is priced at 11.62x consensus FY2027 adjusted earnings with a 12.06% free-cash-flow yield and a 3.82% dividend. It has committed to repurchasing $7.5B of stock in fiscal 2026 — 7.2% of the market value — and the collapse in the share price has roughly quadrupled the rate at which those dollars retire shares, with the diluted count down 1.7% in two quarters against 0.55% for the whole of FY2025. At a constant forward multiple that combination returns roughly 10% a year with no change of view required from anybody. The lagging operational measures are all pointing the right way: revenue per employee up 5.7%, utilisation at 93%, attrition down to 14%, remaining performance obligations up from $34B to $38B in nine months, adjusted operating margin up 20 basis points in the most recent quarter.

What you are accepting, stated without softening. New bookings — the leading indicator, and the disclosure Accenture itself describes as the best available trend information on new business — fell 1% in fiscal 2025 and 3% in local currency in the quarter ended 31 May 2026. Book-to-bill has gone from a derived 1.25x to 1.16x to 1.03x. Business optimisation charges have run three consecutive years. Research and development spend has fallen 38% in two years. Goodwill of $22.536B is 34.5% of total assets and tangible book is $8.48 a share. The U.S. federal business is under disclosed pressure. And the company's own 10-K risk factors state that it faces competition "including from our clients and ecosystem partners, who may develop their own internal AI-related capabilities, as well as new AI-native companies," and that "some services and tasks currently performed by our people have been and will continue to be replaced by automation." Our knowledge base is thin, split, and its only specific bullish call was made at the top and has been wrong by 40%.

And the entry is stretched. RSI 74.76 after a 37.0% rally off the low, 18.9% below a falling 200-day average, with the decisive disclosure 51 days away. Hence Tactical, hence staged: a third to a half now, the rest on a retest of the $150-158 zone or on evidence at the 24 September print.

Pre-registered KILL criteria — what would break the thesis:

• Local-currency new bookings declining again in FQ4'26 (2026-09-24) and in FQ1'27, taking trailing book-to-bill below 1.0x. This is the primary criterion and it dominates the others.
• Fiscal-2027 revenue guidance implying 2% or less local-currency growth, given the company's own estimate of roughly 2 points of currency benefit.
• Revenue per employee falling year-on-year in local currency for two consecutive quarters — the direct mechanical signature of automation compressing the pyramid.
• Utilisation falling below approximately 90% while headcount grows. That combination is bench-building and it precedes margin damage by about two quarters.
• Adjusted operating margin below 15%, against 15.9% for the nine months to May 2026 and 15.6% for FY2025.
• A goodwill impairment of any material size against the $22.536B balance. The FY2025 write-down of $271M on two divested acquisitions is the small-scale precedent.
• A fourth consecutive year of business optimisation charges appearing in FY2027, at which point they are structural cost, not restructuring.
• The $4.979B of July note proceeds funding a large transformational acquisition rather than repurchase or tuck-ins. That would change the risk profile of a company we are buying partly for its balance sheet.

Pre-registered FLIP TO CORE (upsize):

• Local-currency bookings growth above 5% for two consecutive quarters and trailing book-to-bill back above 1.15x.
• Fiscal-2027 guidance of 5% or better local-currency revenue growth on 24 September.
• Disclosure that the July note proceeds were deployed into repurchase below approximately $180, which would be management acting on the same valuation view this dive takes.
• Revenue per employee accelerating above 4% in local currency, which would mean the automation is accruing to Accenture rather than to its clients.
• A price below approximately $140 (about 9.5x FY2027E) in the absence of a fundamental break — at which point the position should be completed rather than reconsidered.

Where ACN fits in the Synthos Framework Portfolio. The cash-return / de-rated quality sleeve, as a 1.5-2% starter position built to 3% on the triggers above. The sizing logic is explicit: the business quality, the balance sheet and the yield justify a full-sized holding, but the conviction lane is Low, the leading indicator is negative, and the entry is technically stretched — three independent reasons to leave room. On overlap within this batch: Accenture, Adobe and Salesforce are three expressions of the same wager, that the market has over-discounted incumbents whose revenue models new capability might displace. They are not interchangeable and should not all be held. Accenture is the cheapest of the three on forward earnings, the only one paying a substantial dividend (3.82%), the only one whose cost base is people rather than code — which cuts both ways, since people can be reduced but code cannot be un-commoditised — and the one whose leading indicator is most clearly published and therefore most cleanly falsifiable. If only one of the three is held, the case for it being this one rests on the yield and on the fact that its bear case is measurable quarter by quarter. Logged as a tracked Synthos call (Buy — Tactical) as of 2026-08-04 at $170.54.

Single biggest risk: a sustained decline in local-currency new bookings. Everything else in the bear case is a symptom; this is the mechanism. Accenture's revenue in any quarter is the conversion of business won in prior quarters, so a shrinking order book shows up in revenue with a lag of roughly four to eight quarters, and by the time it is visible in the revenue line the multiple will have moved first. Book-to-bill at 1.03x in FQ3'26, against a derived 1.25x two years earlier, is the warning. It is not yet a verdict — bookings are lumpy by the company's own admission, dependent on the timing of a small number of large managed-services signings, and remaining performance obligations rose 12% over the same nine months — but it is the number that decides this investment, and it prints on 24 September.

Most fragile assumption in the price: that Accenture captures the surplus from automating its own delivery rather than surrendering it to clients in price. The whole bull case reduces to one ratio — revenue per employee growing faster than cost per employee, in local currency, for years. It is currently doing so (about $91,489 against about $86,577 twelve months earlier). But the 10-K names the exact failure mode: "If we are unable to introduce or if our clients do not accept new pricing or commercial models that reflect the value of these AI-enabled solutions, our results of operations may be adversely affected." A firm that automates its own work and continues to bill by the hour has cut its own revenue. A firm that automates its own work and successfully re-prices to outcomes has expanded its margin. Which of those two Accenture becomes is not knowable from this file, is not knowable on 24 September, and is the reason the exponential score is 2 and the position is staged.

Provenance & disclosures

• Traceability: 7 tagged knowledge-base claims name Accenture (breadth 6 sources, claim count 7, net conviction mixed-negative). Search terms ACN and Accenture across 51,928 distilled claims. 7 raw hits; 5 used; 2 discarded or downweighted. Discards and downweights, with reasons: (i) compound_and_friends-CQCA0iLGOxY appears twice — same date, same document, near-identical text, one tagged horizon "fact" and one "thesis" — and is counted once, since counting it twice would overstate breadth; (ii) the jordi_visser hit of 2025-10-05 is match: text rather than match: entity (its entity list is empty and Accenture appears as one of four example employers in a labour-market claim), so it is treated as supporting colour and not as a claim about the equity. Downweighted rather than discarded: the a16z / Aaron Levie claim of 2026-04-28, because the speaker runs a listed enterprise-software company with a direct commercial interest in the enterprise-deployment narrative and is therefore promoter-class for this claim; and the raoul_pal claim of 2025-11-06, which names Accenture in an entity list but is substantively a macro claim about employment. Attribution follows the speaker where speaker and channel differ: Aaron Levie (channel a16z) and Avi Felman (channel anthony_pompliano_show, attribution "guest"). Under the standing house rule the Pompliano channel's own voice is low-weight while guests are first-class, so the Felman claim is not discounted for channel. Sources named in full: odd_lots, compound_and_friends, jordi_visser, raoul_pal, a16z, anthony_pompliano_show. All quotations are verbatim from stored claim text. Two claims (Visser, Pal) are partially contradicted by the filings, which show the workforce growing from approximately 779,000 to approximately 799,000 between 31 August 2025 and 31 May 2026.
• Data as-of: income statement 2026-05-31 (FQ3'26, reported 2026-06-18, 47 days before this dive) · balance sheet 2025-08-31 in the payload, supplemented by the FQ3'26 10-Q's narrative disclosures at 2026-05-31 ($10.2B cash; $5.0B senior notes; $100M commercial paper) · cash flow 2025-08-31 annual, with trailing figures derived from km_ttm · segments per the 10-K filed 2025-10-10 for the fiscal year ended 31 August 2025 · estimates 2026-08-04 · prices 2026-08-04, quote stamped 2026-08-04T19:58:56Z, i.e. two minutes before the 16:00 Eastern close · knowledge-base claims 2026-08-04. Accenture's fiscal year ends 31 August, so all "FY" references are August year-ends and FY2026 closes 26 days after this dive. No figure in this document comes from memory, recall or external retrieval.
• Data rejections (detailed in Section 8): totalDebt of $8.183B rejected for folding in $3.034B of operating lease liabilities that the 10-K states are not finance leases, corrected to $5.149B of borrowings and $6.34B of FY2025 net cash; netDebt/enterpriseValueTTM corrected from $102.584B to $99.26B using the 10-Q's $10.2B cash and $5.0B notes at 31 May 2026, with EV/EBITDA restated from 8.094x to 7.83x, EV/FCF from 8.153x to 7.89x and EV/revenue from 1.403x to 1.358x; seg_geo rejected as two years stale and using segment names retired in Q1 FY2025; seg_prod flagged as interleaving type-of-work and industry-group disaggregations across years (each year reconciles exactly to revenue, so no double-count, but it is not a time series) and the label "Outsourcing" corrected to managed services; inc_a research-and-development rejected as a series, the vendor's zeros for FY2020-FY2024 replaced by the 10-K's $1.2B (FY2024) and $1.3B (FY2023); earn_cal EPS actuals identified as non-GAAP against GAAP quarterly income statements, with the reconciliation verified to the disclosed business-optimisation impacts; est EBITDA and EBIT rows not used; est FY2029 and FY2030 rows not used (4 analysts, 17% discontinuity); inc_q FQ1'26 operating income identified as adjusted while the same quarter's net income and EPS are GAAP, corrected from $3.152B to $2.844B and from a 16.82% to a 15.17% operating margin; priceToBookRatioTTM (3.274) corrected to 3.16x; quote intraday 52-week extremes reported alongside the closing-basis tech figures actually used; the near-close quote timestamp disclosed; negative PEG ratios, Graham net-net and net-current-asset-value fields rejected as uninterpretable for this business. The share-class check demanded for this name PASSED: the vendor's $104.361B market capitalisation matches 611,942,109 Class A shares net of 55,568,101 treasury shares at $170.54, not the gross 667,510,210 on the 10-Q cover.
• Basis note — the most important one in this dive: analyst estimates and earn_cal actuals are on Accenture's adjusted (non-GAAP, ex-business-optimisation) basis; the quarterly income statements are GAAP. Consensus FY2026 EPS of $13.861 equals the 10-Q's adjusted nine-month $10.67 plus the $3.19 fourth-quarter estimate, exactly. Trailing adjusted diluted EPS is $13.70 against GAAP $12.52. All forward multiples in this dive are on the adjusted basis and are labelled at each use. Where a GAAP multiple is shown it is identified as such.
• Local-currency note: Accenture reports revenue and bookings growth both in U.S. dollars and in local currency, and the two diverge materially in fiscal 2026. The FQ3'26 10-Q discloses that favourable currency translation made dollar revenue growth approximately 2.5 points higher in the quarter and 2.7 points higher over nine months than local-currency growth. Every growth figure in this dive states which basis it is on. The company's estimate that full-year fiscal-2026 dollar growth will be about 2 points above local currency is management framing about a future period and is half-weighted and labelled.
• Management-framing note: the $7.5B fiscal-2026 repurchase figure comes from the 8-K of 2026-06-23 and is company guidance, half-weighted and labelled at each use. The verifiable facts around it are that available authorisation was $3,244M at 31 May 2026, that $59.1B has been authorised cumulatively since August 2001, and that repurchases in FQ2'26 and FQ3'26 ran $1,231M and $1,147M respectively at average prices of $239.60 and $199.33. The 10-K risk-factor quotations used in this dive are also company voice, but they run adverse to the company's interest and are treated as disclosure rather than as promotion.
• Filing-extraction limitation: the text extractions of the 10-K and both 10-Qs dropped every financial-statement and segment table, retaining only narrative prose and inline XBRL-tagged figures. Consequently the FQ3'26 balance sheet is available only through narrative disclosure, the FY2025 revenue split by geographic market could not be reproduced, and quarterly bookings split between consulting and managed services — specifically requested for this name — could not be recovered and are reported only in the qualitative form the MD&A supplies. Stated as a gap.
• Estimate coverage: solid by the standards of this batch — 16 analysts on FY2026 EPS, 18 on FY2027, 11 on FY2028, and 18-20 on the revenue lines. FY2029 (4 analysts) and FY2030 (4) are not used. The FY2027 row is the primary forward anchor for all three fair-value anchors.
• Peer note: of the ten names in the vendor peer set, only Infosys ($50.94B) and Leidos Holdings ($16.43B) are genuine comparables; AMAT, KLAC, QCOM, TXN, APH, SONY and PANW are semiconductor, hardware or security-software businesses. No peer multiples are supplied in the file, so no peer-multiple comparison is drawn. The Adobe cross-reference ($102.37B market capitalisation from the peers block, $257.64 close) uses only figures present in the payload and in the batch price table.
• Fair-value caveat: the $132 / $190 / $235 anchors are multiples of consensus FY2027 adjusted EPS of $14.679 (9.0x / 12.9x / 16.0x), cross-checked against free-cash-flow yield (15.4% / 11.2% / 9.0% on flat free cash flow), against dividend yield, and against the observed 52-week closing range. Stated arithmetic, not a discounted cash flow. The base case requires approximately 11% of multiple expansion and that leg is flagged as fragile; the no-re-rating path independently produces roughly $181 in twelve months and $194 in twenty-four at a constant 12.30x forward multiple, which is what actually supports the rating.
• Timing: FQ3'26 was reported 2026-06-18, 47 days before this dive, so the most recent quarter is fully in the data. The next print is 2026-09-24, 51 days away, and it is a fiscal-year-end report carrying initial FY2027 guidance — materially more consequential than an ordinary quarter. Two events post-date every balance sheet in this file and are disclosed rather than modelled: the 2026-06-23 announcement raising the fiscal-2026 repurchase programme to $7.5B, and the 2026-07-10 closing of $4.997B of senior notes.
• 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.