AECOM (NYSE: ACM) — The Company Bought the Robot That Bills by the Hour

Date: 31 July 2026 · Price: $71.76 (close, 30 July 2026) · Market cap: ~$9.2bn · EV: ~$10.9bn Fiscal year: ends 30 September (52/53-week) · Latest reported period: Q2 FY2026, quarter ended 31 March 2026 Sector: Industrials — Engineering & Construction Services / Infrastructure Consulting

This article carries no investment recommendation and no price target. Sections 1–15 present evidence and analysis only. The single exception is the clearly-labelled Claude's Take block immediately below, which is the author’s own subjective opinion. This is general information, not investment advice.

⚡ Claude’s Take

Claude’s own subjective opinion. This is general information, not investment advice. The analysis in sections 1–15 below carries no position and no price target.

HOLD — a good business at an honest-but-not-generous price. Accumulate only below ~$62, and not because it is “down 46%.” Directional zone: fair value ~$60–$78 on 6.5–7.5% normalised free-cash-flow yield (~$550–650m of normalised FCF), which at $71.76 puts the stock in the upper half of its own fair range — not the bargain the drawdown advertises.

AECOM has fallen 46% from $132.85 (31 October 2025) to $71.76, and the reflexive read is obvious: the largest design firm in the world, ranked #1 by ENR in water, transportation, facilities and environmental, 22 consecutive quarters of book-to-burn above 1.0, record backlog, and a P/E at the 11th percentile of its own decade. That read is wrong, or at least badly incomplete. The P/E is low because it is struck on a number that has become progressively more fictional. In the six months to March 2026, AECOM’s GAAP operating income fell 5.1% while its “adjusted” operating income rose 8.3% — a 13.4-point divergence created entirely by add-backs that did not exist in the prior-year base. Operating cash flow fell 78% to $74m; free cash flow for the half was $14.5m against a full-year guide of ~$400m; and unresolved claims — revenue booked against customers who have not agreed to pay — went from ~$180m to ~$400m to ~$680m in eighteen months, an increase that almost exactly matches the cash that went missing. Strip the recurring “one-time” items (restructuring in six straight years, and $504m of cumulative losses from operations “discontinued” in 2020 that still generate no revenue) and normalised owner earnings are roughly $4.50, not the guided $6.00 — 16x, not 12x.

The framing is neither falling knife nor abandoned value: it is a category de-rating that the company then validated. The factor model is unusually clear here — ACM carries no Momentum, Quality or LowVolatility loading at all, and a Value loading of effectively zero, so none of the standard cheap-stock or broken-stock labels fit; 72% of its return variance is idiosyncratic. The decisive session was 12 February 2026, when ACM fell 12.4% three days after a beat-and-raise — alongside WSP −12.5%, Stantec −11.7%, Tetra Tech −11.6% and Jacobs −7.6%. That was the market repricing billable-hour engineering as AI-disintermediable, and it is not a judgement about AECOM. Then AECOM validated it with a cash-flow quarter that drew seven plaintiff-firm investigations. What makes AECOM the most interesting name in the group is that it is the only one that has bought the weapon: roughly 4% of its market cap spent on an “autonomous engineer” whose vendor advertises a 90% reduction in engineering time, inside a firm whose own 10-K says it earns money “through the billing of our employees’ time.” Management then set a 20%+ margin target on the productivity — and, tellingly, adds the AI’s own amortisation back to the margin metric that carries the target, worth ~54bp of the promised ~350bp. Rudd’s concession on the Q2 call is the honest version: on AECOM’s two AI-enabled wins, “It’s not necessarily that we’re going to see more revenue… We will see improved margins.” Fewer hours, better margin — provided the client never asks for the saving back. That is the whole debate, and four quarters will not settle it.

Two facts stop me going lower than HOLD, and two stop me going higher. In the stock’s favour: it is the cheapest name in its cohort on forward earnings and EV/EBITDA (10.8x and 9.2x) while earning the cohort’s highest recent return on capital, and the transportation “cliff” the market is pricing looks like a timing risk, not a quantum one — H.R. 8870 cleared committee 62–2 at $580bn. Against it: AECOM’s seven-year average ROIC is ~8.3%, roughly its cost of capital and squarely in Greenwald’s advantages-absent band, so FY24–25’s 12–14% is the exception rather than the franchise return; and the record backlog is hollower than the headline — Americas total backlog grew just 1.8% while Americas design revenue grew 8%, and essentially all the growth sits in the uncontracted bucket.

Conviction: medium. The bull trigger that would flip me: H2 FY2026 free cash flow converting the ~$385m still owed against guidance with the claims balance falling — that would prove the cash problem was Middle East timing, not contract friction, and the 11th-percentile multiple would then be a genuine gift. The bear trigger: another quarter of claims growth above $750m, or a write-down/settlement of the DOE and Montana legacy items, which would confirm that “adjusted” has been financing the story for six years. Tag: the cheapest P/E in a decade, on the least trustworthy E in a decade.

📈 Stock Price Action — Five-Year Event Map

AECOM spent four years as a quiet compounder — $60.55 (July 2021) to an all-time high of $132.85 on 31 October 2025, roughly a double — and then gave nearly half of it back in nine months. At $71.76 the stock is −46.0% from that high, sits 17% below its $86.47 200-day EMA, and trades near the bottom of a 52-week range of $66.86–$132.85. Three single-day declines greater than 11% did almost all of the damage; only two of the three were about AECOM.

1. Jul 2021 – Dec 2022. Having sold Management Services and exited self-perform construction in 2020–21, AECOM was re-rated from a low-multiple contractor to a professional-services firm. The IIJA’s November 2021 enactment supplied the demand narrative. (FACT: the move. INTERPRETATION: the attribution.)
2. Jan 2023 – Dec 2024. The operating story worked: segment adjusted margin rose from 14.7% (FY23) to 15.8% (FY24), G&A stayed flat in dollars, and the share count fell from 140.1m to 136.5m.
3. Jan – Oct 2025. AECOM hit its 17%+ long-term margin target five quarters early (Q3 FY25) and set records on nearly every metric, peaking at $132.85 — roughly 22x forward adjusted EPS.
4. 19–21 Nov 2025. The 18 November investor day raised long-term targets (20%+ margin exit rate by FY28; 15%+ adjusted EPS CAGR) and raised the dividend 19%. The stock fell 19% in three sessions because the FY26 guidance underneath it cut free cash flow from $685m to ~$400m to fund AI restructuring, and guided margin to 16.6% — flat. Four analysts cut targets; Barclays moved to Equal-Weight ($135 → $100), citing the “complexity” the AI-driven targets added.
5. 11–12 Feb 2026. AECOM reported Q1 FY26 on 9 February — a beat, a guidance raise, record backlog on a 1.5x book-to-burn — and rose 1.1% the next day. Then on 12 February the entire category repriced: ACM −12.4%, WSP −12.5%, Stantec −11.7%, Tetra Tech −11.6%, Jacobs −7.6%, AtkinsRéalis ~−10%. Morningstar titled its note “AI Disintermediation Fears Drag Down Engineering and Construction Stocks.” This move was not about AECOM’s results.
6. 12 May 2026. Q2 FY26 raised guidance for a second consecutive quarter — and disclosed $4m of operating cash flow, negative $27m of free cash flow, and constant-currency NSR growth of 2% against a 6–8% organic guide. Peers fell only 2–3% that day; this one was AECOM’s own.
7. Jun – Jul 2026. The stock has based in the high-$60s/low-$70s. The June dividend was held flat at $0.31 after November’s 19% raise, and the CEO, CFO and President made the first open-market purchases in five years — totalling $0.70m.

Price moves are FACT (AZI daily price history); attributed causes are INTERPRETATION. No price target, recommendation, or technical level is expressed or implied in this section.

1. Executive Summary

AECOM is the largest architectural and engineering design firm in the world by design revenue (ENR 2025 Design Survey) and the #1-ranked firm in water, transportation design, facilities design, environmental engineering, environmental consulting and environmental science. It sells engineering judgment by the hour to governments and infrastructure owners across roughly 51,000 employees, and since exiting self-perform construction in 2020–21 it has been a clean, capital-light professional-services business: capex is 0.7–0.9% of revenue, and free cash flow ran a remarkably stable 8.8–9.8% of net service revenue for five straight years.

The five-year operating record is genuinely good and should not be dismissed. Net service revenue grew from roughly $6.0bn to $7.57bn while headcount stayed flat at ~51,000; gross margin on revenue expanded in every single year, 5.99% → 7.54%; segment adjusted operating margin rose 14.2% → 16.5%; general and administrative expense was $155.1m in FY2021 and $157.8m in FY2025, flat in dollars against 19% NSR growth. Book-to-burn has exceeded 1.0 for 22 consecutive quarters and backlog reached a record $26.2bn in Q2 FY2026. This is a real business with real operating leverage on a fixed overhead base.

The problem is that the operating leverage stopped, the cash stopped, and the reported earnings did not. In the six months to 31 March 2026: revenue fell 2.0%; gross margin on NSR fell 10bp; GAAP operating income fell 5.1% to $469.8m — while “adjusted” operating income rose 8.3% to $544.0m and management raised full-year adjusted EPS guidance twice. Operating cash flow fell 78.4% to $74.0m, free cash flow was $14.5m against a reaffirmed ~$400m full-year guide, and owner free cash flow (after $30.8m of stock compensation) was negative. The reconciling gap between GAAP and adjusted operating income widened tenfold year-on-year, from $7.2m to $74.2m.

The cash did not vanish; it moved onto the balance sheet. Contract assets rose $379.0m in six months on falling revenue, and unresolved claims — amounts recognised as revenue from customers who have not agreed to pay them — went from ~$180m (FY24) to ~$400m (FY25) to ~$680m at 31 March 2026. The $280m six-month increase in claims almost exactly matches the $268m six-month fall in operating cash flow. AECOM simultaneously increased its receivables-factoring programme by $51.2m, flattering even the depressed reported figure. Claims now equal 30% of book equity and roughly one year of normal free cash flow.

The “one-time” items are not one-time. AECOM has booked restructuring in six consecutive fiscal years (FY23 $188.4m, FY24 $98.9m, FY25 $59.4m, H1 FY26 $41.5m, with more guided) and has recorded after-tax losses from operations “discontinued” in 2020 in every year since — $504.5m cumulatively, on businesses that generated exactly zero revenue in H1 FY2026. Normalising for recurring restructuring, real intangible amortisation and the legacy drag produces owner earnings of roughly $4.50 per share against guided adjusted EPS of $6.00 — 16x rather than 12x at $71.76.

Capital allocation is the clearest management indictment. AECOM spent $3,028m on buybacks between FY2021 and H1 FY2026 to retire a net 15.0m shares — $201.88 of cash per net share retired, against a $71.76 share price. In September 2025 alone it bought 1,970,954 shares at an average of $126.91, weeks before the all-time high; the $790m deployed across FY25–H1 FY26 at an average of $108.96 is worth $520m today, a $270m loss equal to ~2.9% of the entire company. Total capital returned over the period equalled 112% of cumulative free cash flow, funded by a $707m increase in net debt. Insiders have made three open-market purchases in five years, totalling $699,390, all after the 46% drawdown, against $62.1m of sales at an average of $93.30.

Against all of that, the de-rating itself was not primarily about AECOM. The decisive 12 February 2026 session — ACM −12.4%, three days after a beat-and-raise — was a category event: WSP −12.5%, Stantec −11.7%, Tetra Tech −11.6%, Jacobs −7.6%. The market repriced billable-hour engineering as AI-disintermediable. AECOM is the most exposed name to that narrative and the only one that has bought the weapon: it spent $375.9m (per its filings; ~$390m per trade press) on AI technology whose vendor advertises a 90% reduction in engineering time, then set a 20%+ margin target on the productivity — while adding the technology’s own amortisation back to the margin metric that carries the target.

The framework verdicts. Industry: structurally mediocre — fragmented, labour-constrained, with barriers that are credentials and relationships rather than economics. Competitive position: a narrow, real advantage in large-programme delivery and technical depth, but not a Greenwald barrier to entry; AECOM’s scale buys it a seat at every table, not a price. Growth: decelerating and lower-quality than the backlog headline implies (only $13.8bn of $26.2bn backlog is contracted). Financial quality: deteriorating at the margin, with a live cash-conversion problem. Capital allocation: poor. Valuation: fair rather than cheap once the add-backs are removed.

No recommendation and no price target appear in this memo outside Claude's Take.

2. Business Overview

2.1 What AECOM actually sells

AECOM’s 10-K is unusually candid about the business model: “Our business focuses primarily on providing fee-based knowledge-based services. We primarily derive income from our ability to generate revenue and collect cash from our clients through the billing of our employees’ time spent on client projects and our ability to manage our costs.” The unit of production is an engineer-hour. Everything in this memo — the margin story, the AI debate, the cash-conversion problem — reduces to what happens to the number of hours, the price per hour, and the speed with which clients pay for them.

The services span the asset lifecycle: advisory and planning; architectural and engineering design; programme management and construction management. The technical disciplines include civil, structural, process, mechanical, geotechnical and electrical engineering, architecture, landscape and interior design, urban and regional planning, project economics, cost consulting and environmental health and safety work. The stated commercial logic is land-and-expand: “our work for a client may begin with a small consulting or planning contract, and may later develop into a broader advisory, design or overall management role for the project or a series of projects, which we refer to as a program.”

Since the 2020–21 exits — Management Services (sold to what is now Amentum), self-perform at-risk civil infrastructure, power construction and oil & gas construction — AECOM has been a pure design and consulting firm. Those businesses sit in discontinued operations in every period presented, and, as the Financial Quality section shows, they are still costing money.

2.2 Segments and the pass-through mechanic

AECOM reports three segments: Americas (US, Canada, Latin America), International (Europe, Middle East, India, Africa, Asia-Australia-Pacific), and AECOM Capital (a legacy real-estate co-general-partner fund in wind-down, whose team was transitioned to a third-party platform in Q3 FY2024).

The single most important presentational fact about AECOM is that GAAP revenue is not a meaningful top line. In the course of delivering projects the firm subcontracts services and incurs other direct costs on behalf of clients; these are passed through at little or no margin but are recognised in both revenue and cost of revenue. Pass-through revenue was $8.6bn of $16.14bn of FY2025 revenue — 53% (FY2024: $8.9bn, 56%). AECOM’s own chief operating decision maker evaluates the segments on “revenue less pass-through revenue” — net service revenue (NSR). So does this memo.

The pass-through intensity is wildly asymmetric between the segments, and this materially changes how the company should be read:

International is roughly twice as important to earnings as the gross-revenue optic implies — 40% of NSR versus 22% of revenue. That matters because International is the segment that is currently shrinking: Q2 FY2026 International NSR fell 3% in constant currency on declines in Asia and the Middle East. Any analysis that reads AECOM off GAAP revenue systematically under-weights the deteriorating half of the business.

The margin arithmetic follows the same logic. FY2025 gross profit of $1,216.7m is only 7.5% of GAAP revenue but 16.1% of NSR. AECOM’s headline “segment adjusted operating margin” — the metric carrying the 20%+ target — is struck on NSR.

2.3 The five-year top-line and margin record

Three observations. First, the margin improvement is real and continuous: gross margin on revenue rose in each of five consecutive years, and G&A was held flat in absolute dollars while NSR grew ~26%. That is textbook operating leverage on a fixed cost base and it is the strongest single fact in the bull case. Second, FY2023 is not a usable base year: GAAP operating income of $324.1m was depressed by a $279.4m equity in losses of joint ventures, against +$35m, +$54m, +$2m and +$27m in the surrounding years. Third, and most important, the improvement has stopped. H1 FY2026 gross margin on NSR was 15.07% against 15.17% a year earlier, GAAP operating margin fell 20bp, and consolidated revenue is now declining outright.

2.4 Clients, end markets and contract types

Two things follow. Direct US federal exposure is only ~7% of revenue — dramatically lower than at Jacobs, Leidos, Booz Allen or CACI, and materially lower than the market’s treatment of the whole “government services” cohort implies. The US federal budget and government-efficiency risk that dominates that cohort is a second-order issue here. But government dependence in aggregate has risen from 43% to 50% in three years while private fell 57% → 50%. AECOM is becoming more exposed to public budgets, not less — the opposite of the direction Jacobs has taken. Public budgets are stable but slow, price-sensitive and competitively bid.

No single client exceeded 10% of revenue in any of the past five years, which removes concentration risk but also tells you something about pricing power: this is a firm with tens of thousands of individually-competed engagements, not a handful of entrenched franchise contracts.

AECOM does not disaggregate revenue by end market in the 10-K, which is a real disclosure gap for a company whose story rests on end-market mix. Only directional MD&A commentary is available: in FY2025 Americas transportation rose $261.1m (+11.7%) and water & environment rose $127.4m (+6.0%), while facilities fell $344.7m (−4.3%). Working backwards from those pairs implies a facilities base of roughly $8.0bn of Americas gross revenue — i.e. facilities, which contains the pass-through-heavy, low-margin construction-management business, is by far the largest Americas end market and it is shrinking. (ASSUMPTION derived from MD&A dollar/percentage pairs, not a disclosed figure.)

Contract mix is deteriorating in a way that deserves more attention than it gets:

The low-risk cost-reimbursable share has fallen 4.6 points in two years while GMP and fixed-price have risen to a combined 62%. The 10-K is explicit about what GMP means: “Cost overruns or costs associated with project delays in completion could be our responsibility.” That exposure sits overwhelmingly in the Americas construction-management business — the one AECOM put up for sale in November 2025 and then decided to keep eleven weeks later.

2.5 Backlog is not contracted revenue

AECOM reports a headline backlog that reached $26.2bn in Q2 FY2026, and a 22-quarter streak of design book-to-burn above 1.0. Both are genuine positives. But the composition matters:

• Of the $26.2bn Q2 FY2026 total backlog, only $13.8bn (53%) is “Contracted.” The remaining $12.4bn is “Awarded” but not contracted.
• In FY2025, of $39.7bn of total backlog (a broader definition including construction management), only $19.7bn was ASC 606 remaining unsatisfied performance obligations — the 10-K states plainly that RUPO is “$20.0 billion lower than backlog.” Roughly 57% of that RUPO converts within twelve months.

This is project revenue that is re-competed, not annuity revenue. The recurring element is relationship- and framework-driven (IDIQ vehicles, master service agreements, UK water AMP regulatory cycles), not contractual.

2.6 Unit economics — the number that matters in a people business

This is the most revealing table in the memo. Headcount has been flat at 50,000–52,000 for five years while NSR grew roughly 26%. Essentially 100% of AECOM’s growth came from revenue per head — bill-rate escalation, mix shift toward higher-value advisory and programme management, and offshoring into the India and Poland enterprise capability centres — and 0% from adding people. The spread between what a head bills and what a head costs widened 22% in two years.

Read one way, this is the single best piece of evidence that AECOM has genuine pricing power and a working productivity engine, and that the 20%+ margin ambition is not fantasy. Read the other way, it is the thesis’ greatest vulnerability: a firm whose entire growth algorithm is price-per-hour is precisely the business model most exposed to a technology that deflates hours. Both readings are correct, and the Competitive Position section takes them seriously.

Verdict: A clean, capital-light, genuinely scaled professional-services business with a demonstrated five-year productivity record — sitting on a revenue base that has stopped growing, a contract mix that has become riskier, a backlog that is only half contracted, and a growth algorithm entirely dependent on the price of an hour.

3. Industry Dynamics

3.1 Sizing the pool honestly

Published “global engineering consulting TAM” figures (~$203bn growing ~4.3%) are press-release grade with no visible methodology and should not carry weight. The one sizing with a transparent method is ENR’s Top 500 Design Firms list, which aggregated $158.7bn of 2025 design revenue, +7.4% year on year — US domestic $136.3bn (+8.1%), international $22.4bn (+3.2%).

Against that, AECOM’s entire global net service revenue of $7.57bn is ~4.8% of the Top-500 pool. This is the first and most important structural fact: the largest design firm on earth holds a mid-single-digit share of a list of 500 firms, with thousands more below it. In Greenwald’s framework, an industry where you cannot count the leaders on one hand and where the leader’s share is 5% is, by construction, an industry without meaningful barriers to entry.

The narrower niches are only modestly better. In US environmental & sustainability consulting (~$25bn), the “Big Four” of Tetra Tech, AECOM, WSP and Jacobs held just over 40% combined in 2024 — down 2.0 percentage points from 2023, and up only 1.3 points since 2021. Consolidation has plateaued.

3.2 The competitive scoreboard

(AECOM’s 16.8% adjusted EBITDA margin is struck including noncontrolling interests. On the consistent adjusted-EBITDA-over-NSR basis used for the peers it is 15.9% — $1,203m ÷ $7,573m — which places AECOM third of the four listed majors. See the Competitive Position section.)

The scoreboard sorts by ownership form, not by skill. Listed, shareholder-owned consultancies earn 14–18% adjusted EBITDA on net revenue. Employee-owned Mott MacDonald and foundation-owned Ramboll earn 4–5% on gross revenue — because they are run to maximise employee compensation, not return on capital. A meaningful share of global engineering capacity is held by firms with no requirement to earn a cost of capital, with private-equity roll-ups occupying the tier below. This structurally caps price: the marginal bidder in many pursuits does not need a 17% margin to stay in business.

Within the wider value chain the margin hierarchy runs: self-perform specialty trades that own scarce craft labour (EMCOR ~9% operating margin, ~31% ROIC) > asset-light design and consulting (14–18% EBITDA on net revenue) > lump-sum megaproject EPC (Fluor ~3%). AECOM sits in the middle tier — the right tier — but that tier’s economics are set by labour arbitrage, not by barriers.

3.3 Barriers to entry — the incumbent’s own verdict

The most important sentence in this section is AECOM’s own, from Item 1 of the FY2025 10-K:

“The markets we serve are highly fragmented and we compete with a large number of regional, national and international companies… The technical and professional aspects of our services generally do not require large upfront capital expenditures and, therefore, provide limited barriers against new competitors.”

The world’s largest design firm states in its own annual report that its industry has limited barriers to entry. Qualifications, professional licensure, past-performance references, framework/IDIQ prequalification and security clearances are real frictions — they slow entry onto a given agency’s shortlist — but they are table stakes held by dozens of firms, not exclusionary barriers. There is no plant, no patent-protected process and no regulated licence limiting the number of participants. Capex is 0.7–0.9% of revenue.

3.4 Marathon capital-cycle read — and why the usual framework does not apply

Capital is flowing in. M&A accelerated through 2025 on sustained private-equity appetite, particularly at the $1.5–8m EBITDA level, with public firms pursuing roll-ups of small accretive businesses. WSP bought POWER Engineers (2024) and agreed to buy TRC Companies (December 2025); Stantec added 3.9 points of acquired growth in 2025; Jacobs bought in the PA Consulting minority at 13x EBITDA (March 2026).

But the classic Marathon capex cycle does not bind here, because capacity in this industry is not built with capital — it is hired. There is no three-year shipyard lag or nine-year mine lag to create the supply discipline Marathon’s framework relies on. Headcount is growing at the strong players and shrinking at the weak: WSP at 75,214 and growing; Mott MacDonald +2.5%; AECOM flat at ~51,000; Arcadis cutting 1,100; Ramboll organic −2.5%; Wood Group revenue −13.3% and absorbed by Sidara in March 2026.

The supply shock in progress is AI, and it runs the wrong way for the industry. If generative and agentic design tools materially raise output per engineer, the effective supply of engineering capacity expands enormously with almost no capital and no lag. In a fragmented, low-barrier industry with sophisticated repeat buyers who procure on qualifications-then-price, the base rate is that such a productivity gain is competed away to the client within one or two procurement cycles. That is the bear case stated in capital-cycle terms, and it is the correct way to think about it.

3.5 Funding drivers

The US surface transportation cliff. The IIJA’s five-year authorisation of federal surface transportation programmes expires 30 September 2026 — nine weeks from this report date. The House response is the BUILD America 250 Act (H.R. 8870), a five-year, $580bn package ($474.4bn guaranteed from the Highway Trust Fund plus $106bn subject to appropriation), which cleared the House Transportation & Infrastructure Committee 62–2 on 22 May 2026 and shores up Highway Trust Fund revenue with new EV fees ($130 per EV, $35 per plug-in hybrid).

The market has priced a funding cliff. The legislative evidence points to a timing risk, not a quantum risk: the headline dollars are an increase, and a 62–2 committee vote signals genuine bipartisan intent. The realistic downside is a lapse into short-term extensions — the base rate, since MAP-21 and the FAST Act were each preceded by dozens of them, and every surface transportation bill since 1991 has needed extensions before a successor passed. Extensions freeze new-start programming at state DOTs, which delays award of precisely the planning, design and programme-management work AECOM sells. That is a real revenue-timing risk. It is not the end of federal transportation funding, and the gap between that perception and the evidence is a candidate variant perception (the Variant Perception section).

Separately, the IIJA outlay tail runs well beyond the authorisation expiry: $591bn has been directed to 72,000+ projects. Management’s statements on how much remains are, notably, not consistent with each other — Q3 FY25 said only 36% of IIJA funding targeted at its markets had been spent, while the Q2 FY26 call said “more than half of the IIJA funding remains to be spent.” Treat “roughly half remains” as the honest range.

Water. The FY2026 President’s Budget proposed a $2.4bn reduction effectively zeroing the Clean Water and Drinking Water State Revolving Funds. Congress did not enact it: P.L. 119-74 provides $3.04bn for EPA water infrastructure, level with FY2025, and EPA announced $7.2bn of FY2026 SRF allotments. In the UK, AMP8 (April 2025 – March 2030) is a £104bn water capital programme, but consultants publicly warned of a slow start in its first six months and several consultancy frameworks only became effective on 1 July 2026. AECOM was named preferred bidder on Scottish Water’s multi-billion-pound programme. AMP8 is a real, funded, multi-year tailwind — but its revenue recognition is back-end-loaded into FY27–FY29. It supports the out-years, not FY26.

The Middle East is deteriorating badly, and this is not hypothetical. Saudi Arabia is expected to spend ~SAR 60bn ($16bn) cancelling NEOM contracts between 2026 and 2030 — more than it spends building them, and more than a third of the 2026 budget deficit. NEOM terminated three contracts worth $6.85bn in March 2026 (Webuild’s Trojena dams and “The Bow”; Hyundai E&C’s Line tunnel); The Line is deferred past 2030; the 2029 Asian Winter Games were indefinitely postponed on 24 January 2026. AECOM’s Middle East NSR declined in H1 FY26 and cost roughly 100bp of Q2 NSR growth. Management’s pivot narrative — Saudi sports and entertainment, Riyadh mixed-use, and $40–50bn of prospective US military infrastructure repair in the region — is a pipeline, not a backlog.

Offsetting positives are real. Australia: AUKUS-related defence wins (an A$3bn partnership) and appointment as Brisbane 2032 Olympics Delivery Partner. Canada: broad-based growth on national and provincial funding. US defence: a $1.5tn budget proposal, with AECOM’s Department of War pipeline up 50% year on year and defence roughly 10% of the portfolio — the Department of War alone is AECOM’s single largest client at 5%+.

Verdict: a structurally below-average industry currently riding an above-average demand cycle. The largest firm on earth holds ~5% of the pool; the incumbent’s own 10-K concedes limited barriers; capacity is people and therefore responds to price within a hiring cycle, and with AI potentially faster than that; a large slice of competing capacity does not require a shareholder return and caps industry pricing; and the dominant buyers are public agencies procuring on qualifications-then-price with scheduled re-competition. Demand is genuinely strong and genuinely funded — but strong demand in a low-barrier industry attracts supply and is competed away.

4. Competitive Position

4.1 Naming the advantage — and refusing to overstate it

In Greenwald’s taxonomy there are exactly three genuine competitive advantages: supply/cost advantages, demand/customer captivity, and economies of scale combined with some captivity. AECOM’s position must be tested against each.

Supply/cost advantage — partial and real. AECOM’s delivery cost base is genuinely lower than a domestic-only competitor’s because of its enterprise capability centres in India and Poland, which allow a US-billed hour to be produced partly with offshore labour. This is the mechanism that actually produced the five-year margin expansion, and it is not available to a regional firm. But it is available to WSP, Stantec, Jacobs, Arcadis and every other global peer, all of whom run the same model. It is a cost advantage versus small firms, not versus the relevant competitive set.

Customer captivity — weak. Public agencies procure on qualifications-then-price and re-compete on a schedule. Framework and IDIQ agreements confer the right to bid, not the right to the work. No client is more than 10% of revenue, which is the arithmetic signature of a business with thousands of separately-won engagements rather than entrenched franchises. Switching costs exist at the project level — replacing the designer mid-programme is genuinely costly — but not at the client level, which is where captivity would have to bind for it to be an advantage.

Economies of scale plus captivity — the strongest claim, and it is narrow. On the largest and most technically complex programmes — a multi-billion-dollar high-speed rail programme, an Olympics delivery partnership, a national water capital programme — the list of firms that can credibly bid is genuinely short, perhaps four to six globally. There AECOM’s balance sheet, global technical bench and past-performance record are a real barrier, and it is where the company plays best. But this is a segment advantage inside a fragmented industry, not a company-wide moat, and the qualified-bidder list is short enough to be oligopolistic only at the very top of the market.

4.2 Running the tests

The market-share stability test — AECOM fails it. Greenwald’s single most reliable diagnostic is whether the leader’s share is stable over time. AECOM’s own 10-Ks describe it as ENR’s second-largest design firm in the 2021, 2022, 2023 and 2024 Design Surveys; largest in the 2025 survey (as stated in the FY2025 10-K filed 19 November 2025); and then in ENR’s 2026 Top 500 list, published April 2026, Jacobs took No. 1 and AECOM slid back to No. 2 — driven by Jacobs’ data-centre business, which grew 62.2% with a pipeline up 500%. One year at the top of a two-horse race that keeps swapping is not dominant-firm longevity. In the narrower US environmental & sustainability niche, the Big Four’s combined share fell 2.0 points in a year.

The relative-growth evidence corroborates it: in FY2025 AECOM grew net service revenue more slowly than every healthy listed peer, and added backlog more slowly than all of them — design backlog +3%, against Jacobs +22%, AtkinsRéalis +20%+, WSP +10% and Stantec +9.5%. A leader losing rank and growing slower than the field, in an industry where the four largest firms are collectively losing share, does not have a durable structural advantage.

The manner of the loss compounds the point. Jacobs took the crown on AI-related infrastructure — the data-centre buildout that ENR itself attributes to an expected $1.75tn of AI infrastructure spending. AECOM is being de-rated for AI disintermediation while a direct competitor overtakes it on AI-driven demand. AECOM’s facilities end market, which contains its data-centre-adjacent work, shrank 4.3% in FY2025.

The ROIC test — passes on two years, fails on seven. This is the test that matters most, and the answer changes entirely with the window:

Greenwald’s point is that a genuine competitive advantage shows up as persistently above-cost-of-capital returns, not two good years. AECOM’s seven-year average ROIC of ~8.3% sits in the “advantages absent” band — roughly at its own cost of capital. Tetra Tech passes the persistence test; AECOM does not. And the two good years are fully explicable without invoking a moat: the 2020–21 exit from self-perform construction, $346m of restructuring across FY2023–25, the offshore capability centres, and — critically — a shrinking capital base, with equity falling from $3,690.6m (FY2019) to $2,492.6m (FY2025) while more than $3.5bn was returned to shareholders.

That last mechanic also explains why AECOM’s FY2025 ROE (24.0%) sits ten points above Tetra Tech’s (13.7%) on a near-identical ROIC: the gap is the negative tangible equity, not superior operations. A return on capital that looks impressive partly because the capital has been bought back is not the same thing as a business that has become better. In Greenwald’s terms all of this is operational effectiveness — emulable, and emulated — rather than a barrier.

The margin test — AECOM is mid-pack, not best, and management’s claim does not survive. On a consistent adjusted-EBITDA-over-net-revenue basis: WSP 18.3% > Stantec 17.6% > AECOM 15.9% ($1,203m ÷ $7,573m) > Jacobs 13.9%. AECOM is the largest firm and the third-most profitable of four. Note that AECOM’s own quoted 16.8% adds back noncontrolling interests, and its 16.5% “segment adjusted operating margin” excludes roughly $150m of corporate G&A and AECOM Capital. The 20%+ FY2028 target that management says “would substantially lead the industry” therefore starts from a base that does not lead it — a management-commentary-versus-evidence contradiction that should be flagged as such.

And the consolidated margin conceals a bimodal business. FY2025 adjusted operating margin on NSR was Americas 19.8% (a record, +90bp, reaching 20.0% in Q2 FY2026) against International 11.5% (flat). The headline 16.5% is the average of one genuinely best-in-class business and one mediocre one — and the mediocre one is 40% of net service revenue and shrinking.

4.3 The AI question, taken seriously in both directions

This is the question on which the stock now turns, and it deserves to be argued rather than asserted.

The bear mechanism. AECOM’s own 10-K says it earns money “through the billing of our employees’ time.” If AI reduces the hours needed to produce a deliverable, then in a fragmented industry with limited barriers and sophisticated repeat buyers who competitively re-bid, the saving accrues to the client, not the firm. The buyer is a public agency with a procurement department and a duty to obtain value; it will observe that a design that took 10,000 hours now takes 3,000 and will price the next one accordingly. AECOM’s growth algorithm makes this acute: 100% of five-year NSR growth came from revenue per head, none from adding heads. A firm whose entire growth is price-per-hour is maximally exposed to hour deflation.

The bull mechanism. Three arguments have force. First, the constraint is supply, not demand. WSP’s CEO put it directly: “We are capacity constrained” — engineer supply grows roughly 1% a year against 4–7% revenue growth. If the industry cannot hire enough engineers to do the funded work, a productivity tool clears a backlog rather than destroying revenue, at least until the backlog is cleared. Second, contract structure decides who captures the saving, and AECOM’s book is 38% cost-reimbursable — where fewer hours mechanically means less revenue — but 62% GMP and fixed-price, where fewer hours means the same revenue at a better margin. Third, AECOM has already been doing this for five years. Holding headcount flat while growing NSR 26% and expanding gross margin every year is the operating-leverage mechanism working; AI is a continuation, not a discontinuity.

Where the evidence actually points. Management’s own framing concedes the revenue mechanism and pivots to margin. Asked by Citigroup’s Andrew Kaplowitz whether an AI-shaped value model leads to shrinking revenue, CEO Troy Rudd answered that clients “are always expecting more value from us. And when we can provide more value they’ve always been willing to pay us for that.” On the Q2 FY2026 call, describing two AI-enabled wins worth nearly $1bn in aggregate, he was more precise: “as we deliver using AI, we have a mechanism where we effectively will share the benefit from doing that… It’s not necessarily that we’re going to see more revenue from these particular contracts. We will see improved margins on those contracts.”

That is a candid and specific answer, and it is a margin story built on a shrinking hour base. It is durable only for as long as AECOM’s AI advantage is durable — which, in an industry with limited barriers where the same tools are being sold to everyone, and where the acquired vendor was until recently selling to the whole market, is the crux. The honest position is that AECOM has bought itself a two-to-three-year lead in a race where leads are not historically defensible.

Verdict: a narrow, real, but non-structural advantage. AECOM has genuine scale, a genuine offshore cost platform, a genuine short-list position on the world’s largest programmes, and a demonstrated five-year productivity record. It does not have a Greenwald barrier to entry — it says so itself — and it lost the industry’s leadership rank this year to a competitor riding the AI-infrastructure wave it is being punished for. This is a good business in a mediocre industry, not a franchise.

5. Growth History and Forward Opportunities

5.1 The record

Two facts sit in tension and both are true. NSR growth has been positive and reasonably steady for four years, and it is decelerating — 6.9% → 5.7% → 3.6%, with Q2 FY2026 constant-currency NSR growth of just 2% against a 6–8% organic guidance. Meanwhile GAAP revenue is now shrinking outright, and FY2025’s flat headline was itself the product of a $307.4m decline in pass-through revenue: neither Americas (+0.3%) nor International (−0.1%) grew gross revenue in FY2025.

5.2 Backlog: the strongest bull fact, honestly framed

Design book-to-burn has exceeded 1.0 for 22 consecutive quarters and total backlog reached a record $26.2bn (+8%) in Q2 FY2026, with the design pipeline at a record and growing double digits. That is a five-and-a-half-year streak of selling more than you deliver, and it is the single most persuasive item in the bull case. It is also a genuine leading indicator: an order book cannot be manufactured by an accounting policy.

Four qualifications keep it honest, and cumulatively they do most of the work of explaining why the market ignored the headline.

First, the audited backlog has gone nowhere. AECOM reports two coexisting backlog definitions — a 10-K figure and a narrower press-release figure ($39.7bn versus $24.8bn at the same date). On the 10-K definition, backlog went $41.2bn (FY2020) → $38.6bn → $40.2bn → $41.2bn → $37.4bn → $39.7bn (FY2025): down 3.6% over six years while NSR rose 24%. (The FY2023 figure was also restated downward by $1.6bn in the FY2024 10-K without explanation.) The ASC 606 remaining unsatisfied performance obligation — the audited, contractual measure — went from $18.7bn (FY2021) to $19.7bn (FY2025): +5% in four years, and sits $20.0bn below the stated backlog, with only ~57% converting within twelve months.

Second, all of the growth is in the uncontracted bucket. In Q2 FY2026, contracted backlog grew +3.7% while awarded-but-not-contracted backlog grew +13.2%; in Q1 FY2026 the split was +2.4% versus +16.5%. Roughly 47–48% of total backlog is uncontracted. The quality ordering is itself the warning: pipeline (double-digit) > awarded (+13–17%) > contracted (+2–4%) > RUPO (+5% over four years).

Third, the segment mix inverts the headline. Americas total backlog grew just +1.8% year on year (contracted +1.4%) while Americas design NSR grew 8% in constant currency — the high-margin engine is burning backlog roughly four times faster than it is building it. Meanwhile International backlog +25% is overwhelmingly one framework: International awarded backlog rose 62.2% ($2,008m → $3,258m), consistent with the Scottish Water preferred-bidder win. The order-book strength is concentrated in the lower-margin segment whose revenue is declining, and it is largely a single contract.

Fourth, book-to-burn is regionally lopsided. In Q1 FY2026 Americas book-to-burn was 1.0x (hurt by the 43-day federal shutdown) against International’s 2.3x.

5.3 Where growth is supposed to come from

The credible pools are real and funded: water and PFAS remediation (AMP8’s £104bn, EPA SRF at $7.2bn of FY2026 allotments, AECOM’s #1 ENR ranking in water); environment; defence and government facilities (a $1.5tn US budget proposal, a Department of War pipeline up 50%, AUKUS work in Australia); transportation, contingent on reauthorisation; and the two higher-margin service lines management is pushing — programme management and advisory.

The credible headwinds are equally real: Middle East contraction (NEOM cancellations, The Line deferred past 2030, ~100bp of drag on Q2 NSR growth), transportation award timing into the September 2026 authorisation expiry, Asia weakness, and a facilities end market that shrank 4.3% in FY2025.

Verdict: low-to-medium-quality growth that is decelerating. It is organic rather than acquired, which is a genuine positive, and it rests on funded public programmes rather than a cyclical private capex boom. But it is decelerating toward the low single digits in constant currency, it is running well below the company’s own 6–8% organic guidance, its order-book strength sits in the shrinking segment, and — most importantly — the entire five-year growth algorithm has been price-per-hour on flat headcount, which is exactly the variable the market has decided to worry about.

6. Financial Quality

6.1 The adjusted-earnings problem, quantified

AECOM reports an adjusted EPS that excludes restructuring and acquisition costs, amortisation of intangible assets, “non-core AECOM Capital” results, certain financing charges, and 100% of discontinued operations. Over five and a half years the gap between what was reported and what was earned is not a rounding difference — it is most of the story.

Cumulative pre-tax add-backs, FY2021 through H1 FY2026: restructuring $544.6m · “non-core AECOM Capital” $354.1m · intangible amortisation $110.9m · financing and prepayment charges $171.5m · discontinued operations $504.5m. After tax, the total is $1,425.8m — equal to 81% of the $1,757.6m of GAAP net income actually earned over the same period. Two independent measures agree: the EPS ratio (1.81x) and the dollar ratio (81%).

The items are not one-time by any reasonable definition:

• Restructuring has been charged in six consecutive fiscal years (FY2021 $48.0m, FY2022 $108.0m, FY2023 $188.4m, FY2024 $98.9m, FY2025 $59.4m, H1 FY2026 $41.5m), and the FY2025 10-K already guides to a seventh year of roughly $45m.
• “Discontinued” operations have generated after-tax losses every year since the 2020 divestitures — $504.5m cumulatively — while generating exactly zero revenue in H1 FY2026. These are pure legacy-liability charges on businesses sold six years ago: a $53.0m loss in Q3 FY2025 on the Montana refinery turnaround project after adverse post-trial orders, and a $61.8m non-cash loss in Q1 FY2026 from re-estimating recovery on the Department of Energy Deactivation, Demolition and Removal claims.
• Intangible amortisation has stepped up twentyfold, from $1.5m in H1 FY2025 to $30.1m in H1 FY2026, and is scheduled at $39.1m in each of FY2027 and FY2028. This is the amortisation of the AI technology AECOM bought — and it is added back to the very margin metric the AI is meant to improve (the Financial Quality section).

One further detail deserves attention: management restated its own adjusted FY2022 EPS from $3.47 to $3.40 when it introduced “non-core AECOM Capital” as an add-back category — that is, once AECOM Capital began losing money, its earlier profits were retroactively stripped out of the adjusted series. Adjusted metrics that are redefined after the fact are not a stable measuring stick.

6.2 The cash-flow break, and its mechanism

For five years free cash flow was remarkably stable at 8.8–9.8% of net service revenue. In the six months to 31 March 2026, operating cash flow fell 78.4% and owner free cash flow turned negative.

The cash did not disappear; it moved onto the balance sheet, and the disclosure that explains it is specific:

Unresolved claims recorded in contract assets and other non-current assets: ~$180m (30 Sep 2024) → ~$400m (30 Sep 2025) → ~$680m (31 Mar 2026).

That is a 3.8x increase in eighteen months, and the $280m rise in the six months to March almost exactly matches the $268m fall in operating cash flow over the same six months. Claims are amounts AECOM has recognised as revenue from customers who have not agreed to pay them; realisation depends on negotiation, arbitration or litigation. At $680m they equal 30% of the company’s $2,270.6m of book equity and roughly one full year of normal free cash flow. Contract assets in total rose $379.0m in six months while revenue fell 2.0%.

Scale the number against the guidance: the claims build in H1 FY2026 alone (~$280m) is 70% of the entire ~$400m full-year free-cash-flow guide, and the trailing five-quarter build of ~$510m is roughly 40% of one year’s adjusted EBITDA. A company whose equity story is “100%+ free-cash-flow conversion” is failing that test in the accounts while continuing to assert it in the guidance.

Management’s characterisation on 11 May 2026 was “longer-than-anticipated claim resolution on certain projects” and delayed Middle East collections, with the assurance that Middle East collections “have already recovered in the fiscal third quarter.” That may be true. But an eighteen-month, 3.8x trajectory is not timing, and it coincides precisely with the contract-mix shift toward guaranteed-maximum-price and fixed-price work described in the Business Overview — which is exactly where claims arise.

Two further adjustments make the reported figure look better than the underlying. First, AECOM sells trade receivables; sold receivables outstanding went $319.5m (Sep 2024) → $268.2m (Sep 2025) → $319.4m (Mar 2026), meaning reported H1 FY2026 cash flow was flattered by an extra $51.2m of factoring even as it collapsed. Underlying cash generation was worse than $74.0m. Second, days sales outstanding, which had improved genuinely from 80.2 days (FY2021) to 72.2 days (FY2025), deteriorated to 83.8 days at 31 March 2026 — an 11.6-day step backwards.

6.3 The non-GAAP mechanic that funds the margin target

AECOM’s “segment adjusted operating margin” — the metric that carries the 20%±by-FY2028 target — is built by adding back amortisation of intangible assets. In Q2 FY2026: Americas segment operating income $227.9m + $10.6m of amortisation = $238.5m adjusted; International $77.0m + $6.6m = $83.6m adjusted. Total added back: $17.1m in the quarter, versus $0.4m a year earlier.

So AECOM capitalised the AI technology it bought, is amortising it over five years inside cost of revenue, and then excludes that amortisation from the margin the AI is supposed to lift. The cost of the AI is adjusted out; the benefit of the AI is left in. At the scheduled $39.1m of FY2027–FY2028 amortisation on roughly $7.3bn of NSR, the add-back alone is worth about 54 basis points of “adjusted” margin — roughly 15% of the entire ~350bp journey from 16.5% to 20%+.

This is a disclosed, conventional non-GAAP treatment, not an accounting violation. But it means a material slice of the promised operating leverage is definitional rather than economic, and an investor underwriting the 20% target is partly underwriting an add-back.

The consolidated consequence, stated as plainly as the numbers allow: in H1 FY2026, GAAP income from operations fell 5.1% to $469.8m while adjusted income from operations rose 8.3% to $544.0m — a 13.4-point divergence. The reconciling gap widened from $7.2m to $74.2m, a tenfold increase, in the same half-year in which management raised full-year adjusted EPS guidance twice.

6.4 Balance sheet

Tangible book equity is negative $1,669.5m, or −$13.02 per share. Goodwill and intangibles of $3,940.1m are 1.74x total book equity and 32.8% of total assets; retained earnings are a deficit of $1,475.6m. The reported book value per share of $17.57 is entirely goodwill. This is why the AZI P/B percentile of 62.5 should be ignored — a price-to-book ratio is not informative when book value contains no tangible net worth.

Liquidity is adequate and there is no near-term maturity wall. The stack: $1,200m of 6.000% senior notes due 2033; a $950m Term Loan A and $1,500m revolver due March 2031 (Amendment No. 16, 10 March 2026); a $500m Term Loan B due April 2031; plus a new standalone $500m revolving credit facility due 9 June 2028, secured by a lien on substantially all assets, signed 10 June 2026 and undrawn. The covenant is a consolidated leverage ratio of ≤4.00x tested quarterly; reported net leverage was 1.2x. Off balance sheet, $903.8m of standby letters of credit are outstanding. The average effective interest rate was 5.3% in H1 FY2026, and the July 2025 refinancing raised the bond coupon 87.5bp (5.125% → 6.000%) at a $9.1m make-whole cost in exchange for a six-year maturity extension.

Two observations. Cash fell $551.4m and net debt rose $555.4m in the six months in which AECOM generated $14.5m of free cash flow and spent $441.9m on buybacks — the buyback was funded from the balance sheet, not from cash flow. And adding a second, secured $500m revolver three months after a full refinancing that already provided $1.5bn of revolving capacity is an unexplained action worth watching; the 8-K gives no purpose, and the candidate explanations (letter-of-credit capacity, a liquidity buffer against the H2 cash catch-up, or preparation for a transaction) are not distinguishable from the disclosure.

6.5 Returns

On our own computation, NOPAT over book invested capital rose from 12.8% (FY2021) to 20.7% (FY2025), or 17.7% lease-adjusted, against a weighted average cost of capital of roughly 8.5–9.5%. The spread is genuinely positive and AECOM leads its peer group.

The caveats are those set out in the Competitive Position section: roughly a third of the improvement is a shrinking denominator, since invested capital fell from $3,836m to $3,329m as buybacks consumed equity while goodwill remained. AECOM’s ROE sits ten points above Tetra Tech’s on a near-identical ROIC, and that gap is the negative tangible equity, not superior operations.

Verdict: economics did improve with scale — and then stopped. From FY2021 to FY2025 AECOM delivered 278bp of NSR margin expansion on flat dollar overhead, which is real operating leverage and deserves credit. But the improvement has stalled, revenue is declining, days sales outstanding have jumped 11.6 days, claims have tripled, cash conversion has broken, and adjusted EPS has exceeded GAAP by 81% over five years on add-backs that recur every single year. The order book is intact; the conversion is broken.

7. Capital Allocation

7.1 The buyback

This is the clearest management indictment in the file. Between FY2021 and H1 FY2026 AECOM spent $3,028.2m on share repurchases and retired a net 15.0 million shares (143.2m → 128.2m). That is $201.88 of cash per net share retired, against a share price today of $71.76. The difference is dilution: $252m of stock-based compensation expensed and $146.4m of stock issued for acquisitions in FY2025 alone. A material share of the buyback is not shrinking the count — it is mopping up issuance.

The timing is worse than the arithmetic:

In September 2025 alone AECOM bought 1,970,954 shares at an average of $126.91 — $250.1m deployed within weeks of the $132.85 all-time high. That block is worth $141.4m today. Across FY2025 and H1 FY2026 the company deployed $790.4m at an average of $108.96, now worth $520.5m: an unrealised loss of $269.9m, equal to roughly 2.9% of the entire market capitalisation, destroyed through timing alone.

The behavioural pattern compounds it. AECOM did not accelerate as the stock fell — it bought heavily after the November crash (2,975,270 shares at ~$103.71) and again after the February crash (1,180,876 at $98.33), then bought zero shares in January 2026 and zero in March 2026 with $883.9m of authorisation available and the stock in the $80s and $90s. (Activity in the June 2026 quarter is not yet disclosed; the Q3 10-Q had not been filed at the report date.) Board authorisations were repeatedly topped back up to $1.0bn — most recently in February 2026.

Total capital returned over the period was ~$3,516m against ~$3,143m of cumulative free cash flow — 112% of free cash flow, before a dollar of M&A, with the gap funded by a $707m increase in net debt. Returning more than you generate is defensible for a genuinely capital-light compounder with a stable moat. It is not defensible at 1.5x the eventual share price while the underlying cash conversion is deteriorating inside the company and visible to no one outside it.

7.2 Dividend

AECOM paid no dividend before FY2022. Dividends per share: $0.45 (FY22), $0.69, $0.85, $1.04, and $1.24 annualised for FY2026 after a 19.2% increase to $0.31 declared on 18 November 2025 — the day before the stock fell 11%. The June 2026 declaration held it flat at $0.31. At $71.76 the yield is 1.73%, costing ~$159m a year against FY2025 free cash flow of $684.9m — a 23% payout. The dividend is comfortably covered; the buyback is what strained the balance sheet.

7.3 M&A and the disclosure gap

M&A was nearly dormant for four years (cash paid: $26.7m in FY22, nil in FY23, $18.7m in FY24) and then restarted abruptly with $212.5m of cash in FY2025. AECOM completed two acquisitions in FY2025 for total consideration of $375.9m, of which $146.4m was AECOM stock. Neither met the quantitative threshold for separate disclosure, so neither target is named and no revenue, EBITDA or multiple appears anywhere in the filings. The purchase price landed almost entirely in intangibles: goodwill +$225.4m and a new $178.4m intangible (variously labelled “intellectual property” and “digital assets”) amortised over five years, carrying zero accumulated amortisation at 30 September 2025 — meaning the deal closed at the very end of FY2025.

Trade and legal press reported on 25 November 2025 that AECOM had acquired the Norwegian AI start-up Consigli for approximately $390m (NOK 4bn); Consigli describes itself as an “autonomous engineer” automating MEP engineering, space analysis, unit optimisation and tender documentation, and claims to reduce engineering time by up to 90%. Its founder joined AECOM as head of AI engineering. The ~$390m press figure does not reconcile to the filing’s $375.9m of total consideration for two acquisitions, and per our source-priority rule the filing governs; the discrepancy is flagged as unresolved (the Open Questions section). Management referred on the Q1 FY2026 call to completing the integration of “our September acquisition,” having “already doubled the size of our team.”

A further detail: a contingent-consideration liability with a maximum value of $17.7m was issued for one of these acquisitions, and in H1 FY2026 AECOM recorded a $5.7m gain in other income from the decline in its fair value — i.e. the acquired business is tracking below its earnout targets, and that shortfall flows through as a credit to earnings. Together with a $7.3m investment fair-value gain, roughly $13.0m of H1 FY2026’s $18.5m of “other income” is non-cash mark-to-market.

The failed divestiture. On 18 November 2025 AECOM announced a strategic review of Construction Management including a possible sale, retaining Goldman Sachs and Wachtell Lipton, and said the business would be classified as held-for-sale and reported in discontinued operations from Q1. On 9 February 2026 — eleven weeks later — the review was complete and the business retained, with no bidder, no price and no explanation disclosed. A process that short, run by advisers that expensive, terminating with retention, is most consistent with bids that did not clear management’s reserve.

7.4 Incentives — the most damning item in the file

The FY2025 annual bonus (paid at 107.0% to all four named executives) was weighted: adjusted EBITDA 35% · segment adjusted operating margin on NSR 25% · days sales outstanding 20% · subjective KPIs 20%. There is no ROIC metric, no per-share metric, and no free-cash-flow metric.

That last omission is not incidental. Free cash flow carried a 30% weight in both FY2023 and FY2024, and was deleted for FY2025 — replaced by DSO. It was removed in the year immediately preceding the year in which free cash flow was guided down 42% and then collapsed 78% in the first half. We do not allege causation, and DSO is a defensible working-capital metric. But an incentive plan that dropped the cash measure the year before the cash broke is a fact that deserves to be on the table.

The long-term plan is better in form than in effect. PSUs (60% of long-term incentive) carry ROIC, adjusted EPS growth and relative total shareholder return at a third each — but the ROIC is company-defined, with a numerator that excludes intangible amortisation and all restructuring and a denominator that excludes held-for-sale businesses and freezes accumulated other comprehensive income, and it has paid at 200% of maximum in both completed cycles. Adjusted EPS growth explicitly counts buyback accretion, which means executives are paid for the repurchases documented in the Business Overview. The only genuinely binding metric is relative TSR — and its target was cut from the 55th to the 50th percentile, its comparator group switched from the E&C peer set to the S&P MidCap 400, and the FY2023 cycle paid 156.8% of target on a 46th-percentile TSR.

After the crash, pay went up. The grants made on 15 December 2025 — 26 days after the −11% session, struck at $97.96 — were 21–32% larger in units, implying FY2026 long-term incentive of roughly $12.0m for CEO Troy Rudd (+9.1%), $3.7m for CFO Gaurav Kapoor (+15.6%) and $4.0m for President Lara Poloni (+14.3%). Rudd’s FY2025 total compensation was $15,973,622 (including $96,193 of “membership dues”), a 215:1 pay ratio.

Governance corroborates the concern. The Chairman and CEO roles were combined effective February 2025 — eight months before the peak — after the 2024 proxy had described their separation as “best-in-class governance.” Dissent against governance-chair Bradley Buss has risen 4% (2024) → 15% (2025) → 21% (2026). A management severance-ratification proposal was defeated with 94% against in 2025. No new director has joined since November 2023. Insiders as a group own 0.457% of shares outstanding.

7.5 Insider transactions

Across the complete five-year Form 4 corpus — 109 filings, 253 individual transactions — there are exactly three code-P open-market purchases:

Against that, $62.12m of code-S open-market sales (665,812 shares at an average realised price of $93.30): Rudd $40.65m at $94.49; Tishman $5.73m; Poloni $5.08m; Kapoor $4.89m at $115.34; Gan $3.83m; Stotlar $1.51m. Dollars sold divided by dollars bought: 88.8x. There were zero director open-market purchases in five years. Rudd’s largest sale window was 15 August 2025 — 53,097 shares at $119–120, ten weeks before the peak. Gan and Poloni continued selling in December 2025 at $97–99, after the first crash, under 10b5-1 plans.

The three purchases are suspiciously round dollar amounts ($300k, $100k, $300k), which is the signature of an ownership-guideline or optics purchase rather than a conviction position. The CEO’s $300,060 is 0.7% of his realised sale proceeds.

Verdict on capital allocation: negative. Value-destroying buyback timing on nearly $800m; capital returns at 112% of free cash flow funded by debt; $375.9m of unnamed acquisitions with no disclosed multiple and an earnout already marked down; a failed divestiture run at speed by two of the most expensive advisers in the market; an incentive plan with no ROIC, no per-share and — from FY2025 — no cash metric; larger equity grants after the collapse; and 88.8x more insider selling than buying. The dividend policy is the one part of this that is sound.

8. Changes and Headwinds — Last Two Years

November 2024 – October 2025: the peak. AECOM beat its 17%+ long-term segment margin target five quarters early in Q3 FY2025, delivered record FY2025 results (NSR $7,573m +6%, adjusted EBITDA $1,203m +10%, adjusted EPS $5.26 +16%), and the shares reached an all-time high of $132.85 on 31 October 2025 — roughly 24.8x adjusted earnings, the richest multiple in a decade.

July 2025: refinancing. AECOM issued $1,200m of 6.000% senior notes due 2033 and redeemed the 5.125% notes due 2027, paying a $9.1m make-whole premium — raising the bond coupon 87.5bp to extend maturity by six years, and borrowing an incremental $191.3m that funded buybacks.

September 2025: the AI acquisition, and the worst buyback month. AECOM closed two acquisitions for $375.9m of total consideration ($146.4m in stock), creating a new $178.4m five-year “digital assets”/“intellectual property” intangible. Trade press subsequently reported this as the ~$390m purchase of Norwegian AI firm Consigli. In the same month the company repurchased 1,970,954 shares at an average of $126.91.

18 November 2025: the pivot the market rejected. In one release AECOM raised long-term targets (20%+ segment margin exit rate by FY2028, 15%+ adjusted EPS CAGR FY2026–FY2029), attributed them to proprietary AECOM AI and Advisory, announced a strategic review of Construction Management with Goldman Sachs and Wachtell Lipton, raised the dividend 19% to $0.31 — and guided FY2026 free cash flow to ~$400m against $685m delivered in FY2025, with the segment margin guided essentially flat at 16.6%. The shares fell 11.1%, then 6.1%, then 2.5%: −19% in three sessions, while peers barely moved. Four analysts cut price targets; Barclays downgraded to Equal-Weight and cut its target from $135 to $100, citing the “complexity” the AI-driven targets added; Baird moved to Neutral.

9 February 2026: the retreat, and a weak quarter dressed well. AECOM completed the Construction Management review in eleven weeks and retained the business — no buyer, no price, no explanation. It raised FY2026 adjusted EPS guidance to $5.85–$6.05, lifted the buyback authorisation to $1bn, and reported record backlog on a 1.5x book-to-burn. The release headline read “Strong performance exceeded expectations on all key financial metrics.” Against guidance, that was true. Against the prior year it was not: Q1 FY2026 revenue fell 5%, GAAP operating income fell 7%, GAAP net income fell 21% and adjusted net income fell 3%, with free cash flow of $42m. Americas design book-to-burn was 1.0x, hurt by a 43-day US federal government shutdown.

11–12 February 2026: the category de-rating. Three sessions after a well-received print, the entire engineering-design cohort was repriced on AI-disintermediation fear: ACM −12.4%, WSP −12.5%, Stantec −11.7%, Tetra Tech −11.6%, Jacobs −7.6%, AtkinsRéalis ~−10%. Nothing AECOM-specific occurred. Over the following months the fear broadened — Stantec −23.7% and WSP −23.1% over roughly six months against an S&P/TSX up 12.5%.

10 March 2026: refinancing again. Amendment No. 16 to the Syndicated Facility Agreement provided a new $1.5bn revolver and $950m Term Loan A (both to March 2031, a two-year extension) plus a $500m Term Loan B (April 2031), with a ≤4.00x consolidated leverage covenant.

11–12 May 2026: the cash quarter. AECOM raised guidance for a second consecutive quarter to $5.90–$6.10, reaffirmed the 20%+/15%+ targets, and reported record 16.5% segment margins and record $26.2bn backlog — alongside operating cash flow of $4m (−98%) and free cash flow of negative $27m, constant-currency NSR growth of 2% against a 6–8% organic guide, International NSR −3%, and an adjusted tax rate of 13.9%. The 10-Q filed the same day disclosed a ~$680m claims balance. The shares fell 12.0% while peers fell 2–3%.

20 May – 25 June 2026: litigation risk emerges. At least seven — by one count nine — plaintiffs’ firms announced securities-fraud investigations, all expressly citing the 11 May cash-flow disclosure. No complaint has been located as of the report date, and AECOM has filed no 8-K disclosing any of it. These are investigations, not filed cases, and the base rate for such announcements converting into material liability is low; but they belong on the risk register.

14 May and 16 June 2026: the first insider buying in five years. CEO Rudd ($300,060), CFO Kapoor ($100,990) and President Poloni ($298,341) made the only three open-market purchases in the entire five-year Form 4 corpus.

3 June 2026: the dividend was held flat at $0.31, after November’s 19% increase.

10 June 2026: an unexplained second revolver. AECOM entered a new, standalone $500m revolving credit facility maturing June 2028, secured by a lien on substantially all assets, three months after a refinancing that already provided $1.5bn of revolving capacity, and six weeks after the negative-free-cash-flow quarter. The 8-K states no purpose. It was undrawn at signing.

Through 2026: the Middle East deteriorated. Saudi Arabia is expected to spend ~$16bn cancelling NEOM contracts through 2030; three contracts worth $6.85bn were terminated in March 2026; The Line is deferred past 2030; the 2029 Asian Winter Games were postponed indefinitely in January 2026.

Verdict: these developments weaken the thesis, and the weakening is mostly self-inflicted. The category de-rating in February was exogenous and arguably unfair. Everything else — the 42% cash guide cut, the failed divestiture, the buyback at $126.91, the claims escalation, the widening add-backs, the flat dividend, the second secured revolver — is AECOM’s own record, and it is a record of a company whose reported earnings improved while its cash, its disclosure quality and its credibility did not. The one genuinely favourable development is legislative: the BUILD America 250 Act cleared committee 62–2, which makes a transportation funding cliff materially less likely than the price implies.

9. Risk Analysis

Catastrophic-loss risk: low. The business is asset-light, leverage is 1.3x, there is no maturity wall before FY2029, the revolver is essentially undrawn and the dividend is covered at a 23% payout. The dominant risk here is a permanent re-rating, not insolvency. Probability of total loss: negligible.

10. Valuation Discussion

10.1 The multiple stack at $71.76

Rebuilt from primary sources — 128,522,920 shares (Q2 FY2026 10-Q cover) at $71.76 gives a market capitalisation of $9,222.8m; net debt of $1,683–1,713m gives an enterprise value of roughly $10.9–11.1bn, or ~$11.6bn lease-adjusted. Neither ROIC.ai’s $10,874m market cap (struck at the 31 March fiscal close of ~$84.80) nor FactorsToday’s $8,732m (implying ~121.7m shares) was used.

Own-history percentiles (AZI, 30 July 2026): P/E 18.67x = 10.98th percentile; P/S 0.59x = 46.00th; P/B 4.08x = 62.51st; composite 39.83rd. Read the P/E and the P/S; ignore the P/B — book value contains $3.76bn of goodwill against negative tangible equity, so a 62nd-percentile price-to-book is a goodwill and leverage artefact, not a valuation signal.

10.2 Against its own history

Today’s EV/EBITDA is below the intra-year low print of every fiscal year since FY2018, and below every year-end print in a decade except FY2017. The adjusted P/E has compressed roughly 48% in ten months while adjusted EPS grew. Over the same period returns on invested capital roughly doubled (8.3% in FY2021 to 14.2% in FY2025). This is emphatically not a returns story.

10.3 Against its peers

(Prices 31 July 2026, yfinance — unofficial; ACM restated to its own rebuilt figures.)

ACM is the cheapest name in the design cohort on forward earnings and EV/EBITDA while earning the highest or second-highest return on invested capital in it. That inverts the usual relationship and is the single strongest quantitative argument for the bulls.

One normalisation deserves emphasis: on gross revenue ACM’s EV/sales of 0.70x looks dramatically cheaper than Stantec’s 1.53x or WSP’s 1.66x — but that is an artefact of AECOM’s 53% pass-through revenue. On the comparable net-revenue basis ACM trades at ~1.43–1.50x, essentially in line with Stantec and below WSP and Tetra Tech. So the market is not discounting AECOM’s revenue; it is discounting its earnings and its cash, by 30–45%. That is precisely consistent with the quality-of-earnings analysis in the Financial Quality section, and it tells you what the debate is actually about.

10.4 Embedded expectations

Perpetuity cross-check against the $9,222.8m market capitalisation at a 9% cost of equity:

At $71.76 the market pays for roughly nominal-GDP growth on a depressed cash number — and, if you accept management’s own 100%-conversion target, essentially zero real growth in perpetuity. The entire valuation gap is the credibility of cash conversion. It is not about growth, and it is not about margin.

Required-return decomposition — what FY2029 adjusted EPS must be for a 9% annual return from $71.76, off the FY2026E base of $6.00:

Read across the table: the market is underwriting one of two propositions, not both. Either management’s 15%+ adjusted EPS CAGR lands and the multiple is permanently re-based to ~10x; or the multiple part-recovers to a still-modest ~13x on roughly a third of the plan. Management’s plan delivered alongside anything like a normal multiple is worth considerably more than the current price.

10.5 Scenarios to FY2029

Common FY2026E base (company guidance, 11 May 2026): NSR ~$7,750m; segment adjusted operating margin 16.8%; G&A ~$155m; adjusted net interest ~$140m; adjusted tax 21–22%; average diluted shares 130m; adjusted EPS $6.00; FCF ~$400m.

The bull EPS exceeds management’s own 15%+ target; delivering the plan exactly gives $9.13, which at 14x is ~$128 (~$99 discounted).

A necessary normalisation caveat. These scenarios run on adjusted EPS, because that is the currency in which management guides and the market trades. On the evidence of the Financial Quality section that currency is overstated: stripping a normalised recurring restructuring charge (~$100m on the five-year average), the real economic amortisation of the acquired AI intangible (~$39m), and the recurring legacy discontinued-operations drag (~$82m after tax) reduces FY2026 earnings power from the guided $6.00 to roughly $4.50 of normalised owner earnings — 15.9x, not 12.0x. Applying a similar ~10–25% haircut across the scenarios moves the discounted values to roughly $35 (bear) / $69 (base) / $102 (bull). Both framings are shown because both are defensible; the difference between them is the investment debate.

On skew. The downside case requires three things to go wrong simultaneously — reauthorisation failure and realised AI fee deflation and margin reversal. The upside case requires only that growth holds at mid-single digits, that the margin expands roughly halfway to target, and that the multiple recovers to 12x — a level ACM traded at or above in every fiscal year from FY2018 to FY2025. The skew is modestly favourable. The load-bearing variable on both sides is identical: cash conversion.

What the market is probably pricing correctly: the H1 FY2026 cash miss is real and large; the September 2026 reauthorisation expiry and IIJA runoff are dated, calendarised risks; $3.76bn of goodwill against negative tangible book leaves no asset floor; and a design-hours business is a legitimate AI-deflation candidate.

What it may be pricing incorrectly: the ~$400m FY2026 FCF guide was set in November 2025 and explicitly carries one-off restructuring and AI investment — treating it as run-rate embeds a permanent 42% cut from FY2025; ACM has the cohort’s best returns on capital and its lowest multiple; the AI question is genuinely symmetric and no evidence exists either way; the transportation cliff looks more like a timing risk than a quantum risk on a 62–2 committee vote; and the factor model shows the marginal holder pricing ACM alongside municipal-infrastructure-products cyclicals rather than professional-services firms.

No price target and no recommendation is expressed in this section.

11. Variant Perception

Consensus, as revealed by the tape rather than by ratings. The adjusted P/E has compressed ~48% in ten months while adjusted earnings grew; the stock sits 17% below its 200-day exponential moving average with an alpha of −0.211. The market believes AECOM is a structurally sound but AI-threatened design consultancy whose adjusted earnings do not convert to cash, whose end markets face a September 2026 funding cliff, and whose record margins are therefore a peak.

Notably, the published sell-side has not capitulated: despite a 46% decline there is no Sell rating, and targets sit at RBC $105, Truist $107, Baird $96, KeyBanc $94 and Goldman $127 — every one 30–45% above spot. Consensus in ratings is offside long while consensus in price is deeply bearish. That divergence is itself informative: the marginal seller is not the analyst community.

The strongest bull case. The de-rating is a cash-timing scare inside a genuinely improving franchise. Backlog is at a record $26.2bn (+8%) after 22 consecutive quarters of book-to-burn above 1.0; the design pipeline is at a record and growing double digits; Americas design NSR grew 8% in constant currency at a 20.0% adjusted margin, an all-time quarterly high; FY2025 ROIC of 14.2% is best in the cohort and nearly double FY2021; guidance has been raised in each of the last two quarters; and roughly $500m a year of buyback is being applied to a 12x multiple. If H2 FY2026 delivers the ~$385m of free cash flow required to meet the guide, the single fact underpinning the bear case evaporates and the stock is trading at a decade-low multiple on rising returns.

The strongest bear case. The adjusted numbers are a construct. Operating cash flow fell 98% in a quarter in which “record” margins were reported. Contract assets rose $379m in six months while cash fell $551m. Unresolved claims have risen from $110m to $680m and are accelerating (+$80m, +$70m, +$80m, +$120m, +$160m over five quarters) — and management’s explanation, that this stems from “two projects bid in fiscal 2019 and 2020,” is difficult to reconcile with a balance of $110m at 30 September 2022. Adjusted EPS has exceeded GAAP by 81% over five years on charges taken in six consecutive years. Tangible book is −$1.67bn. The 16.5% margin is being earned in an industry the company’s own 10-K describes as having limited barriers to entry, and AECOM lost its ENR No. 1 rank this year. And the terminal risk — AI performing design at a fraction of the hours — attacks revenue directly in a business whose entire five-year growth came from price per hour. On this reading, a 20%-margin FY2028 target and a 15% FY2029 EPS CAGR are exactly what a company promises when the cash has stopped arriving.

The five assumptions that matter most.

1. Cash conversion. Does FY2026 free cash flow land near $400m, and does FY2027 return to ~100% of adjusted net income? Everything else is second-order.
2. AI: deflator or lever? Management asserts lever and has staked a 350bp margin bridge on it; the market has priced deflator. No evidence exists on either side beyond assertion and one candid CEO quote pointing at margin, not revenue.
3. Funding continuity. Reauthorisation passing or extending around September 2026; state and local budgets holding.
4. Margin durability. Is 16.5% a peak or a waypoint? Americas is already at 20.0%; International, at 11.1% and declining, is the swing factor.
5. Where the multiple rests. 10x / 12x / 14x on FY2029 adjusted EPS is roughly $60 / $100 / $146 undiscounted.

What would falsify the bull case. Q3 FY2026 — which prints on 10 August 2026, ten days after this report — failing to show a large cash catch-up, or the FY2026 free-cash-flow guide being cut. A second consecutive sub-100%-conversion quarter converts “timing” into “the model.” Equally: Americas design book-to-burn falling below 1.0, or any client-led design-fee repricing explicitly attributed to AI.

What would falsify the bear case. H2 FY2026 free cash flow of ~$385m or more, hitting the guide, with the claims balance falling rather than merely decelerating. A clean surface-transportation reauthorisation before or shortly after 30 September 2026. And an FY2027 guide showing the segment margin still climbing toward 20% on mid-single-digit-plus NSR growth.

The factor-positioning input. The disagreement here is fundamental, not positional. Roughly 72% of ACM’s return variance is idiosyncratic, and the stock carries no Momentum, Value, Quality or LowVolatility loading at all — meaning no systematic strategy has an axe in this name, and the selling cannot be explained as a factor unwind. What it does load on is Market (0.83), SmallSize (0.45), Industrials (0.45), DividendYield (0.18) and Infrastructure (0.15), with a negative Growth loading. Its factor neighbours are Core & Main, Federal Signal, Eagle Materials and water ETFs — not one design peer appears in the top twenty.

That inversion is the sharpest positioning insight available: the marginal holder prices AECOM as a mid-cap municipal-infrastructure-products cyclical, not as an asset-light professional-services firm with 14% returns on capital. It explains why the stock de-rates on infrastructure-budget fear rather than on services-sector news, and it means the “capital-light compounder” frame management uses is not the frame the market is applying. Meanwhile the 21-day regime has rotated hard into Value, LowVolatility and DividendYield — everything AECOM now is — while Industry: Infrastructure is the single worst-performing exposure in the model (−4.6% over 21 days, z −1.74; −9.8% over 252 days). The style is a tailwind; the sub-sector is the headwind, and the sub-sector is winning. (Regime observations are interpretation and mean-revert.)

12. Fact vs. Interpretation

13. Open Questions

1. What is actually in the $680m claims balance? Counterparty, contract, age and probability weighting are undisclosed. Management attributes it to “two projects bid in fiscal 2019 and 2020” — which is difficult to reconcile with a claims balance of $110m at 30 September 2022. This is the single most important undisclosed item in the file.
2. Can H2 FY2026 bridge $15m of first-half free cash flow to a ~$400m full-year guide? That requires ~$385m in two quarters. Q3 prints 10 August 2026.
3. Why was a second, secured $500m revolver added in June 2026, three months after a full refinancing that already provided $1.5bn of revolving capacity and six weeks after a negative-free-cash-flow quarter? The 8-K states no purpose.
4. What did the Construction Management sale process reveal? Goldman Sachs and Wachtell Lipton were retained in November 2025 and the business was retained eleven weeks later with no bidder, price or explanation disclosed.
5. What were the two FY2025 acquisitions, and at what multiple? Neither is named anywhere in the filings. The ~$390m press figure for Consigli does not reconcile to the filed $375.9m for two deals combined.
6. Will AECOM continue to add back digital-asset amortisation through FY2028, and is the 20%+ target stated on that basis? (On the evidence, yes to both.)
7. What is AECOM’s employee turnover? The company discloses none. In a business whose entire growth algorithm is revenue per head, attrition is a first-order metric.
8. Does the Senate act on surface transportation before 30 September 2026, or does the programme lapse into short-term extensions?
9. Why does the FY2025 10-K MD&A contain a self-contradictory International transportation sentence (−$67.9m/−5.2% and +$18.8m/+1.3% in the same passage), and why was FY2023 backlog restated down $1.6bn without explanation? These are disclosure-quality flags in their own right.
10. Did AECOM repurchase shares in the June 2026 quarter at $67–85? The Q3 10-Q was not filed at the report date. Two consecutive months of zero repurchases (January and March 2026) with a fully loaded authorisation is an unexplained pattern for a company that spent $250m in a single month at $126.91.
11. What is the sell-side consensus, as distinct from company guidance? No free source was available; every forward figure in this memo is guidance, so the guidance-versus-consensus gap is unmeasured.
12. What caused the 24–28 July 2026 group bounce (+10.3% over four sessions across ACM, J, TTEK and PSN)? No catalyst was identified; recorded as unattributed rather than guessed.

14. What Must Be True

The bull case

The bear case

The single decisive observation for both cases is the same: H2 FY2026 cash conversion, reported on 10 August 2026 and completed in November 2026. Everything else in this memo is second-order to it.

15. Source Appendix

The full source appendix — every primary filing, transcript, data feed and third-party source relied upon, with URLs and access dates — is presented as Appendix B to this article.

Principal primary sources: AECOM Forms 10-K for FY2021–FY2025 (CIK 0000868857); Forms 10-Q for FY2024–Q2 FY2026; the complete Form 8-K corpus since July 2021 (52 filings) including the earnings exhibits of 18 November 2025, 9 February 2026 and 11 May 2026; DEF 14A proxy statements for 2024, 2025 and 2026; and the complete Form 4 corpus (109 filings, 253 transactions).

Principal data sources: SEC EDGAR XBRL company facts; AZI daily price history and own-history valuation percentiles; ROIC.ai for ratios, multiples and earnings-call transcripts; FactorsToday for factor loadings, risk-adjusted returns and factor-similarity peers; yfinance for peer market data (unofficial — reconciled where material).

Sections 1–15 contain no investment recommendation and no price target; the Claude's Take block at the head of this article is the author’s own subjective opinion, is clearly labelled as such, and is general information rather than investment advice.

APPENDIX A — Standard Diligence Questionnaire

AECOM (NYSE: ACM) — 31 July 2026

Supplemental to the analysis above. Answers are labelled FACT / INTERPRETATION / ASSUMPTION where it matters.

General

What thoughtful questions have other investors asked about this company?

The Q&A on the last three earnings calls converged on one question, asked three different ways: does AI shrink AECOM’s revenue? Citigroup’s Andrew Kaplowitz put it most directly on the Q1 FY2026 call — “does a new value model shaped by AI not lead to shrinking revenue for AECOM? And… that EBITDA to employee calculation… is up, I think, 50% over the last 5 years. Does that rate of improvement now shift up substantially?” Baird’s Andrew Wittmann pressed the cash-flow and claims question directly on the Q2 call. The second recurring question is whether the 20%+ FY2028 margin target is credible when FY2026 was guided essentially flat at 16.6–16.8%. The third, largely unasked by the sell-side and in our view the most important, is what is actually inside the $680m claims balance.

Notably, the published sell-side has not capitulated: there is no Sell rating after a 46% decline, with targets at RBC $105, Truist $107, Baird $96, KeyBanc $94 and Goldman $127 — all 30–45% above spot. (FACT.) The thoughtful questions are being asked on the calls, not reflected in the ratings.

Cyclicality & Earnings Nature

Are earnings at a cyclical high or low? A high, on the reported measures. FY2025 delivered a record segment adjusted operating margin (16.5%), record adjusted EBITDA ($1,203m), record adjusted EPS ($5.26) and record backlog, on the back of peak IIJA-funded activity. Americas segment margin reached an all-time 20.0% in Q2 FY2026. (FACT.) The counter-argument is that cash earnings are at a low, not a high — H1 FY2026 free cash flow was $14.5m against a five-year run-rate of ~$600m — so the answer depends on which earnings you mean. (INTERPRETATION.)

Driven by the external environment or internal actions? Both, in roughly equal measure. Externally: the IIJA, UK AMP8, Australian and Canadian infrastructure programmes, and (until 2025) Middle East mega-projects. Internally, and genuinely to management’s credit: the 2020–21 exit from self-perform construction removed the loss-making, capital-consuming part of the business; offshore capability centres in India and Poland lowered delivery cost; and G&A was held flat in dollars ($155.1m FY2021 → $157.8m FY2025) against 26% NSR growth. (FACT.) Roughly a third of the ROIC improvement, however, is a shrinking capital base from buybacks rather than operating gain. (INTERPRETATION.)

How stable are revenues? Moderately. Revenue is project-based and re-competed, not contractual annuity. Book-to-burn has exceeded 1.0 for 22 consecutive quarters, which is genuine evidence of stability, but only $13.8bn of $26.2bn of backlog is contracted and the audited ASC 606 performance obligation grew just 5% in four years. No client exceeds 10% of revenue. Half of revenue comes from governments, which is stable but slow and price-competitive. (FACT.)

Outlook for products/services? The funded demand is real: water and PFAS (AMP8’s £104bn, EPA SRF allotments of $7.2bn), environment, defence (a $1.5tn US budget proposal; AECOM’s Department of War pipeline +50%, its largest single client), transportation (contingent on reauthorisation), and the higher-margin advisory and programme-management lines, with advisory targeted to double NSR in three years. Nuclear fusion work (UK STEP; Type One Energy/TVA) is described by management as “9 figures of NSR in the coming years.” (FACT — management commentary, unvalidated.)

How big will this market be — growing, shrinking, domestic or international? The transparent proxy is ENR’s Top 500 Design Firms aggregate: $158.7bn of 2025 design revenue, +7.4% (US $136.3bn +8.1%; international $22.4bn +3.2%). Growing modestly, and disproportionately domestic. AECOM’s global NSR is ~4.8% of that pool. (FACT.) International is the weaker half for AECOM specifically: Q2 FY2026 International NSR fell 3% in constant currency, and the Middle East is contracting hard (~$16bn of NEOM contract cancellations expected through 2030).

Business Quality & Competitive Moat

Is the industry getting more or less competitive? More. The Big Four’s share of US environmental & sustainability consulting fell 2.0 points in 2024, and consolidation has plateaued since 2022. Private-equity roll-ups are actively adding capacity at the $1.5–8m EBITDA level. AECOM lost the ENR No. 1 rank to Jacobs in 2026. And the AI supply shock, if real, expands effective engineering capacity with near-zero capital and near-zero lag. (FACT for the share data; INTERPRETATION for the AI conclusion.)

How profitable is the business (ROIC, ROE)? FY2025 ROIC 14.2% and ROE 24.0% on a consistent third-party basis — the best in the cohort (Tetra Tech 14.1%, Stantec 10.1%, Jacobs 6.6%). But the seven-year average ROIC is ~8.3% (FY19 4.96%, FY20 4.57%, FY21 8.30%, FY22 8.83%, FY23 4.58%, FY24 12.49%, FY25 14.24%) — roughly the cost of capital. The ROE is ten points above Tetra Tech’s on a near-identical ROIC because tangible equity is negative $1,669.5m. (FACT.) Two good years do not establish a franchise return. (INTERPRETATION.)

How profitable is the industry — how many competitors, what barriers to entry? Listed peers earn 14–18% adjusted EBITDA on net revenue; employee-owned Mott MacDonald and foundation-owned Ramboll earn 4–5% on gross revenue because they optimise for compensation rather than return on capital. On barriers, AECOM answers the question itself in its 10-K: “The technical and professional aspects of our services generally do not require large upfront capital expenditures and, therefore, provide limited barriers against new competitors.” (FACT — direct quote.)

Can the business be easily understood? Yes. It sells engineering hours to infrastructure owners. The two genuine complications are (a) that 53% of GAAP revenue is zero-margin pass-through, so net service revenue is the only meaningful top line, and (b) a persistent and widening gap between GAAP and adjusted earnings.

Can it be undermined by foreign low-cost labour? It already has been, and AECOM is on the right side of it: the India and Poland capability centres are a principal source of the five-year margin expansion. The risk runs the other way — competitors have the same option, and the offshore cost advantage is therefore competitive parity among the global majors rather than an advantage. (INTERPRETATION.)

Do brands matter? Modestly. The ENR ranking, past-performance references and technical reputation help pre-qualify AECOM for the largest programmes. They do not command a price premium: AECOM is the largest firm and the third-most profitable of four.

What is the nature of competition? Qualifications-based selection followed by price, on individually-competed pursuits with typically 3–8 qualified bidders. Public agencies re-compete on a schedule and are legally obliged to compete. (FACT.)

Customers’ switching costs? Real at the project level — replacing a designer mid-programme is genuinely disruptive — and weak at the client level, which is where captivity would need to bind. Management claims a re-compete win rate above 90%, but this is unaudited with no disclosed denominator. The 10-K is explicit that government clients “can terminate, renegotiate, or modify any of their contracts… at their convenience” and that backlog “excludes contract renewals or extensions that are at the discretion of the client.” (FACT.)

Financial Condition & Balance Sheet

Assets not fully recognised on the balance sheet? The workforce, the client relationships and the design-data libraries — none capitalised, and collectively the entire earning power of the firm. This is normal for the model and is why book value is uninformative here.

Off-balance-sheet liabilities? Yes, and they matter: $903.8m of standby letters of credit (insurance programmes and contract performance guarantees); $628.5m of operating lease liabilities; $319.4m of sold trade receivables (a factoring programme that was increased $51.2m in H1 FY2026 while reported cash flow collapsed — treat as debt-like); surety bonds; and a $30m revolving-credit commitment to the buyer of the divested civil-infrastructure business, $25.6m drawn at 31 March 2026. Adding leases and sold receivables takes economic net debt to ~$2,661m, or ~2.5x. (FACT.)

How conservative is the accounting? Not conservative, and deteriorating. The load-bearing item is $680m of unresolved claims recognised in contract assets and other non-current assets — revenue booked against customers who have not agreed to pay — up from ~$180m eighteen months earlier and equal to 30% of book equity. Adjusted EPS has exceeded GAAP by 81% over five years on charges recurring in six consecutive years. Adjusted segment margin adds back the amortisation of the AI asset the margin is meant to be driven by. Management restated its own adjusted FY2022 EPS from $3.47 to $3.40 when it introduced a new add-back category. Two coexisting backlog definitions are published ($39.7bn versus $24.8bn at the same date), FY2023 backlog was restated down $1.6bn without explanation, and the FY2025 10-K contains a self-contradictory International transportation passage. (FACT.)

How CapEx-hungry is the business? Barely at all — capex is 0.7–0.9% of revenue ($136.7m on $16.1bn in FY2025). The capital intensity is in working capital, not fixed assets, and that is exactly where the problem now sits: DSO went from 72.2 days (FY2025) to 83.8 days (March 2026).

Capital Allocation & Management

How much FCF does the business generate, how does management use it, what is the philosophy? Free cash flow ran a stable $568m–$708m (8.8–9.8% of NSR) for five years, then collapsed to $14.5m in H1 FY2026 against a ~$400m full-year guide. Cumulatively FY2021–H1 FY2026: $3,143m generated; $3,028m spent on buybacks plus ~$488m of dividends — 112% of free cash flow, with the gap funded by a $707m increase in net debt. The stated philosophy is a “returns-focused capital allocation policy.” (FACT.)

Significant acquisitions recently? Two in FY2025 for $375.9m of total consideration including $146.4m of stock. Neither is named in any filing, neither met the disclosure-significance threshold, and no multiple is disclosed anywhere. The price landed almost entirely in goodwill (+$225.4m) and a new $178.4m five-year “digital assets”/“intellectual property” intangible. Trade press reported this as the ~$390m acquisition of Norwegian AI firm Consigli, which does not reconcile to the filed figure. The $17.7m earnout is already being marked down, producing a $5.7m gain in H1 FY2026 other income — i.e. the acquired business is missing its targets, and the shortfall flatters earnings. (FACT, with the press figure flagged as unreconciled.)

Buying back shares? Aggressively, and badly timed. $3,028m spent to retire a net 15.0m shares — $201.88 of cash per net share retired against a $71.76 share price. In September 2025 alone, 1,970,954 shares at an average of $126.91 ($250.1m), weeks before the all-time high. The $790.4m deployed across FY2025–H1 FY2026 at an average of $108.96 is worth $520.5m today: −$269.9m, or ~2.9% of the entire market capitalisation. Zero shares were repurchased in January and March 2026 with $883.9m authorised. (FACT.)

Issuing large amounts of new shares to insiders? Stock-based compensation was $252m over FY2021–FY2025 — modest at ~0.8% of NSR — but combined with $146.4m of acquisition stock it absorbed most of the buyback, which is why $3.0bn retired only 15.0m shares.

Compensation policy of directors/management? The most troubling part of the file. The FY2025 annual bonus weighted adjusted EBITDA 35%, segment margin 25%, DSO 20% and subjective KPIs 20% — no ROIC, no per-share metric and no free-cash-flow metric. Free cash flow had carried a 30% weight in both FY2023 and FY2024 and was deleted for FY2025, the year before free cash flow was guided down 42% and then collapsed. PSU ROIC is company-defined (excluding intangible amortisation and all restructuring) and has paid at 200% of maximum in both completed cycles; adjusted EPS growth explicitly counts buyback accretion. The relative-TSR target was cut from the 55th to the 50th percentile and its comparator switched from E&C peers to the S&P MidCap 400; the FY2023 cycle paid 156.8% of target on a 46th-percentile TSR. Grants made 26 days after the November crash were 21–32% larger in units. CEO Rudd’s FY2025 total compensation was $15,973,622 (215:1 pay ratio). (FACT.)

Motivations of management? The revealed preference is adjusted-EPS growth and margin optics. Insiders own 0.457% of the company. Across five years there were three open-market purchases totalling $699,390 against $62.12m of sales at an average of $93.30 — an 88.8x ratio, with zero director purchases. The three purchases came only after a 46% decline, in round dollar amounts consistent with ownership-guideline compliance. Governance corroborates: Chairman and CEO roles were combined in February 2025 after the 2024 proxy called their separation “best-in-class governance”; dissent against the governance chair has risen 4% → 15% → 21%; a management severance-ratification proposal was defeated 94%-against in 2025; and no new director has joined since November 2023. (FACT.)

Valuation & Market Data

Is the stock an ADR, MLP, or K-1 issuer? No. AECOM is a Delaware corporation, NYSE-listed, issuing a standard Form 1099. No ADR, MLP or K-1 complications.

Dividend policy? Initiated FY2022. DPS $0.45 → $0.69 → $0.85 → $1.04 → $1.24 annualised, a ~20% CAGR since initiation. The quarterly dividend was raised 19% to $0.31 on 18 November 2025 and held flat at the June 2026 declaration. At $71.76 the yield is 1.73%, costing ~$159m against FY2025 free cash flow of $684.9m — a 23% payout. Comfortably covered; the dividend is the sound part of the capital-allocation record. (FACT.)

How profitable is the business? See above: 16.5% segment adjusted operating margin on NSR (16.1% gross margin on NSR; 7.5% on gross revenue); 14.2% FY2025 ROIC against a ~8.3% seven-year average.

Is net income diverging from cash from operations? Yes — severely, and this is the central finding of the engagement. Operating cash flow to net income ran 4.07x, 2.30x, 12.59x, 2.06x and 1.46x across FY2021–FY2025 — a steady, unbroken deterioration — and then broke: H1 FY2026 operating cash flow of $74.0m against $353.9m of income from continuing operations, a ratio of 0.21x. In Q2 FY2026 specifically, adjusted EPS rose 27% and GAAP EPS rose 22% while GAAP operating income fell 4% and operating cash flow fell 98%. The reconciling item is the claims and contract-asset build. (FACT.)

Risks & Downside

What factors would cause the stock to decline? In descending order of probability-weighted impact: (1) H2 FY2026 failing to deliver the ~$385m of free cash flow required to meet guidance, or a cut to that guidance — reported 10 August 2026; (2) a further escalation in the claims balance above ~$750m; (3) realised evidence of AI-driven fee deflation, whether from AECOM or a peer; (4) a surface-transportation reauthorisation lapse freezing state DOT new-start programming; (5) margin reversal from the record 16.5%, most plausibly via International, already at 11.5% and declining; (6) further Middle East deterioration; (7) a securities-fraud complaint actually being filed.

Risk of a catastrophic loss? Low. The business is asset-light with capex under 1% of revenue, net leverage of 1.3x (1.7x lease-adjusted), no maturity wall before FY2029, a $1.5bn revolver essentially undrawn plus a new $500m secured facility, and a dividend covered at a 23% payout. The realistic bear case is a permanent re-rating to roughly $51 undiscounted (the scenario analysis in the Valuation section), not distress.

Chance of a total loss? Negligible. There is no plausible path to zero for a profitable, cash-generative, 51,000-person firm with a record order book and 1.3x leverage. The genuine asymmetry to watch is that tangible book equity is negative $1.67bn, so there is no asset floor beneath the equity — the entire value is the going concern. That raises the severity of a permanent earnings impairment; it does not create insolvency risk at current leverage.

Recent News & Events

Has the business environment changed recently? Yes, in three distinct ways. (1) The market’s frame changed on 12 February 2026, when the entire engineering-design cohort was re-rated on AI-disintermediation fear (ACM −12.4%, WSP −12.5%, Stantec −11.7%, Tetra Tech −11.6%, Jacobs −7.6%) with no AECOM-specific disclosure. (2) The Middle East deteriorated materially — Saudi Arabia is expected to spend ~$16bn cancelling NEOM contracts through 2030; $6.85bn of terminations occurred in March 2026; The Line is deferred past 2030. (3) US transportation funding reached a dated cliff — IIJA authorisation expires 30 September 2026, with the House’s $580bn BUILD America 250 Act having cleared committee 62–2 on 22 May 2026 but not yet enacted. A 43-day federal government shutdown in Q1 FY2026 also delayed awards. (FACT.)

Significant acquisitions? Covered above: two unnamed FY2025 acquisitions for $375.9m, understood from press reporting to include the AI firm Consigli.

Change in accounting policies? No change in accounting policy, but three presentational changes that matter: (a) Construction Management was slated for held-for-sale/discontinued-operations treatment in Q1 FY2026 and then reversed when the sale was abandoned, leaving the FY2026 NSR guidance never restated to a comparable basis; (b) a new “digital assets”/“intellectual property” intangible class with a five-year life appeared, whose amortisation is added back to adjusted margin; © adjusted FY2022 EPS was retroactively restated from $3.47 to $3.40 when “non-core AECOM Capital” was introduced as an add-back category. (FACT.)

Recent changes — new markets, facilities, management? Headquarters are in Dallas, Texas (relocated from Los Angeles). Chairman and CEO roles were combined under Troy Rudd effective February 2025. Consigli’s founder joined as head of AI engineering, and management stated the AI team had “already doubled” in size. New market entries of note: Scottish Water (preferred bidder on the largest contract that client has ever let, from a standing start), nuclear fusion (UK STEP; Type One Energy/TVA), Brisbane 2032 Olympics Delivery Partner, and AUKUS-related defence work in Australia. On the debt side, three financing actions in twelve months: the July 2025 bond refinancing, Amendment No. 16 in March 2026, and a new $500m secured revolver in June 2026 whose purpose is undisclosed. (FACT.)

Anything else material? Between 20 May and 25 June 2026, at least seven plaintiffs’ firms announced securities-fraud investigations, all expressly citing the 11 May 2026 cash-flow disclosure. No complaint has been located and AECOM has filed no 8-K regarding any of it. These are announcements of investigations, not filed litigation, and the base rate for such announcements producing material liability is low — but they belong on the risk register. Q3 FY2026 results are due after the close on 10 August 2026, ten days after this report, and are the single most informative near-term event for the thesis. (FACT.)

APPENDIX B — Source Appendix

AECOM (NYSE: ACM) — 31 July 2026

All sources accessed 31 July 2026 unless otherwise noted. Primary sources are listed first.

1. SEC filings — primary (CIK 0000868857)

The full trailing-60-month corpus was enumerated and downloaded in full from EDGAR: 253 filings since 1 July 2021, of which 92 documents were saved (5 Forms 10-K, 15 Forms 10-Q, 52 Forms 8-K, 5 DEF 14A, 5 DEFA14A, 1 PRE 14A, 5 Forms 11-K, 4 ARS), plus 116 insider filings enumerated. Form breakdown by count: Form 4 ×109, 8-K ×52, SC 13G/A ×17, 10-Q ×15, 144 ×14, 10-K ×5, DEF 14A ×5, Form 5 ×3, Form 3 ×3.

Annual reports (Form 10-K)

Quarterly reports (Form 10-Q) — principal

• Q2 FY2026, period ended 2026-03-31, filed 2026-05-12 — acm-20260403.htm. The load-bearing filing: balance sheet (cash $1,034.3m, total debt, goodwill $3,762.3m, equity $2,270.6m), cover share count 128,522,920 at 2026-05-08, H1 cash-flow statement, Note 3 (discontinued operations; digital-assets intangible table and amortisation schedule; DOE $61.8m and Montana $53.0m legacy charges), Note 4 (~$680m claims; sold receivables $319.4m), Note 7 (contingent consideration), Part II Item 2 (repurchases). https://www.sec.gov/Archives/edgar/data/868857/000086885726000014/acm-20260403.htm
• Q1 FY2026, period ended 2025-12-31, filed 2026-02-10 — acm-20260102.htm. https://www.sec.gov/Archives/edgar/data/868857/000086885726000011/acm-20260102.htm
• Forms 10-Q for FY2024 and FY2025 quarters (2025-02-04, 2025-05-06, 2025-08-05, 2024-02-07, 2024-05-07, 2024-08-06) — used for the claims time series and quarterly repurchase tables.

Current reports (Form 8-K) — principal

Proxy statements (DEF 14A)

• 2026-01-20 — acm-20260119.htm: FY2025 annual-incentive metric weightings (adjusted EBITDA 35% / segment margin 25% / DSO 20% / subjective 20%), PSU design, relative-TSR comparator change, December 2025 grant sizes, CEO compensation ($15,973,622), pay ratio (215:1), beneficial ownership (insiders 0.457%).
• DEF 14A 2025 and 2024 — used to establish that free cash flow carried a 30% annual-bonus weight in FY2023 and FY2024 and was deleted for FY2025; the 2024 proxy’s description of Chairman/CEO separation as “best-in-class governance”; and the 2025 severance-ratification proposal defeated 94%-against.

Insider filings (Forms 3/4/5)

The complete Form 4 corpus — 109 filings / 253 individual transactions, 31 July 2021 to 16 June 2026 — was parsed from raw XML (note: EDGAR’s filing index returns the XSL-rendered path /xslF345X06/ownership.xml; the XSL segment must be stripped to reach the raw XML). Findings: three code-P open-market purchases totalling $699,390 (Rudd 2026-05-14; Kapoor 2026-05-14; Poloni 2026-06-16); 665,812 shares of code-S sales for $62.12m at an average $93.30; 35 of 37 sales under Rule 10b5-1 plans; zero director open-market purchases. Example raw filing: https://www.sec.gov/Archives/edgar/data/868857/000172309926000002/wk-form4_1781626413.xml

2. Structured data feeds

Rebuilt capital structure (not taken from any aggregator): 128,522,920 shares (Q2 FY2026 10-Q cover, 2026-05-08) × $71.76 = market capitalisation $9,222.8m; balance-sheet debt $2,717.2–2,747.7m less cash $1,034.3m = net debt $1,683.0–1,713.4m; enterprise value ~$10.9–11.1bn, or ~$11.6bn lease-adjusted.

3. Earnings-call transcripts

Retrieved from ROIC.ai; key quotes below:

• Q2 FY2026, call 2026-05-12 — Rudd on AI-enabled wins (“It’s not necessarily that we’re going to see more revenue from these particular contracts. We will see improved margins on those contracts.”); Kapoor quantifying AI spend ($13m in Q2, ~66bp of NSR) and the Scottish Water “pain share” mechanism; the Middle East collections explanation.
• Q1 FY2026, call 2026-02-10 — Kaplowitz (Citigroup) on AI and revenue; Rudd’s response; Kapoor’s gross-to-net margin bridge (“90 to 100 bps… once you net the investments… expect 30 bps”); reference to completing integration of “our September acquisition.”
• Q3 FY2025, call 2025-08-05 — baseline; IIJA spend-to-date commentary.
• NOT RETRIEVED: the Q4 FY2025 call (November 2025) — not carried by ROIC.ai and not sourced elsewhere. The 18 November 2025 press releases were read directly instead.

Methodology note: ROIC.ai’s call-listing endpoint does not filter reliably by ticker; the transcript endpoint resolves correctly using the exchange-qualified symbol.

4. Industry, regulatory and market sources

• ENR 2026 Top 500 Design Firms — $158.7bn aggregate 2025 design revenue (+7.4%); US $136.3bn (+8.1%), international $22.4bn (+3.2%). https://www.enr.com/articles/62878-enr-2026-top-500-design-firms-ai-boom-buoys-design-revenue
• Jacobs press release, “Jacobs ranks No.1 on ENR’s 2026 Top 500 Design Firms list” — confirming Jacobs took No. 1 and AECOM slid to No. 2, on a data-centre business up 62.2% with a pipeline up 500%. https://www.jacobs.com/newsroom/news/jacobs-ranks-no1-enrs-2026-top-500-design-firms-list
• Environment Analyst — US environmental & sustainability consulting shares; Big Four just over 40%, down 2.0pp in 2024. https://environment-analyst.com/global/111519/
• Morningstar, “AI Disintermediation Fears Drag Down Engineering and Construction Stocks” (company report 1431283) — the 12 February 2026 category de-rating: Jacobs ~−8%, AtkinsRéalis ~−10%, AECOM ~−12%. Direct fetch returned HTTP 403; headline and figures obtained via the search index. https://www.morningstar.com/company-reports/1431283-
• The Globe and Mail, “AI fears sink Stantec and WSP shares…”, 18 May 2026 — Stantec −23.7%, WSP −23.1% versus S&P/TSX +12.5%; Ian Gillies (Stifel) on the lost “pedestal”; Alexandre L’Heureux (WSP) on capacity constraint and 1%/yr engineer supply growth. https://www.theglobeandmail.com/business/article-ai-fears-sink-stantec-and-wsp-shares-turning-the-high-flying/
• CRS, Surface Transportation Reauthorization — R48644 and R48845. https://www.congress.gov/crs-product/R48845
• H.R. 8870, BUILD America 250 Act — $580bn over five years, $474.4bn HTF-guaranteed, approved by House T&I 62–2 on 22 May 2026. https://www.congress.gov/bill/119th-congress/house-bill/8870 · House T&I release · Holland & Knight, “A Closer Look at the BUILD America 250 Act,” May 2026. https://www.hklaw.com/en/insights/publications/2026/05/a-closer-look-at-the-build-america-250-act
• Semafor, “Saudi’s NEOM faces $16 billion bill to cancel NEOM contracts,” 7 June 2026. https://www.semafor.com/article/06/07/2026/saudis-neom-faces-16-billion-bill-to-cancel-neom-contracts
• New Civil Engineer, “‘Time is running out’: Contractors and consultants warn of slow start to AMP8,” 11 November 2025. https://www.newcivilengineer.com/latest/time-is-running-out-contractors-and-consultants-warn-of-slow-start-to-amp8-11-11-2025/
• Peer results releases — WSP Q4-2025 (2026-02-25); Stantec FY2025 (2026-02-26); AtkinsRéalis FY2025 (2026-02-27); Arcadis FY2025; Ramboll 2025; Mott MacDonald 2025; Parsons FY2025; Tetra Tech FY2025.
• Capstone Partners, AEC Services Sector M&A Update; Auxo Capital Advisors, PE Roll-Ups in AEC 2026 Guide — private-equity capacity additions.
• AECOM acquisition of Consigli — AEC Magazine, “Aecom acquires AI start-up Consigli”; Verdantix, “AECOM Takes A Bold Swing At AI Transformation With Consigli Acquisition”; Mogin Law LLP / Lexology, “AECOM’s $390 Million Bet on Artificial Intelligence.” All secondary; the ~$390m figure does not reconcile to the filed $375.9m and is flagged as unresolved.
• Analyst actions (secondary, via StreetInsider / GuruFocus / Investing.com summaries, not primary research notes): Barclays (Adam Seiden) Overweight → Equal-Weight, target $135 → $100, later $110 → $90; Baird (Andrew Wittmann) Buy → Neutral, target to $96; RBC $142 → $111, later $105; Truist $116 → $109, later $107; KeyBanc $94; Goldman Sachs $127.
• Securities-litigation investigation announcements (Businesswire, 20 May – 25 June 2026): Johnson Fistel (5/20), Kirby McInerney (5/21), Law Offices of Frank R. Cruz (5/26), Glancy Prongay & Murray (5/27), Howard G. Smith (5/27), Bragar Eagel & Squire (5/28), Pomerantz (6/09–6/25), Schall (6/11). Announcements of investigations only; no complaint was located and no 8-K has been filed.

5. Data-quality flags and unresolved items

1. The ~$390m Consigli press figure does not reconcile to the filing’s $375.9m of total consideration for two FY2025 acquisitions combined. The filing governs.
2. AECOM publishes two coexisting backlog definitions — $39.7bn (10-K) versus $24.8bn (press release) at the same date — and restated FY2023 backlog downward by $1.6bn in the FY2024 10-K without explanation.
3. The FY2025 10-K MD&A contains a self-contradictory International transportation passage (−$67.9m/−5.2% and +$18.8m/+1.3% in the same discussion).
4. AECOM does not disclose end-market revenue in dollars. The Americas facilities base of ~$8.0bn cited in the Business Overview is back-solved from MD&A dollar/percentage pairs and is assumption-grade.
5. Management’s “adjusted EBITDA margin” denominator does not reconcile to reported NSR (Q2 FY2026: $312m ÷ $1,948m = 16.02% against a stated 16.5%).
6. Management’s two IIJA spend-to-date statements are inconsistent — “only 36% spent” (Q3 FY2025) versus “more than half remains to be spent” (Q2 FY2026).
7. No employee turnover or attrition disclosure was located, in a business whose entire growth algorithm is revenue per head.
8. No free sell-side consensus source was available; every forward figure in this memo is company guidance, so the guidance-versus-consensus gap is unmeasured.
9. The FY2026 NSR base is muddied by the Construction Management reversal: the original $7.2–7.4bn guidance was on an ex-CM basis and was never restated after the business was retained. The ~$7.75bn FY2026E enterprise NSR used in the Valuation section is an assumption, built up from H1 actuals of $3,799m.
10. The catalyst for the 24–28 July 2026 group bounce (+10.3% across ACM, J, TTEK and PSN over four sessions) was not identified and is recorded as unattributed rather than guessed.
11. Q3 FY2026 (April–June) repurchase activity is not yet disclosed; the 10-Q had not been filed at the report date. Results are due after the close on 10 August 2026.