Autodesk (ADSK): a powerful franchise priced for mediocrity

The Gemini Brief - Investment Deep Dives
The Gemini Brief – Investment Deep Dives
Autodesk (ADSK): a powerful franchise priced for mediocrity
Loading
/

Slide Deck

Autodesk is a rare business that genuinely earns its moat. Its dominance in AEC design software, anchored by Revit’s proprietary file format lock-in and 40 years of workflow entrenchment, produces 91% gross margins, 97% recurring revenue, and $2.4 billion in free cash flow — financial characteristics consistent with a business possessing durable competitive advantage. FY2026 revenue reached $7.21 billion (+17.5%), the best growth year since the subscription transition era, driven by accelerating BIM adoption and successful execution of the new direct-billing model. The stock, however, trades at ~$238 (~21x trailing FCF, ~19x forward non-GAAP P/E) — near multi-year valuation lows — implying only 7.5–8% annual FCF growth for the next decade, far below the 12–16% FCF growth management is guiding for FY2027 alone.

The gap between Autodesk’s business quality and its current market valuation demands explanation. The answer lies in a cluster of governance incidents (the 2024 accounting investigation, CFO removal, Starboard activist campaign), macro anxiety about construction cyclicality, and legitimate questions about whether AI disrupts or reinforces the company’s franchise. This report dissects each of these factors with the rigor they deserve.


What Autodesk sells and why customers can’t leave

Autodesk operates a software oligopoly serving three verticals that design and build the physical world. Its FY2026 revenue breaks down as follows:

Product familyFY2026 revenue% of totalYoY growth
AECO (Architecture, Engineering, Construction & Operations)$3,583M49.7%+22%
AutoCAD / AutoCAD LT$1,787M24.8%+14%
Manufacturing$1,379M19.1%+16%
Media & Entertainment$332M4.6%+5%
Other$125M1.7%+6%

The mix has shifted meaningfully over five years. AECO has grown from ~43% to nearly 50% of revenue, reflecting the single most powerful secular tailwind in the business: global government BIM mandates. The UK has required BIM Level 2 on public projects since 2016, with over 70% of UK professionals now using BIM. Italy, Germany, Singapore, and Malaysia have all implemented or expanded mandates. India is developing national BIM policy. These are not reversible trends — they are regulatory ratchets that permanently expand Autodesk’s addressable market.

The competitive landscape is a textbook oligopoly. The global design software market is controlled by four firms: Autodesk, Dassault Systèmes, Siemens Digital Industries Software, and PTC, which collectively command roughly 65% of the CAD market. But market structure varies sharply by vertical:

  • AEC (Autodesk’s strongest position): Revit is the dominant BIM platform in North America and much of Europe. Autodesk holds an estimated 35–45%+ share in core AEC design. Competitors Bentley Systems (infrastructure), Nemetschek/ArchiCAD (Europe), and Trimble (field technology) are significant but operate in adjacent niches rather than head-to-head against Revit’s BIM authoring workflow.
  • Manufacturing (contested): Autodesk’s Fusion and Inventor compete in the mid-market but face formidable enterprise competitors in Dassault’s CATIA/SolidWorks and Siemens NX/Teamcenter. Autodesk is not the dominant player here.
  • Media & Entertainment (structurally challenged): Maya remains the VFX industry standard, but at only $332M in revenue growing 5%, this segment is small and faces relentless pressure from Blender (free, open-source) and real-time engines like Unreal.

The competitive advantage evidence is clearest in AEC. Revit’s .rvt file format is the single most important moat asset. Unlike AutoCAD’s .dwg format (which now has credible compatible alternatives like BricsCAD), .rvt is fully proprietary with no functional open-source reader. Years of Revit families, templates, parametric models, and multi-discipline coordination workflows create switching costs that are measured not in dollars but in months of lost productivity. Industry estimates suggest firms lose 4–8 weeks of billable time during a full platform migration. Multiple large-firm architects and engineers report that government projects and major developers contractually require Revit-format deliverables, creating regulatory lock-in on top of technical lock-in.

Almost every real-world switching example involves AutoCAD (2D drafting), not Revit. UK surveyor THS Concepts switched to BricsCAD citing a 58% AutoCAD price increase over six years. Mazda and PERI adopted BricsCAD for cost reasons. But switching away from Revit? That’s exceedingly rare. As one Revit forum user bluntly stated: “Without a valid alternative to Revit, you don’t have much to stand on.”


Financial returns confirm the moat, with caveats

A competitive advantage must manifest in superior financial returns. Autodesk’s numbers largely deliver, though with important context.

MetricFY2022FY2023FY2024FY2025FY2026
Revenue ($B)$4.39$5.01$5.50$6.13$7.21
Gross margin~89%~90%~90%90.6%91.0%
Non-GAAP op. margin~32%~34%~36%36%38%
Free cash flow ($B)$1.48$2.03$1.28$1.57$2.41
FCF margin33.6%40.6%23.3%25.6%33.4%
ROEN/MN/M~100%+~42%~40%
ROIC~10–14%~19%~7–12%~12.6%~14–17%

Gross margins of 91% place Autodesk at the top of the design software peer group (Bentley ~80%, PTC ~84%, Dassault ~85%, Nemetschek ~75%). This is the signature of a business with near-zero marginal cost of delivery and strong pricing power. Non-GAAP operating margins of 38% are improving but still trail PTC’s 48% — a gap Starboard Value has hammered, arguing Autodesk should reach 45% by FY2028.

ROIC tells a more nuanced story. At ~14–17% on an invested capital base that includes $4.3 billion of goodwill from acquisitions, Autodesk clears its cost of capital (~10–11% WACC) but does not generate the exceptional returns one might expect from a monopoly-like franchise. The mediocre ROIC partly reflects the dilutive effect of stock-based compensation consuming ~11% of revenue ($788M in FY2026) — a real economic cost that management’s non-GAAP metrics conveniently exclude. SBC consumed roughly 33% of FY2026 free cash flow, and while this ratio is improving, it remains a meaningful drag on per-share value creation.

The critical comparison: despite spending $1.3–1.4 billion annually on buybacks, Autodesk has reduced diluted shares from ~222 million to ~213 million over five years — a net reduction of only ~4%. The vast majority of buyback spending simply offsets SBC dilution. This is not capital allocation that creates shareholder value; it is capital allocation that prevents shareholder value from being destroyed by compensation practices.


The subscription transition succeeded, but the billing transition raised trust issues

Autodesk’s 2016–2020 shift from perpetual licenses to subscriptions was a genuine strategic triumph. Revenue cratered 18.9% in FY2017 as upfront perpetual sales were replaced by ratable subscription recognition, but recurring revenue climbed from under 50% to 97% — transforming the business from a cyclically vulnerable software vendor into a predictable annuity stream. Revenue has more than tripled from the FY2017 trough.

The more recent billing transition — the New Transaction Model (NTM) shifting from reseller buy-sell to direct agency billing — is strategically sound but was marred by execution issues. Under the old model, resellers controlled pricing, offered discretionary discounts, and owned customer billing relationships. AEC Magazine estimated resellers were giving away ~$600 million in discounts. The NTM gives Autodesk direct control over pricing, data, and cash collection. By FY2025’s end, 84% of business had migrated to the new model.

However, the NTM transition coincided with — and was intertwined with — the most damaging governance episode in recent Autodesk history. In March 2024, an internal investigation revealed that during FY2022–FY2024, management made operational decisions about multiyear billing, collections timing, and discretionary spending that were “informed by their anticipated effects on the company’s external free cash flow and/or non-GAAP operating margin targets.” In plainer language: management manipulated the timing of cash flows to hit publicly guided FCF targets. In FY2023 specifically, after announcing a shift to annual enterprise billing, the company reversed course and aggressively pursued multiyear upfront contracts to meet its lowered FCF guidance — upfront billings “substantially exceeded historical levels.”

The Audit Committee found no fraud requiring financial restatement. Both the SEC and U.S. Attorney’s Office closed their investigations without enforcement action by August 2025. But CFO Deborah Clifford was removed (reassigned to a strategy role, effectively terminated). The incident revealed that management was willing to reverse publicly communicated strategic decisions to manage non-GAAP optics — a serious trust issue for any investor relying on management’s forward guidance.


Growth is real but decelerating from a high base

Autodesk’s growth decomposition reveals a business still expanding meaningfully, though the easiest gains from the subscription conversion are behind it.

PeriodRevenue CAGR
3-year (FY2023–FY2026)~13.3%
5-year (FY2021–FY2026)~13.7%
10-year (FY2016–FY2026)~11.2%

Growth has been overwhelmingly organic (85–95%+). The largest acquisition, Innovyze ($1B), added perhaps 2–3 percentage points to FY2022 growth. The FY2026 acceleration to 17.5% partly reflects NTM billing benefits and strong AECO demand rather than a permanent step-up in underlying growth.

Forward growth drivers are credible but require execution:

  • AECO secular expansion remains the strongest driver. Global BIM mandates are expanding, data center construction is surging (~30% YoY), and Autodesk’s “Make” revenue (construction cloud products) reached $796M in FY2026, growing 22%. This is a real, substantial new business.
  • Pricing power is genuine. Autodesk raises prices 3–5% annually with limited churn impact. The NTM’s elimination of reseller discounting represents an additional hidden price increase. Net revenue retention has held steady at 100–110%.
  • Collections and EBA upsell provide mix-shift tailwinds as customers move from individual products (~$1,975/year for AutoCAD) to AEC Collections (~$3,225/year) or token-based Enterprise Business Agreements. EBAs now represent ~15% of revenue.
  • Remaining perpetual/non-compliant conversion: At the 2020 Investor Day, Autodesk identified ~2 million additional targetable subscriptions. This opportunity is diminishing but not exhausted.

Management’s FY2027 guidance of $8.1–8.17B revenue (+12–13%) and $2.7–2.8B FCF was well above prior consensus. The FY2029 non-GAAP operating margin target is 41%, implying 300 basis points of expansion from current levels.

Management’s TAM claims deserve skepticism. The 2019 Investor Day cited a $59 billion “Design and Make” TAM — credible when combining global CAD ($24B), BIM ($10–15B), construction tech, and PLM markets. The more recent $100 billion+ figure stretches into aspirational territory, requiring successful expansion into operations, digital twins, and AI services Autodesk doesn’t yet meaningfully deliver. At $7.2B revenue, current penetration of the realistic $50–65B TAM is roughly 11–14% — substantial room to grow, but not the 93% untapped opportunity management implies.


AI is a double-edged sword, tilting in Autodesk’s favor

The AI question is arguably the most important long-term variable for Autodesk’s investment case. The company is investing aggressively: Neural CAD (purpose-built 3D generative AI producing fully editable BREP geometry), Autodesk Assistant (agentic AI across Fusion, Revit, AutoCAD), and a $200M strategic investment in World Labs.

The bull case for AI strengthening the moat is compelling. Autodesk possesses decades of proprietary professional CAD/BIM training data that generic foundation models cannot replicate. Its AI research lab has published ~100 peer-reviewed papers on CAD geometry — making it the world’s leading institution in this niche. CEO Anagnost’s framing is apt: unlike consumer AI applications where “probably right” suffices, buildings and manufactured products must be precisely correct. Domain-specific AI in high-stakes industries favors incumbents with deep expertise and existing distribution.

The bear case cannot be dismissed, however. AI-native startups like Snaptrude (text-to-BIM), TestFit (generative site design), and Swapp (automated architectural drawings) are targeting specific workflow segments. More structurally, NVIDIA’s Omniverse and the USD open standard could erode Autodesk’s proprietary file format lock-in by commoditizing 3D interoperability. And if AI genuinely automates 80–90% of routine design tasks (as Autodesk itself claims Neural CAD can do), the logical consequence is fewer designer seats — directly threatening the per-seat subscription model that generates 97% of revenue. Autodesk’s proactive shift toward consumption-based pricing (EBAs, Flex tokens at ~17% of FY2026 revenue) is the right strategic response but creates its own transition risks.

On balance, AI more likely strengthens Autodesk’s position than destroys it. The company has first-mover advantage, proprietary data, distribution to millions of users, and the financial resources ($1.65B annual R&D) to integrate AI faster than startups can build full-platform alternatives. But the risk is nonzero and warrants ongoing monitoring.


Capital allocation is adequate but not exceptional

Over the past decade, Autodesk has allocated capital across three primary channels:

Category~10-year cumulativeAssessment
R&D~$11–12B (22–25% of revenue)Appropriately high for platform innovation
M&A~$4.5–5BMixed results (see below)
Buybacks~$4–5B (primarily FY2022–FY2026)Mostly offsets SBC dilution
Dividends$0No dividend policy

The M&A track record is mixed. BuildingConnected ($275M) appears excellent — its bidding network now manages $56 billion in monthly project bids with 15 of the top 20 ENR contractors as customers. Spacemaker ($240M) was strategically astute, providing the cloud-native architecture for the entire Forma platform. PlanGrid ($875M) was expensive for a startup but anchored the construction cloud strategy. Innovyze ($1B) raises concerns — Autodesk paid nearly 4x what EQT paid just four years earlier ($270M in 2017), and the standalone ROI remains unquantified.

The balance sheet is conservatively managed: $2.6B cash against $2.7B debt yields near-zero net debt, with interest coverage exceeding 20x. This is appropriate for a company with highly predictable cash flows and no immediate need for financial leverage.


Governance improved after a necessary reckoning

CEO Andrew Anagnost deserves credit for executing the subscription transformation — revenue has tripled since his 2017 appointment. But his $25.2 million FY2025 compensation (175:1 CEO pay ratio), combined with the accounting investigation on his watch, invites scrutiny. Starboard’s criticism that Autodesk has “missed or is on course to miss every investor day commitment since 2018” may be overstated but contains a kernel of truth about margin execution pace.

The Starboard settlement was net positive. Adding Jeff Epstein (former CFO of multiple public companies, Bessemer operating partner) and Christie Simons (Micron board, audit expertise) strengthened a board that needed it. The new CFO, Janesh Moorjani (ex-Elastic), brings external rigor. Starboard’s 44% position selldown mid-campaign raises questions about its own sincerity, but the governance improvements it catalyzed are lasting.

Insider activity shows predominantly routine selling ($45M+ over 24 months, mostly 10b5-1 plan sales of vested equity). The sole notable purchase — Board Chairman John Cahill buying $534,200 during the March 2025 proxy fight — is a modest but meaningful confidence signal.


Valuation embeds conservative assumptions the business may not justify

MetricTrailing (FY2026)Forward (FY2027E)5-year average
P/E (GAAP)~45x—~60x
P/E (non-GAAP)~23x~19x—
EV/Revenue~7.0x~6.2x8–12x
EV/EBITDA~25x~20–22x37–80x
EV/FCF~21x~18x—

Autodesk trades at or near multi-year lows on virtually every valuation metric. The trailing GAAP P/E of ~45x represents a 25% discount to the 5-year average of ~60x. EV/Revenue at 7x is below the historical 8–12x range. The PEG ratio of 0.93 suggests the market is pricing in growth deceleration more severe than either management or analysts expect.

A reverse DCF using FY2026 FCF of $2.41B, a 10% discount rate, and 3% terminal growth rate implies the market is pricing in only ~7.5–8% annual FCF growth over the next decade. This appears conservative when:

  • Management guides FY2027 FCF growth of 12–16%
  • Revenue growth has averaged ~13% over 5 years with ongoing margin expansion
  • The FY2029 margin target of 41% implies continued FCF expansion even if revenue growth decelerates

Against peers, Autodesk’s forward non-GAAP P/E of ~19x is the cheapest in the group (Bentley 27.5x, Nemetschek 28.7x, PTC 22x, Trimble 23.5x). Only Dassault trades cheaper at ~14x, but Dassault has suffered a severe re-rating (-50% over 12 months) amid sector-specific weakness.

All 26 covering analysts rate the stock Buy or Strong Buy, with zero Sell ratings. The average price target of ~$340–370 implies 40–55% upside. The stock is trading near the lowest analyst target.


The five risks that matter most

1. Construction cyclicality is real. Despite 97% recurring revenue making Autodesk far more resilient than in 2008 (when quarterly revenue fell 30%), prolonged weakness in commercial construction and manufacturing capex would slow new seat growth and pressure the narrative. US construction spending declined 4.7% in real terms in 2025.

2. The accounting investigation damaged management credibility. No fraud was found, but the revealed willingness to manipulate non-GAAP metrics through operational decisions means investors must discount management’s forward FCF guidance somewhat. Trust, once broken, takes years to rebuild.

3. Pricing pushback could accelerate. Autodesk has raised prices aggressively (including the hidden price increase of eliminating reseller discounts). Customer forums show rising frustration. If AI tools reduce designer productivity requirements, the value proposition of $3,225/year AEC Collections becomes harder to justify.

4. AI-driven seat compression is a tail risk. If generative AI genuinely automates 80–90% of routine design tasks (Autodesk’s own claim), fewer designers per project means fewer subscriptions — even if per-seat value increases through AI features. The timing is uncertain, but the direction is clear.

5. Go-to-market restructuring creates near-term execution risk. The February 2025 layoff of 1,350 employees (9% of workforce) and sales organization optimization could disrupt customer acquisition in FY2027. Management explicitly flagged “higher probability and potential impact of disruption” from these changes.


Bull case vs. bear case

The bull case requires believing that Autodesk’s AEC moat is deepening (via BIM mandates, construction cloud adoption, and AI integration), that margins will expand toward 41%+ as Starboard-driven discipline takes hold, and that the current valuation represents a temporary dislocation driven by macro anxiety and the governance overhang. In this scenario, Autodesk generates $3.5–4B in FCF by FY2029 on $10B+ revenue, and the stock re-rates toward historical multiples. A re-rating to even 25x FCF on $3.5B would imply an enterprise value of ~$87B — roughly 75% upside.

The bear case requires believing that construction cyclicality is entering a prolonged downturn, that AI disruption compresses long-term seat counts faster than Autodesk can transition to consumption pricing, that margin expansion stalls as competitive pressure forces continued heavy R&D and sales spending, and that the NTM transition and go-to-market restructuring create lasting channel friction. In this scenario, revenue growth decelerates to mid-single digits, FCF stagnates around $2.5–3B, and the stock trades at 15–18x FCF — implying a range of $175–250, roughly flat to down from current levels.


Conclusion

Autodesk is a genuinely good business trading as if it were a mediocre one. The 91% gross margins, 97% recurring revenue base, and $2.4B in free cash flow are the financial fingerprints of a durable competitive advantage — one rooted primarily in Revit’s proprietary file format lock-in, deep workflow entrenchment in AEC, and the structural tailwind of global BIM mandates. The subscription transition is complete. Growth is real and predominantly organic. The balance sheet is clean.

The concerns are also real: the accounting investigation revealed management’s willingness to prioritize optics over transparency, SBC consumes a disproportionate share of free cash flow, ROIC — while above cost of capital — does not scream monopoly-grade returns, and the manufacturing and M&E verticals lack the moat depth of AEC. AI is a wildcard that probably helps Autodesk but introduces genuine long-term uncertainty about the per-seat business model.

What stands out most is the valuation disconnect. At ~21x trailing FCF and ~19x forward non-GAAP earnings, the market implies only 7.5–8% long-term FCF growth — well below the 12–16% the company is guiding for the next year and the ~13% it has compounded over five years. Either the market knows something the numbers don’t show, or this is a franchise temporarily mispriced due to a convergence of governance noise, macro anxiety, and software sector de-rating. The weight of the evidence suggests the latter, but the governance issues warrant a discount for uncertainty around management’s credibility — a discount that may already be more than fully reflected in the stock price.

Frequently Asked Questions

General Questions

What thoughtful questions have other investors asked about this company?

The most incisive questions from the Q4 FY2026 earnings call and analyst community center on several themes. First, the durability of the NTM billing tailwind — multiple analysts pressed on how much of the 17.5% FY2026 growth was “pull-forward” from eliminating reseller discounts versus sustainable organic expansion. This is the right question because the NTM is a one-time structural repricing event that flatters the growth rate for 2–3 years and then normalizes. Second, Starboard and several sell-side analysts have asked why operating margins remain at 38% when PTC runs at 48% on similar gross margins — the implication being that Autodesk’s cost structure is bloated relative to what the franchise economics should support. Third, the AI seat compression question: if Autodesk’s own Neural CAD automates 80–90% of routine design tasks, doesn’t the logical endgame involve fewer seats? Management’s answer — consumption-based pricing and higher value per seat — is directionally correct but untested at scale. Fourth, construction cyclicality: with US real construction spending declining in 2025, how resilient is the “97% recurring” revenue claim when customers can simply not renew or downgrade? The honest answer is that Autodesk has never tested its current subscription model through a deep construction recession.


Cyclicality & Earnings Nature

Are earnings at a cyclical high or cyclical low?

Neither extreme, but tilting toward the favorable end of the cycle. FY2026’s 17.5% revenue growth was the fastest since the subscription transition era, partly driven by the one-time NTM repricing tailwind and strong AECO demand. Operating margins at 38% are still 300+ basis points below management’s FY2029 target of 41%, suggesting margin expansion runway remains. However, US nonresidential construction put-in-place has softened, and the global manufacturing PMI has been mixed. Earnings are above mid-cycle but not at a peak — the NTM tailwind will fade over FY2027–FY2028, which is why management guided FY2027 revenue growth of 12–13% (decelerating from 17.5%). The most honest characterization: earnings quality is high (97% recurring), but the growth rate is temporarily elevated by the billing transition.

Are earnings driven primarily by the external environment or internal company actions?

Predominantly internal actions layered on top of a favorable external backdrop. The subscription transition (2016–2020), the NTM billing shift (2023–2025), regular 3–5% annual price increases, and cross-sell from individual products to Collections are all management-driven. The external tailwinds — BIM mandates, data center construction boom, infrastructure spending — are real but wouldn’t generate 13%+ CAGRs without Autodesk’s pricing discipline and product bundling strategy. In a downturn, the recurring subscription base provides a floor that perpetual-license Autodesk (pre-2016) never had. That said, new seat growth is clearly macro-sensitive — net new subscriber additions slow when construction firms stop hiring.

How stable are revenues?

Extremely stable by any reasonable standard. 97% of revenue is recurring subscription revenue. Net revenue retention runs 100–110%, meaning the existing customer base grows even before adding new customers. Autodesk has not experienced a revenue decline since FY2017 (the deliberate subscription transition trough). Even in the mild 2023 industrial slowdown, revenue grew 12%. The key risk is not revenue collapse but growth deceleration — if construction enters a prolonged downturn, revenue growth could slow to mid-single digits, but an actual revenue decline would require mass non-renewals, which the switching cost analysis suggests is unlikely for Revit/BIM users.

Outlook for the company’s products and services?

Strong in AEC, adequate in Manufacturing, challenged in M&E. The AEC outlook is the best it’s been in a decade: BIM mandates are expanding globally (India developing national policy, Germany requiring BIM on federal projects, EU pushing standardization), data center construction is booming, and Autodesk Construction Cloud is gaining traction against Procore in field execution. Manufacturing is a more contested market where Autodesk’s Fusion platform is competitive in the mid-market but faces Dassault and Siemens in enterprise. M&E at 4.6% of revenue and 5% growth is essentially a legacy franchise being slowly eroded by Blender and Unreal Engine — it’s not strategically important enough to worry about but also not a growth driver.

How big will this market be? Is it growing? Shrinking? Domestic or international?

The realistic TAM for Autodesk’s current product portfolio is roughly $50–65 billion, comprising global CAD (~$24B), BIM/AEC (~$10–15B), construction technology (~$10B), PLM/manufacturing (~$10–15B), and M&E (~$3–5B). This market is growing at roughly 8–10% annually, driven primarily by BIM adoption in emerging markets, construction digitization, and the shift from 2D to 3D design workflows. Autodesk’s revenue of $7.2B implies roughly 11–14% market penetration. The geographic split is roughly 40% Americas, 35% EMEA, 25% APAC, with the fastest growth coming from APAC and emerging markets where BIM adoption is earlier-stage. Management’s $100B+ TAM claim stretches into adjacent markets (digital twins, operations, AI services) that Autodesk doesn’t meaningfully serve today — treat that number as aspirational, not investable.


Business Quality & Competitive Moat

Is the industry getting more or less competitive?

The answer differs sharply by vertical. In AEC/BIM authoring, the industry is getting less competitive — government BIM mandates effectively standardize on Revit (and to a lesser extent ArchiCAD in parts of Europe), raising barriers for new entrants. The construction cloud space is getting more competitive (Procore, Oracle Aconex, Trimble). In Manufacturing CAD, competition is stable but intense — the same four players (Autodesk, Dassault, Siemens, PTC) have dominated for 20+ years with minimal share shifts. In M&E, competition is getting dramatically more intense as Blender (free, open-source, improving rapidly) erodes the value proposition of paid tools. The net effect: Autodesk’s strongest and largest segment (AEC) is becoming more defensible, which is what matters most for the investment case.

How profitable is this business? What is the return on capital invested? Return on equity?

Gross margins of 91% are among the highest in all of software. Non-GAAP operating margins of 38% are good but not best-in-class (PTC runs 48%). GAAP operating margins are significantly lower (~23%) due to $788M in SBC and $400M+ in amortization of acquired intangibles. ROIC runs roughly 14–17% depending on how you treat goodwill — adequate but not spectacular. ROE is optically high (~40%) but misleading because Autodesk has negative tangible book equity (accumulated deficit of ~$4.5B from years of buybacks and the subscription transition’s depressed earnings). The honest assessment: this is a very profitable business whose returns on capital are good but diluted by aggressive SBC and historically expensive acquisitions. If you strip out goodwill and measure returns on tangible capital employed, the numbers look much better — but that’s also a less conservative measure.

How profitable is this industry? Are there a lot of competitors? What are the barriers to entry?

The design software industry is one of the most profitable in technology. All major players — Autodesk, Dassault, PTC, Bentley, Nemetschek — operate at 75–91% gross margins and 25–48% operating margins. The industry is an oligopoly with extremely high barriers to entry: it takes 10–20 years to develop a competitive CAD/BIM platform, customer switching costs are measured in months of lost productivity, proprietary file formats create lock-in, and the installed base of trained professionals creates a self-reinforcing ecosystem (universities teach Revit and AutoCAD, graduates demand those tools, firms standardize on them). The last meaningful new entrant in core CAD was Onshape (2015), which PTC acquired for $470M — confirming that even successful entrants get absorbed by incumbents. Open-source alternatives (FreeCAD, BRL-CAD) exist but lack the reliability, support, and ecosystem integration required for professional use on real projects.

Can this business be easily understood?

Yes, at the core level: Autodesk sells mission-critical design software on annual subscriptions to architects, engineers, and manufacturers who cannot easily switch away. The complexity comes from the multiple product lines, the ongoing billing transition, and the heavy use of non-GAAP metrics that obscure true economics. But the fundamental business model — recurring subscriptions for sticky software with pricing power — is straightforward.

Can this company be undermined by foreign, low-cost labor?

No. Autodesk sells software tools, not labor. The users of Autodesk’s software (architects, engineers) can themselves be offshored, but that doesn’t reduce demand for the software — it may actually increase it as firms in India, Southeast Asia, and Africa adopt BIM tools. If anything, globalization of the AEC workforce expands Autodesk’s addressable market.

Do brands matter?

Yes, significantly. “AutoCAD” is effectively a generic term for 2D design software in much of the world, similar to how “Photoshop” means image editing. “Revit” is the default specification in AEC project requirements. Brand recognition reduces customer acquisition costs and creates a self-reinforcing cycle through university education programs (Autodesk provides free educational licenses, training the next generation on its tools). However, brand alone doesn’t create the moat — the switching costs and file format lock-in do the heavy lifting.

What is the nature of competition?

Competition is primarily on product capability, ecosystem integration, and workflow entrenchment rather than on price. Autodesk’s competitors don’t typically try to undercut on price (BricsCAD being a notable exception in 2D CAD). Instead, they compete on specific vertical capabilities: Dassault/Siemens dominate aerospace and automotive manufacturing, Bentley owns infrastructure engineering, Nemetschek/ArchiCAD is strong in European architectural practice. The competitive dynamic is more like geographic/vertical market-sharing than head-to-head price war — which is the hallmark of a healthy oligopoly.

What are the customer switching costs?

Extremely high for Revit/BIM workflows, moderate for AutoCAD, low for M&E tools. A mid-size architecture firm switching from Revit to ArchiCAD faces: 4–8 weeks of productivity loss per employee during retraining, months of library migration (custom Revit families, templates, standards), potential incompatibility with client/contractor BIM requirements, and disruption to multi-discipline coordination workflows. Industry participants consistently report that Revit switching costs amount to 6–12 months of disrupted operations for a typical firm. For AutoCAD (2D drafting), switching costs are lower — BricsCAD and others read .dwg files natively — which is why AutoCAD faces more competitive pressure than Revit.

What are the barriers to entry?

The barriers stack: (1) Technical complexity — building a production-grade BIM platform takes 10–20 years of development; (2) File format lock-in — .rvt is proprietary with no open-source equivalent; (3) Ecosystem network effects — millions of trained users, university programs, third-party plugin developers; (4) Regulatory lock-in — BIM mandates often implicitly or explicitly require Revit-compatible deliverables; (5) Data gravity — decades of project files, templates, and institutional knowledge stored in Autodesk formats; (6) Distribution — Autodesk has direct relationships with millions of subscribers and a global reseller network. A well-funded startup could theoretically build a competitive point solution (and some are trying with AI), but replicating the full-platform ecosystem is practically impossible.


Financial Condition & Balance Sheet

Does the company have assets that are not fully recognized in the balance sheet?

Yes, substantially. Autodesk’s most valuable asset — its installed base of ~7 million+ subscribers paying recurring annual fees — does not appear on the balance sheet. The brand value of “AutoCAD” and “Revit” as industry-standard names is similarly unrecognized. Additionally, the proprietary training data from decades of professional CAD/BIM usage (which powers Autodesk’s AI initiatives) has significant economic value that is entirely off-balance-sheet. These are the assets that make the business worth ~$50B despite a tangible book value that is deeply negative.

What off-balance sheet liabilities does the company have?

Operating lease commitments (~$400–500M), purchase obligations for cloud infrastructure and third-party services, and the ongoing economic cost of future SBC grants (which will dilute existing shareholders by ~2–3% annually before buybacks). There are no material contingent liabilities, pending litigation, or structured off-balance-sheet vehicles. The SEC and DOJ investigations related to the 2024 accounting matter were both closed without action.

How conservative is the company’s accounting?

Mixed. On the positive side, Autodesk’s subscription revenue recognition is straightforward (ratable over the contract term), and the company received clean audit opinions from EY. On the negative side, the 2024 investigation revealed that management made operational decisions to manage non-GAAP metrics — not technically fraudulent but clearly aggressive. The heavy reliance on non-GAAP metrics (which exclude $788M in SBC, $400M+ in amortization) creates a significant gap between reported “adjusted” profitability and GAAP reality. Non-GAAP EPS of $9.20 versus GAAP EPS of $5.14 in FY2026 tells you the magnitude of the adjustment. I’d rate the accounting as “moderately aggressive” — not manipulative, but definitely presenting the business in the most favorable light possible.

How CapEx hungry is this business?

Minimally. Capital expenditures run approximately $100–150M annually (roughly 1.5–2% of revenue), consisting primarily of data center infrastructure and office buildouts. This is a software business with negligible physical asset requirements. Free cash flow conversion from operating cash flow is consistently 90%+. The low CapEx intensity is one of Autodesk’s most attractive characteristics — nearly all operating cash flow drops to free cash flow.


Capital Allocation & Management

How much free cash flow does the business generate? How does management use this free cash flow? What is their philosophy?

FY2026 FCF was $2.41 billion (33.4% FCF margin), up 54% year-over-year. Management allocates FCF almost entirely to share buybacks — $1.3–1.4B annually — with no dividend. The stated philosophy is to return excess cash to shareholders via repurchases while maintaining investment-grade credit. The problem, as noted in the deep dive, is that most buyback spending simply offsets SBC dilution rather than creating per-share value. Net share count has declined only ~4% over five years despite ~$6B+ in cumulative repurchases. This is the single most legitimate criticism of Autodesk’s capital allocation — the company is spending billions to stand still on share count.

Has the company made any significant acquisitions recently?

No major acquisitions since Innovyze ($1B, closed March 2021). The most recent notable transaction was the $200M strategic investment in World Labs (AI-powered 3D world generation, founded by Fei-Fei Li) in late 2025/early 2026. This signals a shift toward minority investments in AI rather than large platform acquisitions, which is arguably more disciplined.

Is the company buying back shares?

Yes, aggressively. Autodesk repurchased approximately $1.3–1.4B in shares in FY2026. But as discussed, the net effect after SBC dilution is minimal. Diluted shares outstanding have gone from ~222M to ~213M over five years — a compound annual reduction of less than 1%. The buyback program is real but functions primarily as an anti-dilution mechanism rather than a value-creation tool.

Does the company issue large amounts of new shares to insiders?

Yes. SBC of $788M in FY2026 represents ~11% of revenue and ~33% of free cash flow. This is high by absolute standards, though in line with large-cap software peers (Adobe, Salesforce, Intuit run similar percentages). The dilutive impact is approximately 2–3% of shares annually before buybacks. This is the biggest single drag on per-share value creation and the area where Starboard’s margin expansion thesis has the most validity — if SBC as a percentage of revenue can decline from 11% toward 8%, the per-share economics improve meaningfully.

What is the compensation policy of directors and management?

CEO Anagnost’s FY2025 total compensation was $25.2M (up 22% YoY), comprising a base salary of ~$1.1M, cash bonus, and the majority in performance-based equity awards tied to revenue growth and non-GAAP operating margin targets. The CEO-to-median-employee pay ratio is 175:1. The performance targets themselves are reasonable, but the use of non-GAAP metrics as compensation triggers creates the exact incentive that drove the 2024 accounting investigation — management was motivated to hit non-GAAP FCF and margin targets, and they made operational decisions accordingly. Board compensation runs ~$350–400K annually per independent director, in line with peers.

What are the motivations of management?

Management is primarily motivated by equity compensation tied to non-GAAP operating margin and revenue growth targets. Anagnost owns a meaningful but not outsized stake (worth roughly $40–60M at current prices). The alignment is adequate but imperfect — the non-GAAP focus creates incentives to manage SBC and FCF timing rather than optimize GAAP returns on capital. The Starboard settlement and board refresh should improve governance discipline, but the fundamental incentive structure hasn’t changed.


Valuation & Market Data

Is the stock an ADR? MLP? K-1?

No. ADSK is a domestic C-corporation listed on NASDAQ. Standard 1099 tax reporting. No K-1 complexity.

Dividend Policy?

No dividend. No stated intention to initiate one. All excess cash is returned via buybacks. Given the SBC dilution dynamic, a dividend might actually be more shareholder-friendly than buybacks at current prices — but management has shown no interest in this approach.

Is net income diverging from cash from operations?

Yes, but in the right direction. FY2026 GAAP net income was ~$1.1B while operating cash flow was ~$2.5B, with the gap explained primarily by non-cash charges (SBC, depreciation/amortization, deferred revenue dynamics). This divergence is normal and healthy for a subscription software business — cash collections run ahead of revenue recognition because annual subscriptions are often billed upfront. There are no signs of earnings quality deterioration in the cash-versus-accrual relationship. The one historical exception was FY2024, when FCF dropped to $1.28B due to the billing transition disruption — but that has since normalized.


Risks & Downside

What factors would cause the stock to decline?

The most probable catalysts for decline are: (1) A sharp construction downturn that slows new seat growth and triggers negative sentiment about the “secular growth” narrative; (2) FY2027 results that miss the ambitious guidance ($8.1–8.17B revenue) due to go-to-market restructuring disruption from the 9% workforce reduction; (3) Evidence that AI-native competitors are gaining traction in BIM authoring (watch Snaptrude, TestFit, and any NVIDIA-adjacent entrants); (4) Further governance issues or management credibility erosion; (5) Broader software sector de-rating if interest rates remain elevated.

What is the risk of a catastrophic loss?

Very low. Autodesk has a clean balance sheet (near-zero net debt), $2.6B in cash, $2.4B in annual FCF generation, 97% recurring revenue, and products so deeply embedded in customer workflows that mass defection is structurally implausible. The business would continue generating substantial cash even in a severe recession. This is not a business that goes to zero or suffers permanent capital impairment under any realistic scenario.

Chance of a total loss?

Effectively zero. Even in a catastrophic scenario — say, AI completely commoditizes CAD/BIM design within 5 years — Autodesk’s $7.2B annual revenue base, $2.4B FCF, and dominant market position would make it an acquisition target for any number of technology companies (Microsoft, Oracle, SAP) long before equity value approached zero. The installed base of millions of subscribers paying recurring fees is worth tens of billions to any strategic acquirer.


Recent News & Events

Has the business environment changed recently?

Yes, in several ways. The Trump administration’s tariff policies introduced in early 2025 have created uncertainty around manufacturing capital expenditure, which could slow demand for Autodesk’s manufacturing products. US construction spending declined 4.7% in real terms in 2025, though data center construction (a growing Autodesk vertical) surged ~30%. The AI landscape has evolved rapidly — Autodesk’s $200M World Labs investment and Neural CAD announcements represent a significant strategic pivot toward AI-native product capabilities. Globally, BIM mandates continue expanding, providing a structural tailwind.

Has the company made any significant acquisitions recently?

No major acquisitions. The $200M World Labs minority investment (late 2025/early 2026) is the most notable recent capital deployment, signaling a preference for AI partnerships over large-scale M&A.

Has the company recently changed accounting policies?

The 2024 internal investigation led to enhanced internal controls around FCF and non-GAAP metric management, though no formal restatement was required. The NTM billing transition changed revenue recognition timing (shifting from net to gross for agency-model transactions), which inflated reported revenue growth by several percentage points during the transition period — an important nuance when evaluating FY2025–FY2026 growth rates. The new CFO (Janesh Moorjani, appointed mid-2024) has emphasized more conservative financial communications.

Recent changes in the business, new markets, new production facilities, what’s changed recently? New management?

The most significant recent changes: (1) Workforce reduction of 1,350 employees (9%) announced February 2025, restructuring the go-to-market organization; (2) New CFO Janesh Moorjani (ex-Elastic) replaced Deborah Clifford following the accounting investigation; (3) Two new independent board members (Jeff Epstein, Christie Simons) added as part of the Starboard settlement; (4) Autodesk Construction Cloud reaching critical scale at $796M annual revenue; (5) Forma (formerly Spacemaker) emerging as the cloud-native early-stage design platform; (6) Neural CAD AI capabilities announced, representing Autodesk’s most substantive AI product initiative to date; (7) FY2027 guidance significantly above consensus, suggesting management confidence in the post-restructuring trajectory.


The bottom line on the full checklist: Autodesk passes the competitive advantage test clearly in AEC, less convincingly in Manufacturing, and fails it in M&E. The financial returns confirm a good business, but SBC dilution and historically mediocre ROIC prevent it from being a great capital compounder on a per-share basis. Capital allocation is the weakest link — billions spent on buybacks that barely offset dilution. The valuation is compelling if you believe the franchise is durable and margins will expand, but governance credibility has been damaged and the go-to-market restructuring introduces near-term execution risk.