Arm Holdings (ARM) FY2025 — Grade F: M-Score Breached, CFFO Collapsed 63%

Grade: F — Major Red Flags

Framework: Schilit *Financial Shenanigans* + Beneish M-Score + forensic accounting principles

Data: SEC EDGAR 20-F (Filed May 28, 2025, FY ended March 31, 2025) + Yahoo Finance

Auditor: Deloitte & Touche LLP — Unqualified opinion (1 critical audit matter: revenue recognition for material long-term revenue contracts)

One-line verdict: Arm's FY2025 20-F shows the rare combination of strong headline growth alongside cash flow that diverged sharply from reported earnings. Revenue rose 24% to "$4,007 million during the fiscal year ended March 31, 2025, from total revenue of $3,233 million during the fiscal year ended March 31, 2024," but operating cash flow collapsed 64% from $1,090 million to $397 million in the same year. The Beneish M-Score of -1.75 sits above the -1.78 threshold, indicating elevated manipulation risk that warrants further investigation. Underlying drivers — IPO-related employment tax payments, accounts receivable timing shifts, and a $246.4 million fair value loss on Ampere — are partially explained in the MD&A, but the divergence is large enough to push this filing into the bottom decile of the screening universe. Layered on top: extreme customer concentration (top four customers = 49% of revenue per the financial statements), 87.1% SoftBank ownership, dependence on Arm China (which the 20-F admits "operates independently of us"), and active litigation with Qualcomm heading to trial in March 2026.

The IP Licensing Model — Why Cash and Earnings Can Drift Apart

The 20-F describes Arm's revenue model in detail in Item 5: "Most of our licenses have two components: license fees and support and maintenance fees (recognized as license and other revenue) and per-chip royalties (recognized as royalty revenue). However, some licenses can have multiple payment milestones that are date-based or event-based (e.g., six months after the effective date of the contract or upon tape-out of the first chip design)."

The license revenue stream is governed by a small number of large multi-year agreements: "Over the term of a license, contractual payments can generally range from hundreds of thousands of dollars to hundreds of millions of dollars, depending on the type of license, its duration, and the type of product that is being licensed... A license may be perpetual, use-limited or time-limited in its application." Per the auditor's critical audit matter, these contracts include "intellectual property licenses, implicit rights to future products, early access rights, software licenses, support, training, professional and design services, and other non-standard performance obligations that may be contract or customer specific."

Royalty revenue is recognized on customer self-reports: "License contracts require the licensee to issue royalty reports, including details of chip sales, to us on a quarterly basis. Royalty revenue is recognized on an accrual basis in the quarter in which the customers ship chips containing our products, using estimates from sales trends and judgment for several key attributes."

The structure creates a wide gap between when revenue is recognized and when cash arrives — a model that can produce both legitimate earnings smoothing and harder-to-detect manipulation if controls fail.

Profitability: Headline Numbers (in $ millions)

From the 20-F's consolidated income statement:

License/royalty mix from the 20-F: "License and other revenue increased $408 million, or 29%, during the fiscal year ended March 31, 2025... primarily driven by new licensing agreements, an increase in revenue from arrangements entered into in prior periods, and renewals of our existing license arrangements with certain customers to gain access to the latest versions of our IP. Royalty revenue increased $366 million, or 20%, during the fiscal year ended March 31, 2025, as compared to the fiscal year ended March 31, 2024, driven primarily by higher chip shipments and an improved mix of products with higher royalty rates per chip."

The 20-F discloses that 21% of FY2025 revenue ($823M) came from related parties — primarily Arm China and other SoftBank-controlled entities. This is structurally elevated for a public company.

The $237 million loss from equity investments was driven specifically by Ampere: "Loss from equity investments, net increased by $217 million, or 1085%, for the fiscal year ended March 31, 2025, as compared to the fiscal year ended March 31, 2024, primarily due to a $246.4 million fair value loss with respect to our investment in Ampere Computing Holdings LLC."

The effective tax rate of ~10% (from a $72M tax benefit on $720M of pretax income) is unusually low. The 20-F explains: "The effective tax rate decreased for the fiscal year ended March 31, 2025, as compared to the fiscal year ended March 31, 2024, primarily due to the tax benefit of windfall gains and research tax credits as a result of an increase in share price." Additionally, Arm benefits from "the U.K.'s patent box regime, which allows certain profits attributable to revenue from patented products (and other qualifying income) to be taxed at an effective corporation tax rate of 10%."

Cash Flow: The 64% Collapse

This is where the FY2025 filing turns problematic. From the 20-F's cash flow summary:

The 20-F's explanation: "Net cash provided by operating activities decreased by $693 million, or 64%, to $397 million for the fiscal year ended March 31, 2025, as compared to $1,090 million for the fiscal year ended March 31, 2024, primarily driven by a decrease of $876 million in other liabilities due to an increase in cash used for employment taxes payable on vested share-based awards related to the IPO share vesting event, a $549 million unfavorable cash flow movement related to accounts receivable due to the timing of payments from customers, partially offset by a $140 million decrease in cash used for accrued compensation and benefits due to the replacement of cash bonuses with equity compensation and favorable movement of $186 million in contract liabilities due to an increase in billings in advance of revenue recognition."

There are three components to the cash flow collapse:

The IPO-related explanation is real but does not fully neutralize the divergence, particularly the $549M AR swing. Even normalizing for the employment tax one-time, cash conversion for FY2025 remains materially weaker than FY2024.

The 18-Point Screening

Revenue Quality

A3 — the divergence is real. Even after accounting for the $876M IPO employment tax payment, the unfavorable AR movement of $549M means Arm extended significantly more credit to customers in FY2025. The 20-F discloses this explicitly as "unfavorable cash flow movement related to accounts receivable due to the timing of payments from customers." DSO climbing from 88 days to 101 days for an IP licensing business — where most contracts are billed upfront — is unusual.

A2 — AR growth outpacing revenue growth. AR rose 41.7% versus 23.9% revenue growth. For a company that recognizes royalty revenue based on customer self-reported chip shipments after they ship, this could reflect either legitimate billing-after-recognition timing or aggressive revenue pull-forward. The auditor's critical audit matter on "Revenue Recognition for Material Long-Term Revenue Contracts" reflects exactly this kind of complexity.

Expense Quality

B2 — CapEx grew at nearly 3x the rate of revenue. The 20-F notes "a $127 million increase in purchases of property and equipment" in the investing activity walk. For an IP licensing business that historically operated extremely capital-light, a 67% jump in CapEx is notable and worth tracking.

The 97% gross margin is structural — Arm sells IP, not chips. Cost of sales of just $121 million on $4 billion in revenue reflects the marginal cost of duplicating digital design files.

Cash Flow Quality

C1 — CFFO covered only 50% of reported earnings. Even with the IPO-related tax payment as a one-time item, this is the weakest CFFO/NI ratio the screen flagged among large-cap technology companies in FY2025. It is the central reason this filing landed in the bottom decile.

The 20-F's three-year CFFO/NI series — 1.41 (FY23), 3.56 (FY24), 0.50 (FY25) — shows a dramatic collapse. Watch FY26 closely to confirm the normalization that the IPO-vesting explanation implies.

Balance Sheet

Goodwill is essentially flat YoY, and total debt is only $356M against $2.8B in cash. The balance sheet itself is healthy.

Acquisition Risk

No acquisition red flags. Arm grows its IP organically.

Manipulation Score

F1 — M-Score sits above the manipulation threshold by 0.03. The components:

When DSRI, SGI, DEPI, and TATA all push in the same direction simultaneously — receivables growing faster than sales, sales growing rapidly, depreciation rate falling, and accruals positive — the Beneish model interprets that as the statistical signature of earnings manipulation. The M-Score of -1.75 signals elevated risk worth investigating, not a definitive finding.

Key Risks from the 20-F (Item 3.D — Risk Factors)

1. SoftBank Owns 87.1% — Public Float Is Effectively a Token

Item 3.D discloses: "As of May 20, 2025, SoftBank Group beneficially owns approximately 87.1% of our total issued and outstanding share capital, representing a majority of the total voting power of our ordinary shares. As a result of SoftBank Group's ownership, we are a controlled company under the Nasdaq corporate governance standards. Because we qualify as a controlled company, we have the option not to comply with certain requirements to which companies that are not controlled companies are subject, including the requirement that a majority of the Board of Directors consists of independent directors."

The 20-F continues: "for so long as SoftBank Group and its controlled affiliates hold shares representing a majority of the votes entitled to be cast by the holders of our outstanding ordinary shares at a shareholder meeting, SoftBank Group will generally be able to determine the outcome of all matters requiring shareholder approval." Public float holders have essentially zero governance power. Related-party transactions with SoftBank-controlled entities — already 21% of revenue — are not subject to independent director oversight in the same way they would be at a non-controlled company.

2. Arm China — A "Commercial Relationship" Outside Arm's Control

Item 3.D's most striking single disclosure is about Arm China: "We utilize our commercial relationship with Arm Technology (China) Co. Limited ('Arm China') to access the PRC market. If that commercial relationship no longer existed or deteriorates, our ability to compete in the PRC market could be materially and adversely affected. Neither we nor SoftBank Group control the operations of Arm China, which operates independently of us."

The financial statements quantify the dependency: "our top five customers (including Arm China) collectively accounted for approximately 56%, 54% and 57% of our total revenue for the fiscal years ended March 31, 2025, 2024 and 2023, respectively, and our largest customer individually, Arm China, accounted for approximately 17%, 21% and 24% of our total revenue, respectively, during those fiscal years."

A single customer that represents 17% of revenue, that the company explicitly states it does not control, that operates in the PRC under increasing US/UK technology export pressure, is an extraordinary structural risk. Arm China's share of revenue is declining (24% → 21% → 17%) — the question is whether that decline reflects Arm's loss of leverage or successful customer diversification.

3. Customer Concentration — Top 4 = 49%, Top 3 = 42%

Per the financial statement disclosure: "the Company had three customers that collectively represented 42% of total revenue, with the single largest customer accounting for 17% of total revenue, the second largest customer accounting for 11% of total revenue, the third largest customer accounting for 11% of total revenue and the fourth largest customer accounting for 10% of total revenue." Adding the fourth largest brings the total to 49%.

Prior years showed even higher concentration: "For the fiscal year ended March 31, 2024, the Company had three customers that collectively represented 42% of total revenue, with the single largest customer accounting for 21%... For the fiscal year ended March 31, 2023, the Company had three customers that collectively represented 44% of total revenue, with the single largest customer accounting for 24%."

Per Item 3.D: "As a result of this customer concentration, we are particularly susceptible to adverse developments affecting our key customers and their respective businesses... this customer concentration also has afforded certain customers significant bargaining power, which has, in some cases, resulted in pricing or other contractual terms that are less favorable to us."

4. Active Litigation With Qualcomm — Trial Set for March 2026

The 20-F provides a detailed update on the Qualcomm/Nuvia dispute under Item 5: "On August 31, 2022, we sued Qualcomm and Nuvia in the U.S. District Court for the District of Delaware... The original claims were narrowed by both parties to the contractual issues arising from the Nuvia ALA and the Qualcomm license and were tried to a jury in December 2024. The jury failed to reach a complete verdict on the three issues presented to it. The jury concluded that certain technology was licensed to Qualcomm under the Qualcomm license and that Qualcomm had not breached the Nuvia ALA but failed to reach a verdict on whether Nuvia breached the Nuvia ALA. Both parties have filed post-trial motions seeking judgment as a matter of law and/or a new trial on the issues that were tried."

A second action by Qualcomm is now pending: "On April 18, 2024, Qualcomm brought a new action in Delaware against us, asserting claims that were rejected for inclusion in the original action. In this new action, Qualcomm asserted that we failed to satisfy certain delivery obligations under the Qualcomm ALA. On December 16, 2024, Qualcomm amended its complaint to add allegations relating to an Arm notice of breach of the Qualcomm ALA and related tort and anti-competition claims... The case is currently set for trial on March 9, 2026."

Qualcomm is one of Arm's largest royalty payers. An adverse outcome that materially altered the Nuvia ALA (or invalidated the Qualcomm license terms Arm has tried to enforce) could affect future royalty streams.

5. RISC-V Free Architecture Is Eating at the Edges

Item 3.D names RISC-V as a structural threat: "We anticipate continued challenges from current and new competitors, including established technologies such as the x86 architecture... as well as by free, open-source technologies, including the RISC-V architecture. Many of our customers are also major supporters of the RISC-V architecture and related technologies. If RISC-V-related technology continues to be developed and market support for RISC-V increases, our customers may choose to utilize this free, open-source architecture instead of our products."

The 20-F flags a specific 2023 development: "in August 2023, a group of our customers and other competitors announced a joint venture aimed at accelerating the adoption of RISC-V, initially focused on the automotive sector." When customers fund a joint venture to develop a free competitor to your core product, the long-term royalty trajectory faces real questions.

6. AI Acceleration Could Shrink the CPU TAM

Item 3.D flags AI as a double-edged sword: "several of our customers are creating their own bespoke accelerators for artificial intelligence ('AI') workloads that are intended to sit alongside the Arm core. If, over time, more of the workloads move to accelerators away from the CPU, the CPU market size may decrease, resulting in a decrease of our license fees and royalties."

7. Foreign Private Issuer Status and Pillar Two Tax Exposure

Item 3.D notes: "While we are a foreign private issuer, we may opt out of certain Nasdaq corporate governance rules applicable to public companies organized in the U.S." Translation: less disclosure than US domestic peers are required to provide.

On taxes: "Pillar Two is effective for the Company for the fiscal year ended March 31, 2025. We have determined that Pillar Two did not have a material impact on the financial statements or results of operations for the fiscal year ended March 31, 2025." The patent box regime that underpins Arm's 10% effective rate could be challenged by future tax reform.

The Critical Audit Matter, in Detail

Deloitte's unqualified opinion identified one critical audit matter — and it lands squarely on the issue this screen flagged. Per the auditor's report: "Revenue Recognition for Material Long-Term Revenue Contracts... The Company routinely enters into material long-term revenue contracts with customers that involve multiple performance obligations consisting of products and services such as intellectual property licenses, implicit rights to future products, early access rights, software licenses, support, training, professional and design services, and other non-standard performance obligations that may be contract or customer specific."

Deloitte continues: "Determining the appropriate revenue recognition for these material long-term revenue contracts may involve significant judgment by management, including the identification of performance obligations and the portion of the transaction price that should be allocated to those performance obligations, both of which impact the amount and timing of revenue recognition... judgments were required when identifying the performance obligations in a given contract... due to the inclusion of complex terms and conditions that may entitle the customer to non-standard products and services, as well as products and services that may be available in future periods or are otherwise implied based on forward looking and historical information. Judgments were likewise required in determining the standalone selling price of certain performance obligations and allocating the transaction price thereto. This is due to a lack of observable standalone selling prices for many of the Company's products and services as well as the inclusion of non-standard offerings for which the standalone selling price must be estimated."

The auditor's procedures included consulting with revenue recognition specialists and reperforming transaction price allocations. The opinion was unqualified — Deloitte signed off on the numbers — but the critical audit matter is unusually direct in identifying revenue recognition as the area requiring the highest auditor judgment.

Summary

Grade: F. Elevated manipulation risk; warrants further investigation, not automatic exclusion.

Arm's FY2025 20-F shows two distinct stories. The growth story is real: revenue up 24% to $4,007 million, gross margin at 97%, operating income from $111M to $831M, and net income tripling to $792M. The balance sheet is pristine — $2.8B in cash against $356M in debt, Z-Score of 8.21 in the safe zone, debt/EBITDA of 0.4x. The IP licensing business is structurally a high-margin, asset-light franchise.

The earnings quality story is troubled. Operating cash flow collapsed 64%, from $1,090M to $397M, while reported net income simultaneously rose 159%. Accounts receivable grew 41.7% against revenue growth of 23.9%. Days Sales Outstanding climbed from 88 days to 101 days. The Beneish M-Score of -1.75 sits 0.03 above the -1.78 manipulation threshold, with three components (DSRI, SGI, DEPI) all flashing the same signal. CapEx jumped 67% — almost 3x revenue growth.

The MD&A and 20-F provide partial explanations. The cash flow collapse was driven significantly by "$876 million in other liabilities due to an increase in cash used for employment taxes payable on vested share-based awards related to the IPO share vesting event" — a non-recurring impact tied to the September 2023 Nasdaq listing. A "$549 million unfavorable cash flow movement related to accounts receivable due to the timing of payments from customers" makes up most of the AR build. These are real explanations. They do not fully eliminate the divergence, and they should normalize in FY26 if the IPO vesting cycle is truly behind the company.

The qualitative risks add weight to the elevated manipulation signal. SoftBank owns 87.1% of the company, leaving Arm structurally controlled and able to opt out of independent-director governance protections. 21% of revenue comes from related parties. Arm China — which the 20-F explicitly states "operates independently of us" — is the single largest customer at 17% of FY2025 revenue. The top three customers represent 42% of revenue, with a fourth customer at 10% bringing the top four to 49%. Litigation with Qualcomm — one of Arm's largest royalty payers — is set for trial on March 9, 2026. RISC-V continues to develop as a free competitor with active backing from Arm's own customers.

Deloitte's unqualified audit opinion includes one critical audit matter: revenue recognition for material long-term revenue contracts. That is precisely the area where the screening engine flagged divergence between revenue recognition and cash collection. Both signals point to the same conclusion: this filing warrants further investigation, particularly through the FY2026 cash flow statement that will show whether the IPO-related distortions truly normalize.

This is not a definitive finding of manipulation. It is a screening grade that says: do not pass this filing without additional diligence on revenue recognition timing, the relationship with Arm China, and the FY2026 normalization of cash conversion. Investors who already hold ARM should monitor the next two quarterly cash flow statements closely. Investors considering a new position should wait for FY2026 data to confirm whether the cash conversion ratio recovers toward or above 1.0x.

**Disclaimer**: This report is based on Arm Holdings plc's FY2025 Form 20-F filed with SEC EDGAR on May 28, 2025. This is NOT investment advice.

Data: SEC EDGAR 20-F + Yahoo Finance

Auditor: Deloitte & Touche LLP (Unqualified opinion, 1 critical audit matter — revenue recognition for material long-term revenue contracts; auditor since 2022)

Fiscal year ended: March 31, 2025

Parent: SoftBank Group (87.1% ownership as of May 20, 2025)