ARCC · Analyze

Date: 2026-08-03 · Price: $19.165 · NAV/share: $19.35 · P/NAV: 0.990x Sector: Financial Services (BDC — externally managed by Ares Capital Management) Verdict: [5.5] HOLD — best-in-class franchise, but recurring earnings have just slipped below the dividend and the stock carries no discount for that.

Data provenance. Every figure that decides the verdict comes from the Q2 2026 Form 10-Q (filed 2026-07-29, period 2026-06-30) and the accompanying 8-K earnings release (Ex. 99.1), pulled directly from EDGAR, plus the Q4-2023/Q4-2024/Q3-2025/Q4-2025 earnings releases for the multi-year series. Vendor feeds were used only for price and peer screening. Per principle-primary-source-beats-vendor, no vendor number is load-bearing here.

⚠️ Two vendor traps confirmed live on this name — do not use these fields: 1. Yahoo payoutRatio = 1.4328 and totalRevenue = $3.106B. The payout ratio is GAAP-EPS-based and is meaningless for a BDC (analysis_notes.md §4 High-Yield overlay: the denominator must be NII). The revenue figure is a net construct — cf. pitfall-vendor-revenue-is-net-construct-for-banks. ARCC's actual total investment income was $768M in Q2 2026 ($3.07B annualised). 2. Yahoo freeCashflow = $759.5M / operatingCashflow = −$680M. FCF is not a meaningful concept for a BDC — portfolio purchases run through operating cash flow, so OCF is structurally negative in any growth year. Ignore §1's FCF ladder entirely for this name; NII is the cash-earnings measure.

---

1. Snapshot

A portfolio-specific passage was removed from the public build.

---

2. Fundamentals — the BDC overlay (analysis_notes.md §4)

2.1 The five mandated BDC tests

Test	Threshold	ARCC	Result
Price / NAV	context	0.990x	✅ Slight discount — but see §5, it is a premium to peers
Base dividend coverage	≥ 100%	97.9%	❌ FAIL
Variable-rate debt	≤ 90%	~74% (78% after the Jan-2031 swap went effective 15 Jul 2026)	✅ PASS, comfortably
TBVPS 5yr CAGR	growth	+3.28%/yr (FY2020 $16.97 → FY2025 $19.94)	🟡 Positive but thin; +0.45%/yr measured to Q2'26
Price / TBVPS	context	0.990x	✅ For a BDC, NAV is tangible book — same number

2.2 Balance sheet and debt structure

Total debt principal $15,924M, carrying value $15,773M, weighted average maturity 4.1 years. Weighted average stated rate (including swaps) is 4.9% on a period-average basis for both the three and six months ended 6/30/26, and 4.8% spot as of the 6/30/26 balance-sheet date — both figures appear in the filing and measure different things. All-in is ~5.3%: $214M of interest and credit facility fees against $16,111M average outstanding debt in Q2. 63.7% of the book is unsecured ($10,150M of notes vs $5,774M of secured facilities and CLOs).

Debt type	$M	Share	Rate basis
Revolving Credit Facility	1,566	9.8%	Floating
Revolving Funding Facility	1,086	6.8%	Floating
SMBC Funding Facility	728	4.6%	Floating
BNP Funding Facility	674	4.2%	Floating
CLO notes/loans (Apr'36, Oct'36, Jan'38)	1,720	10.8%	Floating
Unsecured notes swapped to floating	6,050	38.0%	Floating (via swap)
Unsecured notes remaining fixed	4,100	25.7%	Fixed
Total	15,924	100%	~74% floating

The asset side is 71% variable rate at fair value (12% fixed, 9% non-income-producing, 1% non-accrual, 7% IHAM equity). So ~71% of assets float against ~74–78% of debt — a deliberately well-matched book. This is the single most important structural fact for the rate scenarios in §2.5: ARCC has hedged away most, though not all, of its rate asymmetry. 96% of the variable-rate investments (ex-SDLP) carry interest-rate floors, but with SOFR at 3.50–3.75% those floors only begin to bind around −300bp.

Liquidity is not a constraint: $383M cash, ~$6.7B available under the credit facilities, no remaining 2026 unsecured maturities after the $1B July repayment, and only $1.4B due in 2027.

2.3 Portfolio composition — the drift down the capital structure

Investment type	FV 6/30/26 ($M)	%	FV 12/31/25 ($M)	%	Δ
First lien senior secured	17,333	59.1%	17,858	60.6%	🔻 −1.5pp
Second lien senior secured	1,293	4.4%	1,487	5.0%	🔻
SDLP subordinated certificates	1,154	3.9%	1,117	3.8%	→
Senior subordinated loans	1,822	6.2%	1,585	5.4%	🔺 +0.8pp
Preferred equity	2,348	8.0%	2,475	8.4%	🔻
Ivy Hill (IHAM)	2,843	9.7%	2,434	8.3%	🔺 +1.4pp
Other equity	2,556	8.7%	2,529	8.6%	→
Total	29,349	100%	29,485	100%

Two things deserve attention. First, only 59% of the book is first lien — the "senior secured lender" description is only three-fifths true. Preferred equity, other equity and IHAM together are 26.4% of the portfolio, which is equity risk dressed in a credit vehicle's clothing. Second, the mix drifted the wrong way in six months: first lien down, senior subordinated up, IHAM up. The July post-quarter commitments are starker still — of $244M committed 1–23 July, only 47% was first lien while 17% was second lien and 19% senior subordinated. Reaching for yield by moving down the stack is exactly what a lender does when spreads are tight, and it is the mechanism by which today's yield becomes tomorrow's non-accrual.

2.4 🚩 The marks: the debt book is underwater and equity is masking it

This is the most important thing in the filing and it is not in any vendor feed.

Investment type	Amortised cost ($M)	Fair value ($M)	Mark
First lien senior secured	17,859	17,333	−526 (−2.9%)
Second lien senior secured	1,495	1,293	−202 (−13.5%)
Senior subordinated loans	1,888	1,822	−66 (−3.5%)
Preferred equity	2,605	2,348	−257 (−9.9%)
SDLP certificates	1,140	1,154	+14
Ivy Hill (IHAM)	2,646	2,843	+197
Other equity	2,042	2,556	+514
Total	29,675	29,349	−326 (−1.1%)

Six months earlier the portfolio was marked $235M above cost. It is now $326M below — a $561M swing, and it is the entire explanation for the NAV decline.

Read the columns, not the total: every credit line is marked below cost, and the aggregate is held to −1.1% only by $711M of unrealised gains on IHAM and other equity. IHAM is a wholly owned, unlisted asset manager; "other equity" is Level 3 private equity. Both are internally valued. The credit book is deteriorating and the reported NAV is being cushioned by the two least observable assets on the balance sheet. That is not an accusation of mismarking — ARCC's marks are audited and Rule 2a-5 governed — but it is a real quality-of-NAV issue, and it is the asymmetry to watch if the cycle turns.

2.5 Credit quality — the trend is unambiguous

Date	Non-accrual @ cost	Non-accrual @ FV	Grade 1+2 (% of FV)
Dec 2023	1.3%	—	—
Dec 2024	1.7%	1.0%	—
Sep 2025	1.8%	—	—
Dec 2025	1.8%	1.2%	3.8%
Mar 2026	2.1%	—	—
Jun 2026	2.4%	1.4%	5.4%

Non-accruals have risen in every observation for three years and the rate of increase is accelerating — flat at ~1.8% through 2025, then +0.3pp in each of the last two quarters. The internal grades corroborate and lead: Grade 2 ("risk materially increased since origination") jumped from $675M to $1,089M, +61% in six months, and Grade 1+2 combined went 3.8% → 5.4% of fair value. Grade-2 assets are the feeder pool for next year's non-accruals.

Context in ARCC's favour, and it is genuine: 2.4% is still below ARCC's own ~3% post-GFC average and the ~4% BDC sector average. Management is right that the level is low. But the CEO's press-release line — "historically low levels of non-accruing loans and problem assets" — describes the level while saying nothing about a direction that has been one-way for three years. That gap between framing and disclosure is itself a data point.

On the call, management named the mechanism honestly when pressed. President Jim Miller: "we have been operating for an extended period of time where non-accrual rates and default experience has been below the long-term average… we think there is a reversion towards the mean." Asked whether anything specific was deteriorating inside ARCC's book, he pointed back to mean reversion rather than the portfolio. CEO Kort Schnabel declined to discuss two credits an analyst named as sliding, but volunteered that they share a sponsor and that ARCC is "engaging in a lot of ongoing and constructive dialogue with the sponsor… confidence in our ability to address the near-term maturities in those two names." That is workout language.

One external discrepancy worth flagging and not resolved: management said four new names went to non-accrual in Q2; PitchBook counted five borrowers, led by AmeriVet Veterinary Partners ($84.6M subordinated, marked ~62c) — a credit that FS KKR had already placed on non-accrual two quarters earlier, in Q4 2025. The count difference is probably a names-vs-loans artifact. The AmeriVet timing is the sharper item: it sits awkwardly beside Schnabel's claim that ARCC has "been proven to be on the more conservative end of the spectrum" on marks.

2.6 Per-share record and economic return

For a BDC the owner's-eye measure is total economic return = NAV change + dividends, not revenue or FCF.

Year	NAV start	NAV end	Δ NAV	Dividends	Economic return
FY2023	$18.40	$19.24	+$0.84	$1.92	15.0%
FY2024	$19.24	$19.89	+$0.65	$1.92	13.4%
FY2025	$19.89	$19.94	+$0.05	$1.92	9.9%
H1 2026	$19.94	$19.35	−$0.59	$0.96	1.9% (≈3.7% ann.)

A clean, monotonic decay: 15.0% → 13.4% → 9.9% → ~3.7%. The dividend has been constant throughout; the entire deterioration is NAV. This is the honest summary of the last three years — ARCC has been paying out a ~10% yield while its book value stopped compounding.

Share count: 467.7M (FY2021) → 718.1M (FY2025), +11.3%/yr. That dilution was accretive while the stock traded at a 5–20% premium to NAV, which it did through 2024 and most of 2025 — issuing above book adds NAV per share. At 0.99x NAV that arithmetic reverses, and management has correctly stopped issuing: shares were flat at 718M through H1 2026. That is genuine capital discipline and it deserves credit. The $1.0B buyback authorisation, however, remains entirely unused — at 0.99x NAV they are neither issuing nor repurchasing.

---

3. 🔴 Dividend coverage & sustainability — the mandated analysis

This is the section analysis_notes.md §4 requires explicitly for both the BDC and High-Yield overlays. The correct denominator is Core EPS (net investment income excluding realised/unrealised gains and the GAAP-only capital-gains incentive fee accrual) — not GAAP EPS, and not FCF.

3.1 A GAAP artefact was flattering reported NII — and I caught it before reconciling

Working from the income statement alone, reported GAAP NII for Q2 2026 was $359M, or $0.50/share — apparently 1.04x covered. But that line contains a −$21M credit for the capital gains incentive fee (−$82M for the half-year). Note 3 is explicit: "As of June 30, 2026 and December 31, 2025, there was no capital gains incentive fee actually payable." It is a GAAP accrual that reverses precisely because the portfolio is being marked down — the worse the marks, the bigger the boost to reported NII. Stripping it gives $338M, or $0.471/share.

That derived figure reconciles to management's own reported Core EPS of $0.47 — an independent cross-check that the adjustment is the right one. Anyone reading GAAP NII off the income statement would have concluded the dividend was covered. It is not.

3.2 Coverage trend — 10 quarters, not a snapshot

Quarter	Core EPS	Dividend	Coverage
Q3 2024	$0.58	$0.48	1.21x
Q4 2024	$0.55	$0.48	1.15x
Q1 2025	$0.50	$0.48	1.04x
Q2 2025	$0.50	$0.48	1.04x
Q3 2025	$0.50	$0.48	1.04x
Q4 2025	$0.50	$0.48	1.04x
Q1 2026	$0.47	$0.48	0.98x
Q2 2026	$0.47	$0.48	0.98x

Fiscal year	Core EPS	Dividends	Coverage	Payout ratio
FY2023	$2.37	$1.92	1.23x	81.0%
FY2024	$2.33	$1.92	1.21x	82.4%
FY2025	$2.01	$1.92	1.05x	95.5%
FY2026 run-rate	$1.88	$1.92	0.98x	102.1%

Direction over level, per the overlay: coverage has fallen in a straight line for three years — 1.23x → 1.21x → 1.05x → 0.98x — and has now crossed below 1.0x. Management's counter is the trailing-twelve-month figure: $0.50 + $0.50 + $0.47 + $0.47 = $1.94 vs $1.92, or 101%. That is arithmetically true and razor-thin, and it rolls over unfavourably next quarter: when Q3'25's $0.50 drops out, a Q3'26 print of $0.47 takes TTM to ~$1.91 — below the dividend on management's own preferred measure.

3.3 Base vs supplemental — a genuine structural strength

ARCC's payout is 100% base. There is no supplemental component and has not been one since at least FY2024 — every quarter in the filing's three-year table shows exactly $0.48. This matters, and it cuts in ARCC's favour: peers such as TSLX and OBDC report headline yields inflated by variable supplementals that can be switched off without "cutting the dividend." ARCC has no such cushion to hide behind, but equally there is no headline yield here that overstates the commitment. The 10.0% is all base. Combined with 68 consecutive quarters of stable-or-rising payments, this is a dividend defended by policy and precedent, not a formula that auto-adjusts downward.

3.4 What is actually funding the dividend

Source (H1 2026, per share)	Amount
Core NII	$0.94
Net realised gains	+$0.14
Total available	$1.08
Dividends paid	$(0.96)
Surplus	+$0.12

Management said as much on the call: TTM core exceeded the dividend "while an additional $0.15 per share of net realised gains has provided further support." So the dividend is currently covered — but by recurring interest income plus realised gains, not by recurring interest income alone. Realised gains are not repeatable on demand, and the remaining unrealised gains are concentrated in exactly the two assets flagged in §2.4 (IHAM +$197M, other equity +$514M). Harvesting them to fund distributions converts a NAV cushion into income — it works, but it is a drawdown of the buffer, not earnings.

Cash quality: PIK income was $121M of $768M total investment income in Q2 2026 — 15.8%, non-cash, against a dividend that is 100% cash. (Down from 17.7% a year ago, so the direction is mildly favourable.) Elevated PIK is normal for a BDC with a subordinated sleeve and does eventually convert to cash on repayment, but it means roughly a sixth of reported income is not cash in the quarter it is earned.

Spillover: $988M, or $1.38/share — 2.9 quarters of dividend. This is a real buffer and it is why there is no near-term risk. Management named its purpose plainly: spillover "can help bridge during periods of slower transaction activity." That is a company describing its own dividend as needing a bridge.

3.5 Stress tests

Rate scenarios. ARCC's own Item 3 disclosure (annualised, from the 6/30/26 balance sheet, considering interest-rate floors) — note the footnote: it excludes the income-based fee, which absorbs ~20% of any swing, so I apply that offset.

Scenario	Δ Net income (disclosed, pre-fee)	After ~20% fee offset	Δ per share	Core EPS	Coverage
Current run-rate	—	—	—	$1.88	0.98x
−100bp	−$93M	−$74M	−$0.104	$1.78	0.93x
−200bp	−$180M	−$144M	−$0.201	$1.68	0.87x
+100bp	+$94M	+$75M	+$0.105	$1.99	1.03x

A portfolio-specific passage was removed from the public build.

Non-accruals @ cost	Δ per share	Core EPS	Coverage
2.4% (today)	—	$1.88	0.98x
3.0% (ARCC's post-GFC average)	−$0.020	$1.86	0.97x
4.0% (BDC sector average)	−$0.055	$1.83	0.95x
5.0%	−$0.089	$1.79	0.93x

Combined bear case — the realistic stress the overlay demands: a full easing cycle (−200bp) and credit normalising to 5% non-accruals gives Core EPS of $1.59 → 0.83x coverage. Holding coverage at 1.0x from there requires cutting the dividend to ~$0.40/quarter — a 17% cut, almost exactly what OBDC (−16%) and Golub (−15%) have already done.

A milder, arguably more likely path (−100bp, non-accruals to 3.5%) gives $1.74 → 0.91x, implying a ~9% cut to restore coverage.

The probability weighting matters here and it currently favours ARCC. The FOMC held at 3.50–3.75% on 29 July 2026 by 9–3, with all three dissents wanting higher rates, and markets are pricing some chance of hikes rather than cuts into year-end. −200bp is a tail scenario over the next twelve months, not a base case. Management is explicit that they want the base rate up: "Even a little bit of an outlook toward a slight widening of base rates going forward, which should be a benefit to us." Over a 3–5 year holding period, however, a full easing cycle is close to inevitable — which is precisely the horizon this position is held on.

3.6 ⚖️ Verdict on the dividend — stated plainly, as required

The dividend is COVERED — but only just, only with help, and the trend is one-way.

Specifically, and separating the three questions the overlay says must not be blurred:

1. Is it covered by recurring earnings? No. Core NII run-rate is $1.88 against a $1.92 payout — 0.98x. It has fallen every year for three years and crossed below 1.0x two quarters ago.
2. Is it covered in total? Yes. Core NII plus realised gains gives 1.13x for H1 2026, TTM core is 101%, and there is $1.38/share of spillover — 2.9 quarters — behind it. A cut in the next twelve months is unlikely, and I would not position for one: the board has 68 quarters of precedent, the Fed is on hold with a hawkish tail, and the competitive backdrop (§4) is improving.
3. Is it being funded from elsewhere? Partly — yes. Realised gains contributed $0.14–0.15/share, and 15.8% of investment income is non-cash PIK. The marginal dollar of the dividend is now coming from gains harvested out of the equity sleeve rather than from interest collected on loans.

The honest characterisation: this is no longer a dividend covered by the lending business. It is a dividend covered by the lending business plus asset sales plus a spillover buffer. It is safe for now and at real risk in an easing cycle. Per the overlay's final instruction — dividend safety and thesis safety are different questions, and here the answer is that the dividend is probably safe while the thesis is quietly deteriorating. That combination is not a buy case.

---

4. Moat — what actually protects a BDC

Moat rating: 2/5 — narrow. Efficient scale and borrower incumbency are the only sources that survive scrutiny. The peer benchmarking below is primary-source verified (Q1 2026 10-Q/8-K for peers, Q2 2026 for ARCC).

Generic moat language is useless for a BDC. There are only four candidates that can matter, and they are testable.

1. Cost of funds — the claim survives, but in a weaker and more specific form than the bull case states.

ARCC borrows at a 4.9% weighted average stated rate for the period (4.8% spot at 6/30/26), ~5.3% all-in including credit facility fees ($214M ÷ $16,111M average debt), with a 4.1-year weighted average maturity and 63.7% unsecured — verified from the 10-Q debt table ($10,150M of unsecured notes against $5,774M of secured facilities and CLOs).

BDC	WA cost of debt	% unsecured
BXSL	4.83% (4.90% all-in)	55.7%
ARCC	4.9% stated / ~5.3% all-in	63.7%
OBDC	5.2%	53.0%
GBDC	5.2%	50.9%
FSK	5.26%	51%
MAIN	5.5%	71% notes-only / 84.8% incl. SBIC
TSLX	5.5%	68%
PSEC	6.05% all-in / 5.06% stated	76.9%

Peer figures are Q1 2026 (period ended 3/31/26); ARCC is the only one of the eight to have reported Q2 yet, so it carries a one-quarter timing advantage here. The table also mixes stated and all-in bases — compare like with like.

So: second-cheapest of eight, and the cheapest among externally managed peers at comparable scale — not "best-in-class." BXSL is the honest counter-example and it matters, because the reason is instructive. ARCC is genuinely the highest-rated BDC across all three agencies (Baa2/BBB/BBB — the only name carrying a full BBB at S&P rather than BBB−), yet it funds at roughly the same cost as BXSL, which sits one S&P notch below. BXSL's 97.6% first-lien book makes its secured borrowing cheap enough to close the gap. Sharper still: ARCC's marginal unsecured issue (5.550% coupon, T+177, swapped to SOFR+170) prices roughly level with its own secured revolver (SOFR+152.5/165/177.5) — the ratings edge is not converting into a pricing edge at the margin.

The correct claim is therefore a funding resilience edge, not a funding cost edge. 63.7% unsecured, a 4.1-year ladder, $6.0B of liquidity and 186% asset coverage mean ARCC's funding cannot be gated — no borrowing-base test, no lender vote, no redemption queue. That is worth a great deal in a credit event and essentially nothing in a spread war. It is a real advantage, correctly located.

The commercial paper programme is marketing, and it quantifies as such: 50–100bp on a $1.0B cap is ~6% of total debt, i.e. 3–6bp blended — and it introduces short-dated rollover risk into an otherwise 4.1-year ladder. Being first in the sector is a genuine signal of institutional access; the P&L effect is a rounding error.

4.0 Peer credit ratings — a sector-wide signal that cuts against the valuation

BDC	Moody's	S&P	Fitch	KBRA
ARCC	Baa2	BBB	BBB	—
OBDC	Baa2 Stable	BBB− Stable	BBB Stable	BBB+ Stable
TSLX	Baa2 Stable (upgraded from Baa3)	unverified	unverified	BBB+ Stable
BXSL	Baa2, outlook CUT TO NEGATIVE 6/3/26	BBB− Positive	BBB Stable	—
GBDC	Baa2, outlook CUT TO NEGATIVE ~6/1–3/26	BBB−/BBB (mapping unverified)	BBB/BBB− (unverified)	—
MAIN	NOT RATED	BBB− Stable	BBB− Stable	—
FSK	Ba1 Stable — CUT FROM Baa3 3/23/26 (lost IG)	unverified	BB+ Negative, cut from BBB− 4/9/26	BBB− Stable, cut from BBB
PSEC	unverified	BB+ (secondary source)	unverified	BB+ Negative, cut from BBB−

Read this as a credit-cycle signal, not a scorecard. In 2026 alone, two peers lost investment grade (FSK in March, PSEC) and two more took negative outlooks in June (BXSL, GBDC) that postdate and supersede the "stable" shown in their own Q1 decks. ARCC is the highest-rated name in the group and that is real.

But it cuts both ways, and the second edge is the one that matters for the verdict: the agencies are independently corroborating that BDC credit is deteriorating broadly — which is exactly the finding in §2.5 and §3. If the whole field is being downgraded while ARCC trades at 0.99x NAV against OBDC's 0.77x, BXSL's 0.90x and FSK's 0.59x, the premium is harder to justify, not easier. Quality is not in question; the price of quality is.

2. Origination scale and sourcing. $29.3B across 619 portfolio companies and 273 distinct private equity sponsors, with the ability to underwrite and hold very large unitranches solo — $13.1B of the first lien book is unitranche, and management cited committing 100% of a ~$2B facility in the quarter. The Ares platform (~$500B AUM) supplies cross-asset credit intelligence a standalone BDC cannot replicate. Real, but a scale advantage rather than a monopoly — Blackstone, Blue Owl, Golub, HPS/BlackRock and Apollo all clear the same bar.

Incumbency is the second genuine moat source, and it is double-edged. 75% of Q2 transactions were with existing borrowers — an existing lender sees the add-on, the recap and the refinancing first, and prices them with information no newcomer has. That is a moat when you can decline the follow-on. It is a trap when declining means impairing what you already hold — and with two named credits under one sponsor now in maturity-extension talks (§2.5), that is not a hypothetical distinction.

⚠️ Two standard bull points are simply wrong and should be struck from the thesis: - There is no index-inclusion moat — there is an index penalty. BDCs were removed from the major indices in 2014 over the acquired-fund-fees-and-expenses (AFFE) rule, which forces funds holding them to report BDC expenses as their own. That structurally suppresses institutional ownership (ARCC's is only 36%). A legislative fix exists — H.R.2225 passed the House 6/23/25 and is stalled in the Senate — and would be a genuine re-rating catalyst if it moved. Today it is a headwind, not a tailwind. - The premium-to-NAV issuance flywheel is switched off. Issuing equity above book is accretive and drove much of ARCC's history — but at 0.99x NAV that engine is off, which is precisely why share count has been flat at 718M (§2.6). A moat that only operates at a premium is not operating.

3. Permanent capital. A listed BDC cannot be redeemed. This is precisely the advantage that is paying off right now: the non-traded/perpetual vehicles that flooded the market with cheap capital are facing severe redemptions (Blue Owl Credit Income holders requested 21.9% of shares in Q1; Blue Owl Technology Income 40.7%), forcing them to retreat from the upper middle market. Structurally durable and currently valuable.

4. Ivy Hill (IHAM) — a genuine differentiator, and a genuine complication. A wholly owned SEC-registered asset manager with $16.3B AUM, earning third-party fee income ($15M management/incentive fees in Q2) and giving ARCC a captive outlet to sell down loans ($1.1B sold to IHAM vehicles in Q2 alone). No peer has an equivalent. But it is also 9.7% of the portfolio, unlisted, internally valued, carried $197M above cost, and growing fast (subordinated loan to IHAM more than doubled, $531M → $946M in six months). It is simultaneously a moat and the least transparent asset on the balance sheet.

4.1 Adversarial stress test — attacking the thesis properly

"Spread compression will crush the returns." The consensus attack, and the decomposition shows it is aimed at the wrong variable. Of the 220bp decline in portfolio yield since YE2023, 171bp (78%) is the base rate and only ~49bp is spread and mix — and residual spread has been flat-to-up since YE2024. ARCC's own Q2 senior spreads were 20bp wider than Q4 2025 with upfront fees 50bp higher, as redemption-driven retreat by retail-funded competitors thinned the field of lenders who can write large cheques: "the number of players that can compete with scale like Ares has shrunk a bit." Post-quarter commitments (1–23 July) funded at a 10.2% yield against exits at 8.3%.

Verdict: the attack fails as stated — but it succeeds in a slower form that is easy to miss. The damage is not in today's spread, it is embedded and forward: new commitments are going on at 9.4% against a 10.3% book, i.e. ~90bp of decay locked into every dollar that turns over. Across a 4–5 year portfolio turnover that is roughly a ~3%/yr earnings drag — a slow bleed rather than a cliff, and the mechanism that quietly moved coverage from 1.23x to 0.98x. Note also the composition caveat in §2.3: part of the headline yield is bought by moving down the capital structure, not by better pricing.

"The banks are back and will refinance unitranches away." Partially true at the large end, and it shows up as elevated repayments — $3.0B of exits against $2.9B of new commitments in Q2, i.e. the portfolio shrank slightly. ARCC's Q2 closing ratio was "moderately below our historical average approximately 5%" because "the quality of the flow dipped a little bit." This is the live constraint on growth, and it is why NII per share is flat-to-down despite a stable balance sheet. It compresses volume, not spread.

"External management is a permanent tax on shareholders." The most substantive attack, it lands hardest, and it is the single largest number in the file.

Q2 fees: base management fee $110M + income-based fee $84M = $194M, against $768M of investment income.

Fee drag, measured four ways
Annualised	$776M
Per share	$1.08 — more than half the dividend
% of gross assets	2.54%
% of NAV	5.6%
% of gross investment income	25%
% of pre-fee earnings	~36%

The cleanest statement of the problem: in 1H 2026 Ares Management was paid $307M while ARCC shareholders earned $263M of GAAP net income. The manager out-earned the owners. For scale, MAIN — internally managed — runs total operating expense ex-interest at 1.3–1.4% of assets, roughly half.

Three structural defects, all visible in the advisory agreement itself: 1. No total-return hurdle and no lookback — fees are payable in a loss quarter, and Q2 2026 was one (net unrealised losses of $183M). 2. The fee base includes non-cash PIK and is explicitly non-refundable per the agreement's own language. Against $121M of Q2 PIK income, ARCC paid roughly $24M per quarter in fees on income it has not collected and may never collect. 3. The base fee scales with gross assets, rewarding balance-sheet growth independent of per-share outcomes. The record shows it: shares outstanding +129% since YE2016 while NAV per share rose only 17.6%.

And this is now being litigated: a derivative action under Section 36(b) of the Investment Company Act (Siegel, S.D.N.Y., filed 2026-05-26) alleges the adviser received excessive advisory fees and seeks disgorgement plus rescission of the advisory agreement. It is preliminary and 36(b) claims rarely succeed, but it names the exact conflict. This is the strongest structural argument for preferring an internally managed BDC such as MAIN, and it belongs in the verdict rather than the footnotes.

"Adverse selection — the good paper leaves and the bad paper stays." Not proven yet, but the mechanism is visible and this is the attack I weight highest.

Exits in Q2 ran at an 8.3–9.1% yield while the retained portfolio yields 10.3% — the loans being repaid are the cheaper, safer ones, because healthy borrowers refinance into the reopened syndicated market and troubled ones cannot. Meanwhile the book is drifting down the capital structure:

Structural risk marker	ARCC	Comparator
First lien % of portfolio	59% (from 61%)	BXSL 97.6%
Equity & equity-like	~26% of portfolio ≈ 56% of NAV	—
PIK as % of investment income	15.8%	~8% BDC average

The honest read: ARCC is holding its headline yield flat by taking more structural risk and by accruing rather than collecting. In fairness, the underlying credit statistics are still sound — weighted average interest coverage 2.2x, LTV in the low-40s, weighted average grade unchanged at 3.1 — so this is a developing risk, not a realised one. The two numbers to track quarterly are first-lien percentage and the PIK ratio; both moved the wrong way this quarter, and the rising non-accrual and Grade-2 series in §2.5 is consistent with the dynamic already being underway.

4.2 Disruption forecast and evergreen assessment

Private credit as an asset class has attracted enough capital to compete away much of its excess return, and 2026 is the first year the sector has faced a genuine test: the first-ever net outflow, capital formation −40% YoY, Moody's downgrading its outlook on the ~$400B BDC sector on redemption pressure, and the SEC's 2026 exam priorities explicitly targeting private-market valuation practices. The "no mark-to-market" advantage of private credit has never been tested by a real default cycle; §2.4 shows what happens to reported NAV when it starts to be.

ARCC is well positioned within a field that is getting harder. Consolidation is a plausible upside — asked about acquiring a distressed peer, Schnabel volunteered that "given the dispersion in performance that we are seeing across managers… the likelihood of a transaction like that occurring is probably higher than it is been in the past."

A portfolio-specific passage was removed from the public build.

---

5. Valuation

For a BDC the framework's ordering inverts: Price/NAV, NII yield and the coverage work carry the weight; Graham is close to meaningless.

5.1 Price / NAV — the primary anchor

$19.165 / $19.35 = 0.990x.

The de-rating is real and recent. From the filing's own quarterly table, ARCC's closing price ranged from a +0.15% to +20.13% premium to NAV across 2024 and the first three quarters of 2025. Since Q4 2025 it has traded persistently at a discount (Q1'26 low −10.92%, Q2'26 high exactly 0%). The market has already repriced ARCC from premium to par.

But par is not cheap relative to the field:

BDC	Price	P/NAV	Notes
ARCC	$19.17	0.990x
MAIN	$55.84	1.67x	Internally managed — structurally deserves the premium
TSLX	$17.75	1.09x	Cut base dividend Q1'26
BXSL	$23.64	0.90x
OBDC	$11.03	0.765x	Cut base dividend 16%
FSK	$11.10	0.59x	Negative TTM EPS — distressed

ARCC trades at a premium to every externally managed peer except TSLX. The market is still awarding it a quality premium. That premium is defensible on the cost-of-funds and platform evidence in §4 — but it means there is no discount embedded for the coverage gap, the rising non-accruals or the NAV-quality issue in §2.4. You are paying full freight for the best house in a neighbourhood that is getting worse.

5.2 Dividend Yield Theory — run through the pitfall guard, then discarded

Per pitfall-dyt-inverts-when-price-caused-the-yield, DYT must not be reported before decomposing why the yield is high. ARCC is exactly the profile that note warns about — a double-digit yield on a name down 16% over twelve months.

Guard question	Answer	Disqualifying?
Why did the yield rise?	Entirely price. The dividend has been $0.48/quarter — unchanged — for at least 14 consecutive quarters. The price fell from ~$22–23 (2025) to $19.17. The numerator has not moved at all.	❌ Yes — "mostly price"
Is the payout policy progressive or formulaic?	Discretionary, board-set quarterly, 68 consecutive quarters stable-or-rising. Not a fixed payout ratio; it will not auto-decline with earnings.	✅ No — this one passes, and it is a meaningful point in ARCC's favour
Was the historical band set during a boom?	Partly yes. The 9.0% five-year average yield spans 2021–26, including the 2023–25 period when SOFR above 5% inflated BDC NII across the board. With SOFR at 3.50–3.75%, the earnings that supported a 9% yield are structurally lower.	❌ Yes — rate-regime artefact
Does cash earnings cover the dividend?	No — 0.98x on core NII (§3).	❌ Yes

Three of four answers are disqualifying. The naive DYT number — $1.92 ÷ 0.0900 = $21.33, implying +11% upside — is reported here solely so the reasoning is on the record, and is discarded. Mechanically weighting it would have pulled the fair value above the current price on the strength of a yield that is high only because the stock fell.

The repair, per the note: reset the band to the peer class ARCC is becoming, not the one it occupied at 5% SOFR. The relevant comparators are large externally managed BDCs with marginal coverage in a normalising credit cycle — TSLX at 9.8% (post-cut), OBDC ~11.2% base (post-cut), BXSL ~11–13%. MAIN's 8.05% is the wrong comp (internally managed, 1.67x NAV). A reset band of 10.0–11.5% gives $1.92 ÷ 0.115 = $16.70 to $1.92 ÷ 0.100 = $19.20 → $16.70–$19.20. That bracket contains the current price at its very top, and it agrees with the NAV and DDM anchors — which is the signature of the reset version being the right one.

5.3 DDM

The dividend has not grown in 14+ quarters and coverage is below 1.0x, so a growth assumption above ~1% is not defensible. Required return for a 1.12x-levered private-credit vehicle in a normalising credit cycle: 10.5–12%.

r \ g	g = 0%	g = 1%
10.5%	$18.29	$20.21
11.0%	$17.45	$19.20
12.0%	$16.00	$17.45

Range $16.00–$20.20, centred ~$18.20.

5.4 Graham and Bogle

Graham √(22.5 × EPS × BVPS) = √(22.5 × 1.35 × 19.35) = $24.24 on TTM GAAP EPS, or $29.06 on core EPS. Both are discarded. Graham's formula assumes an operating company where book value and earnings are independent sources of value. For a BDC the book value is the earning asset and the earnings are simply the yield on it — the formula multiplies a number by a function of itself and double-counts. It also runs on GAAP EPS, which for ARCC swings with unrealised marks ($0.13 in Q1'26, $0.24 in Q2'26). Reported for completeness only; zero weight.

Bogle expected return = dividend yield + earnings growth ± multiple change = 10.0% + ~0% (core EPS declining ~6% YoY, offset by portfolio growth) + ~0.5%/yr (modest re-rating room from 0.99x toward ~1.02x) ≈ ~10.5%/yr, with a realistic band of 7% (a 15% dividend cut, no re-rating) to 12%. This is the most useful single framing: ARCC is priced to deliver roughly its coupon and little else.

5.5 Weighted fair value

Model	Output	Weight	Rationale
Price / NAV (0.92–1.02x × $19.35)	$17.80–$19.75	Highest	The correct primary anchor for a BDC
DDM (g 0–1%, r 10.5–12%)	$16.00–$20.20	High	Flat, well-defended dividend suits DDM
DYT — reset band (10.0–11.5%)	$16.70–$19.20	Medium	Post-guard version
DYT — naive (9.0% 5yr avg)	~~$21.33~~	Zero — discarded	3 of 4 guard tests failed
Graham	~~$24.24~~	Zero — discarded	Structurally inapplicable to a BDC
Bogle	~10.5%/yr expected return	Context	Not a fair value

Weighted fair value: $17.50 – $19.50 (midpoint ~$18.50)

At $19.165 the stock trades at the top of its fair-value range. It is not expensive and it is not cheap. There is no margin of safety for the coverage gap, the three-year non-accrual trend, or the NAV-quality asymmetry — and by the framework's core principle, a good business at full price is a wait, not a buy.

For reference, the analyst mean target is $20.69 (13 analysts, 4 strong buy / 7 buy / 3 hold). My range is deliberately more conservative. Worth noting how those targets have moved: every single rating action in the last ten months lowered the price target — seven cuts, zero raises, including JP Morgan to $18.50 (below spot) and a Wells Fargo downgrade to Equal-Weight in June. Ratings intact, targets sliding, is what analysts do when NAV erodes but they are not ready to abandon a thesis.

---

6. Sentiment

Q2 2026 (reported 2026-07-29) met consensus on Core EPS at $0.47 and marginally missed on total investment income ($768M vs $769M). The market shrugged: −1.73% on the print, fully recovered within three sessions, and +2.16% today to $19.165 — above the pre-earnings level. This was not a repricing.

Insider activity — verified against the pitfall, and genuinely positive. Per pitfall-yahoo-insider-purchases-counts-rsu-grants, the insider_purchases summary was not used; transaction rows were read individually for price and description.

Date	Insider	Role	Shares	Price	Assessment
2026-02-09	Jana Markowicz	COO	15,000	$19.20	Open-market buy, $288K
2026-02-06	Scott Lem	CFO	5,186	$19.29	Open-market buy, $100K
2026-02-05	Mary Beth Henson	Director	4,000	$19.14	Open-market buy, $77K
2026-02-05	Kort Schnabel	CEO	12,500 ×2	$19.13	Open-market buys, $478K
2025-10-31	Kort Schnabel	CEO	13,000	$20.39	Open-market buy, $265K
2025-03-03	James Miller	President	40,000	$23.32	Open-market buy, $933K
2025-02-11	Ann Torre Bates	Director	6,000	$22.75	Open-market buy, $137K
2025-02-18	M. Arougheti	Officer/Dir	300,172	$0.00 "Stock Gift"	Not a purchase — excluded

Cash paid at prevailing market prices, across CEO, President, CFO, COO and two directors, with zero sales anywhere in the dataset back to February 2025. This is the unambiguous form real insider buying takes, and the February 2026 cluster was bought at $19.13–$19.29 — essentially today's price. Two honest caveats: nothing since 2026-02-09, including no buying during the subsequent slide to the $17.40 low or around the Q2 print; and Yahoo drops the 10b5-1 field, so pre-scheduling cannot be ruled out without Form 4 footnotes (not pulled).

Sector backdrop is materially worse than ARCC's own numbers — and this is the main reason the coverage question matters. OBDC cut its base dividend 16% (originations −41% YoY) and Golub cut 15%, with analysts modelling further cuts. The BDC sector posted its first-ever net outflow, capital formation fell 40% YoY, and Moody's downgraded its outlook on the sector. ARCC is now one of the holdouts still paying an uncut dividend at sub-1.0x recurring coverage — which is precisely why its 98% figure attracts scrutiny.

Short interest is 6.31% of shares outstanding (45.3M shares, 8.5-day cover) — elevated for a large-cap income name, though some is likely hedging rather than directional.

---

7. Phase 3 — the tension, and how I weight it

There is a real disagreement between the credit picture and the competitive picture, and it should not be smoothed over.

Fundamentals says: coverage has fallen for ten straight quarters and is now below 1.0x; non-accruals have risen for three years and are accelerating; every credit line is marked below cost with NAV propped by Level 3 equity; economic return has decayed from 15% to ~4%. On this evidence the franchise is slowly eroding.

Moat/Sentiment says: spreads are 20bp wider, fees 50bp higher, new money is going out at 10.2% against 8.3% exits, competitors are in redemption-driven retreat, the Fed is on hold with a hawkish tail (a tailwind to floating-rate NII), and the CEO, President, CFO and COO all bought stock with cash at today's price. On this evidence the next twelve months should be better, not worse.

Both are true, and they operate on different clocks. The competitive improvement is cyclical and immediate — it should show up in Q3 and Q4 2026 originations and could well push core EPS back above $0.48. The credit deterioration is structural and lagging — Grade-2 assets rose 61% in six months and those convert to non-accruals over the following twelve to twenty-four months, and the adverse-selection dynamic in §4.1 (cheap loans repaying, expensive ones staying) works quietly in the background regardless of how good this quarter's spreads are.

I weight the credit trend higher, for three reasons. First, it is the one that ends in a dividend cut, and the dividend is the entire reason to own this. Second, it has been monotonic for three years while the spread improvement is two quarters old and partly bought by moving down the capital structure. Third — and decisively — the valuation gives no compensation for being wrong. At 0.99x NAV, a premium to nearly every external peer, the improving-competition case is already the price. If it plays out, you earn the coupon; if the credit case plays out, you take a NAV mark and a distribution cut.

The insider buying is the strongest single argument against my caution and I do not dismiss it — but it is six months stale and was executed above the subsequent trading range.

---

8. Verdict

[5.5] HOLD — quality franchise, fair price, deteriorating coverage. Do not add here.

ARCC is the best-run large BDC and the evidence supports that: the highest credit ratings in the sector (Baa2/BBB/BBB), a 4.9% cost of funds that is second-cheapest of eight peers, 63.7% unsecured funding on a 4.1-year ladder that cannot be gated, $29.3B across 619 borrowers and 273 sponsors, permanent capital while rivals face 20–40% redemption requests, a unique captive asset manager in Ivy Hill, and 68 consecutive quarters of stable-or-rising dividends. Management has stopped issuing equity now that the stock trades at NAV, which is the right discipline. The near-term competitive environment is genuinely improving in its favour.

Two bull points should nonetheless be struck: the funding advantage is resilience, not price — BXSL funds cheaper at 4.83% despite a worse rating, and ARCC's marginal unsecured issue prices level with its own secured revolver. And the moat is narrow (2/5): efficient scale and borrower incumbency, nothing more. There is no index-inclusion moat — BDCs were removed from the indices in 2014 over AFFE and the fix (H.R.2225) is stalled in the Senate — and the premium-to-NAV issuance flywheel is switched off at 0.99x book.

But the thing you buy it for has quietly weakened. Recurring earnings no longer cover the dividend — 0.98x, down from 1.23x three years ago in an unbroken line — and the gap is being filled with realised gains and a spillover buffer that management itself describes as a "bridge." Non-accruals have risen for three consecutive years and are accelerating, with Grade-2 assets up 61% in six months feeding the pipeline. Every credit line in the portfolio is marked below cost, with reported NAV held up by $711M of unrealised gains on the two least observable assets on the balance sheet. The manager out-earned the owners in 1H 2026 — $307M of fees against $263M of shareholder net income. And NAV per share has compounded at ~3.3%/yr over five years — this is a rate-cycle income instrument, not a compounder, and should be sized as one.

At 0.99x NAV — a premium to OBDC (0.77x), BXSL (0.90x) and FSK (0.59x) — none of that is in the price. The rating agencies sharpen the point rather than soften it: two peers lost investment grade in 2026 and two more went to negative outlook in June, which corroborates that BDC credit is deteriorating across the field. ARCC deserves to be the highest-rated name in that field; it does not obviously deserve to be the most expensive one relative to book while its own coverage sits below 1.0x. The framework's first principle is that a good business at full price is a wait, and that is where this sits.

Key risks (named)

1. Rate cuts are the primary threat to the dividend. −200bp takes coverage to 0.87x standalone; combined with credit normalising to 5% non-accruals it reaches 0.83x, implying a ~17% cut — the same magnitude OBDC and Golub have already taken. Mitigant: the Fed held 9–3 on 29 July with all dissents hawkish, so this is a 2027+ risk, not a 2026 one.
2. Credit normalisation is already underway. 1.3% → 1.7% → 1.8% → 2.1% → 2.4% non-accruals; Grade 1+2 at 5.4% of fair value. Two named credits under the same sponsor are in maturity-extension talks. A peer non-accrued AmeriVet two quarters before ARCC did.
3. NAV quality. The credit book is $1.05B below cost in aggregate; only $711M of Level 3 equity gains (IHAM, other equity) keep total NAV near par. A real default cycle tests both the marks and the "no mark-to-market" premise of private credit.
4. External-management fee load — the largest single number in the analysis. ~$194M/quarter = $776M/yr = $1.08/share = 2.54% of gross assets = 5.6% of NAV = 25% of gross investment income. In 1H 2026 the manager was paid $307M while shareholders earned $263M. No total-return hurdle, no lookback (fees were payable in a loss quarter), the fee base includes non-cash PIK and is non-refundable (~$24M/qtr charged on uncollected income), and the base fee scales with gross assets — shares +129% since YE2016 against NAV/share +17.6%. Now subject to a Section 36(b) derivative action (S.D.N.Y., filed 2026-05-26). Structurally inferior to an internally managed BDC such as MAIN, which runs opex ex-interest at 1.3–1.4% of assets.
5. Adverse selection and structural risk drift. Exits yield 8.3–9.1% against a 10.3% retained book — good paper refinances away while troubled paper stays. First lien is 59% vs BXSL's 97.6%; equity-like is ~26% of the portfolio but ~56% of NAV; PIK is 15.8% of income vs a ~8% sector average. Underlying credit stats remain sound (2.2x interest coverage, low-40s LTV), so this is developing rather than realised — track first-lien % and the PIK ratio quarterly.
6. Forward spread decay. New commitments go on at 9.4% against a 10.3% book — ~90bp embedded decay per turnover, roughly a ~3%/yr earnings drag over 4–5 years. This is the mechanism that moved coverage from 1.23x to 0.98x and it continues regardless of this quarter's favourable spreads.
7. Sector contagion, now corroborated by the rating agencies. First-ever net BDC outflow, capital formation −40% YoY, SEC scrutiny of private-market valuations — and in 2026 FSK and PSEC lost investment grade outright while BXSL and GBDC took negative outlooks in June. ARCC is the highest-rated name in the group and a holdout on an uncut dividend in a field that is cutting; that is the bull case and the warning in one sentence.

Position context — this matters more in Beta than in the main account

A portfolio-specific passage was removed from the public build.

Main account: a 2-share position is a rounding error. Nothing to do. If the user ever wants real BDC income exposure in this account, the coverage work above argues for MAIN (internally managed, no fee drag, 0.90 payout ratio) over adding to ARCC at 0.99x NAV — but that is a separate decision and MAIN's own 1.67x NAV premium is a hurdle.

A portfolio-specific passage was removed from the public build.

Zones

	Level	Basis
Fair value	$17.50 – $19.50	Weighted, NAV-anchored
Add / entry	$16.50 – $17.75	0.85–0.92x NAV — a genuine discount that pays for the coverage gap. Overlaps the $17.40 52-week low.
Trim	above 1.05x NAV (~$20.30)	Expressed as a multiple of book, which for a BDC is the meaningful multiple — it rises automatically as NAV changes, unlike a frozen dollar level. ≈10.8x core EPS run-rate.
Break trigger	Base dividend cut · non-accruals >3.5% at cost · NAV <$18.50 · core EPS <$0.45

What would change this verdict

• Upgrade to ACCUMULATE if core EPS returns to ≥$0.50 (coverage >1.04x) on the improving spread/fee environment and non-accruals stabilise at or below 2.4% — or if the price reaches the $16.50–$17.75 entry zone with credit no worse than today.
• Downgrade to TRIM if non-accruals exceed 3.5% at cost, or NAV falls below $18.50, or core EPS prints below $0.45, or the board cuts the base dividend.

Next checkpoint: the Q3 2026 print, expected late October 2026. The specific things to read: core EPS against $0.47 (and whether TTM falls below $1.92); the non-accrual rate against 2.4%; the Grade-2 bucket against $1,089M; and whether the equity marks in IHAM/other equity are still carrying the NAV.

---

9. Data gaps and things I could not verify

Stated explicitly rather than filled in:

1. ✅ RESOLVED — the 4.8% vs 4.9% cost-of-debt conflict. Two independent analyses cited the same 10-Q for different figures. Going back to the document: both are correct and both are quoted verbatim from it. The filing states the WA stated rate "as of June 30, 2026 were 4.8% and 4.1 years" (spot, balance-sheet date, Note 5 and the MD&A repeat) and separately tabulates "Weighted average stated interest rate on outstanding debt" for the three and six months ended 6/30/26 and 6/30/25 at 4.9% in all four columns (period average, MD&A results-of-operations). This report uses 4.9% period-average as the headline for peer comparison, notes 4.8% spot, and adds ~5.3% all-in ($214M interest and credit facility fees ÷ $16,111M average debt). No error in either source — a spot-vs-period-average difference.
2. ✅ CLOSED — ARCC's 63.7% unsecured share is now verified, computed directly from the Q2 10-Q debt table ($10,150M unsecured notes ÷ $15,924M total principal), not relayed.
3. Peer figures are Q1 2026; ARCC's are Q2 2026. ARCC is the only one of the eight to have reported Q2, so the comparison carries a one-quarter timing advantage. The peer cost-of-debt table also mixes stated and all-in bases (PSEC and BXSL show both; others show one).
4. ARCC's own agency ratings are shown as Baa2/BBB/BBB but were not verified by me from a primary source (agency press release or the 10-K exhibit). They come from the relayed peer sweep. Also unverified there: TSLX's S&P and Fitch (an earlier analysis conflated Sixth Street Lending Partners with TSLX itself — the relay's figures are used here); PSEC's Moody's and Fitch; FSK's S&P; GBDC's S&P-vs-Fitch letter mapping; and whether S&P or Fitch acted alongside Moody's on BXSL/GBDC in June. BXSL's 4.83% originated from a secondary source citing the 8-K but was independently confirmed from the filing by the relay.
5. Spillover / undistributed taxable income is NOT disclosed in the Q2 2026 10-Q. The $988M / $1.38-per-share figure used in §3.4 comes from the earnings call and presentation, not from a filing line item, and I did not independently verify it. This matters because it is load-bearing for dividend durability. Note it is not the same as GAAP "accumulated undistributed earnings" on the balance sheet, which is a different and smaller number ($531M at 6/30/26, down from $958M at 12/31/25) — taxable spillover excludes unrealised marks, so the two are not interchangeable. The corroborating evidence that taxable income still exceeds distributions is the $14M H1 excise-tax accrual, which is a filing figure.
6. Non-accrual count discrepancy unresolved — management said four new names in Q2; PitchBook counted five borrowers. Probably a names-vs-loans artifact; not reconcilable from public sources.
7. 10b5-1 status of the insider purchases not verified. Requires Form 4 footnotes on EDGAR, which were not pulled. The buys are confirmed as open-market cash purchases; whether they were pre-scheduled is unknown.
8. The 20% income-based-fee offset applied to ARCC's rate-sensitivity table is my adjustment, not the company's. ARCC's Item 3 disclosure explicitly excludes the income-based fee; I applied a 20% offset because the fee is 20% of pre-incentive NII above the hurdle. The true offset varies with the hurdle calculation and could differ modestly.
9. The spread decomposition (171bp base rate / ~49bp spread of the 220bp yield decline since YE2023) and the ~3%/yr forward earnings drag are relayed analysis, not figures I extracted. The inputs I did verify from the filing are the 9.4% funded yield vs the 10.3% book yield, and the 9.5% vs 10.1% period portfolio yield YoY.
10. FY2026 forward figures are run-rate extrapolations (4 × Q2 core EPS), not guidance. ARCC gave no forward earnings guidance on the Q2 call.
11. A web search surfaced "NAV per share $19.90" for ARCC — that figure is stale and wrong for Q2 2026. The primary source is $19.35, corroborated by Yahoo's bookValue of $19.347. Noted so it does not re-enter a later analysis.

---

Sources

Primary (EDGAR, via curl with declared User-Agent): - ARCC Form 10-Q, period 2026-06-30, filed 2026-07-29 — accession 0001628280-26-050307 - ARCC Form 8-K Ex. 99.1 (Q2 2026 earnings release), filed 2026-07-29 — accession 0001628280-26-050303 - ARCC Form 8-K Ex. 99.1 (Q4/FY2025), filed 2026-02-04 · (Q3 2025), filed 2025-10-28 · (Q4/FY2024), filed 2025-02-05 - SEC XBRL companyfacts API, CIK 0001287750 — NAV/share series 2012–2026

Secondary: Q2 2026 earnings call transcript and presentation (roic.ai) · Yahoo Finance MCP (get_stock_info, get_recommendations, get_holder_info:insider_transactions, price history) · .mcp/fin.py (ARCC + peer screen OBDC/MAIN/TSLX/BXSL/FSK) · PitchBook (Q2 non-accrual detail) · CNBC (July 2026 FOMC) · Moody's BDC sector outlook (Apr 2026)

Relayed (peer benchmarking, §4 and §4.0): peer cost-of-debt, unsecured-share and agency-ratings tables were supplied by a parallel primary-source sweep of peer Q1 2026 10-Q/8-K filings, plus a dedicated moat analysis (fee-drag decomposition, spread attribution, AFFE/index history, marginal-issue pricing). These are not figures I extracted from filings myself — see §9 items 3, 4 and 9 for exactly which parts remain unverified. ARCC's own figures throughout this report are primary-source.

Knowledge base: Knowledge/Themes/bdc-income.md (sweep, live to 2026-10-15) · Knowledge/Playbook/pitfall-dyt-inverts-when-price-caused-the-yield.md · Knowledge/Playbook/pitfall-yahoo-insider-purchases-counts-rsu-grants.md · Knowledge/Playbook/pitfall-vendor-revenue-is-net-construct-for-banks.md · Knowledge/Playbook/principle-primary-source-beats-vendor.md

The bdc-income sweep note's central claim held up on this name. It says: "in a credit-stress environment yield is an anti-signal — the highest yields in the group belonged to the names whose coverage was deteriorating, and a BDC screen sorted by yield is sorted backwards." ARCC's 10.0% yield is entirely price-driven against a flat dividend and 0.98x recurring coverage. The note also recorded ARCC as filling "the external slot" alongside MAIN as the quality anchor — nothing here displaces that, but the coverage inflection is new information since the 2026-07-14 refresh.