Figure — Risk Report

Moderate-elevated risk · 5.0 / 10 TradFi dependency · Nasdaq: FIGR · Upstream of PRIME and wYLDS · verified 2026-09-13

Scope: counterparty analysis. This report assesses Figure as the company upstream of PRIME and wYLDS. It is not equity investment research and makes no recommendation about FIGR shares.

Summary

If you hold PRIME or wYLDS, Figure is the company underneath both. This report covers Figure itself rather than any token, so that the token reports can point here instead of re-explaining it.

Figure does two quite different things, and keeping them apart is the whole point of reading this. One part, Figure Certificate Company, is an SEC-registered issuer whose annual accounts are audited by KPMG; it issues the certificate that ultimately backs wYLDS, and it holds Treasuries, money-market funds and overnight Treasury-backed repo against it. That part is genuinely solid. The other part, Figure Lending, is the largest non-bank originator of home-equity credit in the United States — over $25bn originated since 2018 — and it is the part with the risk. It writes loans, then sells them, securitizes them, or pledges them into the lending facility that PRIME’s backing is lent into.

The direction of travel on that loan book is the reason this report is not scored higher. Delinquency on the loans Figure holds for sale rose three periods running — from about 3.9% to about 5.5% to roughly 6.6% — and then fell to 5.05% in Q2 2026. ⚠️ That fall is not the reversal it looks like. Across the same quarter nonaccrual UPB rose from $2.4M to $14.1M — roughly 17% to 72% of the 90-plus bucket, and ⚠️ a six-month move, 31 December 2025 to 30 June 2026, rather than a quarterly one — and the securitized 60-plus line broke flat at 0.8% → 1.1%. The rate and its composition moved in opposite directions, and composition is the half that describes what is actually in the book. The book grew throughout, so the earlier rises were not a shrinking-denominator effect — and the later fall is partly the same arithmetic running the other way.

There is a second figure that looks far better, and understanding why they differ matters more than either number. Loans that Figure has already securitized show delinquency under 1%. But the lending facility’s own rules require any loan more than 60 days late to be removed from the collateral pool — so late loans are pulled out of the clean pools and land back on Figure’s own balance sheet. The flattering number and the unflattering number are measuring different populations, and the rules actively move loans from one to the other. The unflattering one is closer to what actually secures the facility.

Set against that, Figure is a real and unusually well-documented business: Nasdaq-listed, an audited subsidiary, an active and rapidly growing securitization programme, and institutional buyers for its loans. Its own filings also disclose a history of losses and no guarantee of profitability, and its group structure is more tangled than its regulatory standing suggests.

At a glance

What it is	The largest US non-bank home-equity lender, listed on Nasdaq, plus a separate SEC-registered subsidiary that issues the certificate backing wYLDS.
Why it matters to you	It originates and services the loans behind PRIME’s yield, and its subsidiary issues the certificate behind wYLDS.
Loan quality trend	⚠️ Mixed, and the headline is the flattering half. Delinquency on loans held for sale fell 6.61% → 5.05% in Q2 2026 — but nonaccrual UPB rose $2.4M → $14.1M over six months (31 Dec 2025 → 30 Jun 2026) and the securitized 60-plus line broke flat at 0.8% → 1.1%. The rate improved while the composition worsened, and the two are argued separately under credit quality.
The regulated part	Figure Certificate Company files with the SEC, is KPMG-audited annually, and holds Treasuries, money-market funds and Treasury-backed repo. Genuinely strong, and separate from the lending risk.
Lender’s cushion	The facility advances up to 95% against pledged loan balances, so the cushion protecting lenders is around 5% — not the ~40% homeowner equity figure the loan-to-value stats might suggest.
Governance	The weakest area. A multi-entity group, a regulated subsidiary whose twelve-month liquidity rests on a written parent commitment, and a related company running a public protocol on licensed Figure software with no disclosure obligations of its own.
Biggest risk	That the loan book keeps deteriorating while the parts of it that matter most to token holders remain the hardest to see.

Score breakdown

The four axes beneath the overall, so the composite can be argued with rather than taken on trust. Each points at the section that carries its evidence.

Axis	Score	What it rests on
Balance sheet	5.5	Loans that cannot be sold or pledged accumulate on the balance sheet, which is why the securitized-pool statistics are the flattering proxy and the held-for-sale book is the honest one. See the loan book.
Credit quality	5.0	⚠️ Re-based to evidence rather than proxy, and one good quarter does not un-base it. Headline delinquency improved 6.61% → 5.05%, but nonaccrual UPB went $2.4M → $14.1M across six months — roughly 17% to 72% of the 90-plus bucket — and the securitized 60-plus line broke flat at 0.8% → 1.1%. The direction of the headline and the direction of the composition disagree.
Funding model	5.5	Four exits — hold, whole-loan sale, securitization, Democratized Prime — genuinely diversify funding, but at least one must stay open for originations to keep moving. ⚠️ Held through the Kiavi close rather than cut: $600M of term leverage replaced originate-to-distribute, which pulls the axis down, but retiring a margin-callable repo line and a credit agreement — with all liens released — removes a stress-correlated run channel and leaves the acquired book unencumbered. A swap of acute risk for chronic cost, not a deterioration. ⚠️ Magnitude undeclared: the filing does not state either retired facility’s size. See how Figure funds itself.
Governance	4.5	The weakest axis, and deliberately so. Dependencies overlap across a listed parent, a lender, a regulated certificate issuer, advisers and transfer agents, and interchangeable use of “Parent” can assign an obligation to the wrong entity. Hastra is a related party rather than a subsidiary, with no public-company disclosure obligation.
Overall	5.0	Sits at the mean of the four and below the two stronger ones, because governance and credit quality are the axes that would transmit a problem to anything downstream — and this report exists as a dependency for wYLDS and PRIME, not as an equity view.

The two halves of Figure

The Figure name covers several legal entities. Figure Technology Solutions, Inc. is the Nasdaq-listed parent. Figure Lending, LLC originates and services loans. Figure Certificate Company, or FCC, is a separate SEC reporting registrant that issues YLDS, the certificate underneath wYLDS.

That distinction determines which risks flow into which token. A wYLDS holder depends mainly on FCC: Hastra holds YLDS and wraps it for retail use. A PRIME holder depends on FCC and Figure Lending because the certificates behind staked PRIME are lent into Democratized Prime, a senior facility secured by loans Figure Lending owns. The regulated Treasury-reserve node is relevant to both, but it does not turn the lending exposure above it into a Treasury claim.

Signum Ltd., doing business as Hastra, is a disclosed Figure related party, but it is not a subsidiary and Figure does not consolidate it. Hastra issues wYLDS and PRIME and runs their redemption process; its closeness to Figure does not give token holders direct reporting rights against FCC.

FCC’s filing uses “Parent” for more than one group entity, so a parent commitment cannot always be assigned to one named company. The practical rule is to keep the certificate issuer, lender and wrapper operator separate.

The loan book

The most useful credit series is Figure’s loans held for sale: loans still on its balance sheet awaiting a whole-loan transfer, securitization or another funding exit. On a consistent basis — loans at least 30 days late plus loans in forbearance, divided by total unpaid principal — delinquency rose from 3.91% at the end of 2024 to 5.46% at 31 December 2025 and 6.61% at 31 March 2026 — and then fell to 5.05% at 30 June 2026, per Figure’s Q2 10-Q filed 2026-08-14 (unaudited interim, like the quarters before it).

⚠️ The headline improved, the improvement is real, and this report is holding its score anyway. All three of those need saying together.

It is real, and we checked the obvious reason it might not be. A book growing fast can flatter a delinquency rate without anything improving, and this book grew 17.8% over the quarter — exactly the shape that manufactures a fake recovery. So the numerator was tested separately: delinquent dollars fell 9.9% while unpaid principal rose 17.8%. Both legs moved favourably, which a denominator artifact cannot produce. The 30–59 day bucket fell 53%, reversing the 365% spike recorded below. (Reconstructing 31 December 2025 from the new filing’s comparative column returns 5.457% against the 5.46% published here, so these figures sit on the basis used here rather than merely near it.)

⚠️ But the trend did not reverse. It matured, by the mechanism set out in the paragraph below, which says the 30–59 spike “matters because it feeds the 60–89 and 90-plus day buckets over the next few quarters.” That is what the quarter shows:

⚠️ Each row states its own prior date, because a 10-Q’s comparative column is not one basis. Income-statement style measures compare to the prior quarter; balance-sheet items compare to the prior year-end. Reading the filing’s comparative column as a single “31 Mar” would mis-date three of these six rows.

	prior read	30 Jun 2026
Headline delinquency	6.61% (31 Mar 2026)	5.05%	improved
30–59 day bucket	$10.724M (31 Mar 2026)	—	−53%
90-plus day bucket	$19.216M (31 Mar 2026)	$19.543M	+1.7% — essentially flat
60–89 day loans	2,285 (31 Mar 2026)	4,267	⚠️ +87%
Nonaccrual UPB	$2.4M (31 Dec 2025)	$14.1M	⚠️ share of 90-plus on nonaccrual, about 17% → about 72%
Securitized w-avg 60+	0.8%	1.1%	⚠️ broke a flat line

The 31 Mar 90-plus figure is derived from the reading below — 3.90% of the $492.725M book — rather than lifted from the comparative column.

⚠️ The wave is in the middle bucket, and it has not landed yet. 90-plus was essentially flat across the quarter (+1.7%) while 60–89 nearly doubled, 2,285 → 4,267 loans. So the 30–59 collapse is not loans curing — the early bucket emptied because its contents aged into the next one, which is now carrying them. The 90-plus bucket is flat because the wave has not reached it, not because it has passed. On this reading the headline improvement and the forward risk are the same event seen at two points in the pipeline. And the nonaccrual move is the sharper read: the 90-plus dollars are not merely larger, a much larger share of them have been reclassified as loans Figure no longer expects to accrue interest on.

⚠️ The securitized line is the one that matters most, because it is that selection-bias mechanism turning. The section below explains that Democratized Prime forces loans more than 60 days delinquent out of the collateral pool, which is why the securitized statistic stayed clean at 0.8% at both December and March. It is now 1.1%. The clean pool is less clean, and that is the number the selection-bias argument said should stay flat.

So the score holds at 5.0, and the reason it holds is the same reason it was 5.0 before. This report says the direction of travel on the loan book is why Figure is not scored higher. At the 90-plus, nonaccrual and securitized ends that direction is still adverse — the improvement is at the front of the pipeline, the deterioration is at the back, and the back is where losses are realised.

What would change it, and the test is restated because the previous one was too coarse to see this quarter. ⚠️ The old form asked for the headline ratio and the dollars to fall together — and that test passed on a quarter in which the 60–89 bucket nearly doubled. An aggregate test cannot see migration between buckets, which is the only thing this book is doing.

⚠️ The bucket detail the aggregate hid: the entire headline improvement is 30–59 falling 53.4%. 60–89 is worse on both views — dollars +86.7%, ratio 0.464% → 0.735% — and it is the leading indicator. ✅ 90-plus dollars rose only 1.7% and improved as a ratio, 3.90% → 3.37%, which on a book growing 17.8% is new originations performing rather than old loans curing.

✅ Worth noting the Q1 pass predicted this in writing — that the 30–59 bucket “rolls into 60–89 and 90+ over the following two quarters” — and 60–89 nearly doubled, at roughly 40% conversion, so real curing happened too. A forecast that half-lands is still the strongest evidence here, because it was made before the data existed.

The falsifiable form, per bucket rather than in aggregate: cut to 4.5 at Q3 if 60–89 rises again on both dollars and ratio, or if 90-plus rises on the ratio. ⚠️ Improving headline delinquency alone will not sustain the hold a second time. The Q3 10-Q is due around 2026-11-14.

⚠️ And a reading habit worth carrying off this page. A favourable headline can move for reasons that do not mean what they look like — here a delinquency rate fell partly because its denominator grew and its worst dollars aged into a slower bucket. wYLDS shows the same shape from the other side, where a coverage ratio improved because its denominator shrank, and thUSD shows it again where matched dollars flattered a sub-100% ratio. When a ratio moves in your favour, check which side of it moved.

The quarter-by-quarter detail below is retained on the March basis.

The deterioration is not an artifact of a smaller book. Between year-end and 31 March, unpaid principal grew from $396.475M to $492.725M, or 24.3%, while delinquent dollars grew 54.2%. The 30–59 day bucket rose from $2.308M to $10.724M — 365% in one quarter. That earliest bucket matters because it feeds the 60–89 and 90-plus day buckets over the next few quarters. The 90-plus bucket had already risen 34% and represented 3.90% of the book on its own.

Those figures cover all loans held for sale, not the exact collateral pledged to PRIME’s facility. They are a proxy. But Figure highlights its reliance on HELOCs, and the proxy moved adversely for three periods before improving at the headline in Q2 — with the back buckets continuing to fill, as set out above. With no public facility-level tape, it is the closest recurring filing measure.

Borrower statistics look reassuring: credit scores around 741 and combined loan-to-value around 59%. But they do not show Figure’s recovery position. The facility collateral is 46.31% first lien, 51.68% second and 2.01% third (read 2026-09-08 — first-lien share has risen from about 40%). ⚠️ Name the third lien rather than leaving it as a remainder: subordinate liens are 53.69% combined, and that figure is easy to meet elsewhere labelled as “second lien” when it is second plus third. with a small third-lien remainder. Second liens are paid after the first mortgage, so loss severity can exceed what the property-equity figure suggests.

The facility’s contractual advance rate is up to 95% of pledged balances, leaving about a 5% lender-level haircut. Property equity absorbs losses first, then this haircut protects the senior facility. A severe event would generally be needed to impair principal, but “59% loan-to-value” is not a 41% cushion belonging directly to PRIME holders.

Why the two delinquency numbers disagree

Figure’s securitized pools reported weighted-average delinquency of 0.8% at both 31 December 2025 and 31 March 2026 — and ⚠️ 1.1% at 30 June 2026, the first break in that flat line. That looks inconsistent with a held-for-sale book moving from 5.46% to 6.61%. It is not inconsistent; the two groups are selected differently.

⚠️⚠️ And the mechanism is what makes the 1.1% serious rather than merely higher. Democratized Prime requires loans more than 60 days delinquent to leave its collateral pool — so this pool is continuously cleaned of exactly the loans the metric counts, and the metric rose anyway. Loans are entering the 60-plus bucket faster than removal clears them. ⚠️ The two proxies also moved in opposite directions this quarter — held-for-sale improved, securitized worsened — and the one that worsened is the one with the favourable selection bias. The troubled loan does not vanish: it moves out of the clean pool and back toward Figure’s balance sheet, where Figure must substitute collateral, repay funding or absorb the problem.

This creates selection bias. The sub-1% statistic describes loans remaining inside securitized pools; the higher held-for-sale statistic includes weaker loans returned to Figure. The clean number shows that trust protections work. It does not show that Figure’s whole originated book is equally clean.

For PRIME holders, Figure’s ability to replace or finance removed collateral is part of the protection. The held-for-sale trend is not a map of PRIME’s pool, but it is closer to the balance-sheet risk supporting that promise than the securitized statistic.

What changed after Q1, and it is structural

⚠️ Most of this report is built on Q1 2026 figures (31 March). Three things have happened since, and two of them change the shape of the company rather than its numbers.

Q2 2026, reported 2026-08-13 and filed on Form 10-Q 2026-08-14:

	Q2 2026	change
Net revenue	$226M	+113% YoY
Net income	$87M	+192%, margin 38.8%
Diluted EPS	$0.35	+338%
Adjusted EBITDA	$119M	+126%, margin 54.6%
Consumer Loan Marketplace volume	$4.3B	+132%
Figure Connect volume	$2.8B	65% of total, from 56%
Loan servicing assets	$155M	on $17.3B of underlying loans

Q3 guidance is $4.8–5.2B of marketplace volume. ✅ On growth and margin this is a strong quarter, and the shift toward Figure Connect is a shift toward capital-light fee revenue — volume that earns a fee rather than volume that has to be funded.

⚠️ Figure has taken on $600M of senior debt at 8.500%. The notes were priced 2026-07-09 and closed 2026-07-14 — $600,000,000 aggregate principal, 8.500% Senior Notes due 2031, sold at par to qualified institutional buyers, net proceeds about $587.5M. This is a funding layer the rest of this report does not describe, and at 8.5% it is expensive money against a loan book whose delinquency trend is the reason this page is not scored higher.

⚠️ And the Kiavi acquisition has closed — it is not pending. Announced 2026-06-10 at a $717M total purchase price and closed 2026-09-01 per the Form 8-K. Figure paid approximately $590M in cash, net of cash acquired, funded primarily from the notes above, and at closing repaid Kiavi’s existing credit facilities and terminated a Master Repurchase Agreement with Deutsche Bank.

What Figure bought is a different lending business from its own. Kiavi lends to residential real estate investors — short-term Residential Transition Loans and longer-term DSCR rental-property loans. ⚠️ That is investor-property credit, not owner-occupied home equity, and it behaves differently in a downturn: an investor with a vacant rental and a borrower with a home they live in do not default at the same time or for the same reasons. The loan book this report analyses is now not the whole loan book.

⚠️ Two flow figures, and they go to two different places — do not merge them. Figure’s and Kiavi’s own announcements say the acquisition adds over $7B in new annual first-lien volume to the Figure Connect marketplace, and more than $100M monthly to Democratized Prime. ✅ The second one is the one that reaches a token holder: Democratized Prime is the facility PRIME’s backing is lent into, so this changes that pipeline’s composition — see PRIME and wYLDS.

⚠️ A joint venture between Figure and Sixth Street is reported to be buying loans off Kiavi’s balance sheet. ✅ That appears in the companies’ announcements and not in the closing 8-K, so it is reported rather than filed — recorded here at that standing.

⚠️ None of the delinquency analysis below has been re-derived against Q2. The three-period trend, the two disagreeing delinquency numbers and the securitization figures are all Q1-based. A quarter in which volume grew 132% is exactly the quarter in which a delinquency ratio’s denominator moves, so those figures should be read as dated rather than current.

How Figure funds itself

Figure can hold loans, transfer them to institutional buyers, securitize them, or pledge them into Democratized Prime. Four exits diversify funding, but at least one must remain open for new originations to keep moving. ⚠️ Since 2026-07-14 there is also a fifth source that is not an exit at all: $600M of 8.500% senior notes due 2031. That is balance-sheet leverage rather than a way of moving loans off it, and it services at a fixed 8.5% whether or not any of the four exits stays open. The figures in this section are Q1 and predate it.

Collateral in unconsolidated securitization vehicles rose from $4.596B at 31 December 2025 to $6.766B at 31 March 2026 and $7.731B at 30 June 2026. ⚠️ Maximum exposure to loss grew faster than the collateral behind it — $347.8M → $444.8M, up 27.9%, against collateral up 14.2%. The trusts are isolated from Figure’s creditors. A third-party special servicer handles delinquent and defaulted loans and cannot be removed without cause—a meaningful independent protection.

Democratized Prime is another outlet. Its home-equity pool had lent roughly $351.8M against $521.2M supplied — utilization about two-thirds — read 2026-03-31, with idle funds earning the YLDS-linked rate. ⚠️ That is a Q1 reading, and the date is load-bearing because two other reports depend on this figure — undated, a March number reads as current on PRIME and wYLDS. ⚠️ The facility is now 88.1% related-party funded, leaving a third-party leg of about $41.4M. Growth can signal a functioning market, or greater reliance on the facility if other buyers pull back.

⚠️ At the Kiavi close, two of Kiavi’s funding lines were retired rather than inherited. Per the 2026-09-01 Form 8-K (Item 2.01), Figure repaid in full all outstanding obligations under the Third Amended and Restated Credit Agreement dated 2024-12-19, with U.S. Bank Trust Company, National Association as paying agent; and Kiavi Funding, Inc. and Kiavi Funding Trust 2 terminated their Master Repurchase Agreement with Deutsche Bank AG, New York Branch, dated 2025-09-19.

⚠️ Read the entity, because this is the exact substitution this report marks down elsewhere: the credit agreement is “among Kiavi, U.S. Bank Trust Company … and the lenders party thereto” — these were Kiavi’s facilities, not Figure’s, and at least one secondary account has already described the first as Figure’s own. ✅ What it means for funding is the same either way: the warehouse and repo leverage that had carried Kiavi’s originations is gone, and the book behind it now sits on Figure’s side of the house, funded by unsecured notes at 8.500% rather than by secured lines against the loans themselves.

✅ And the same filing says what happened to the collateral: “Upon such repayment and termination, all related financing documents were terminated and all liens securing the obligations thereunder were released.” The acquired book is unencumbered — it can be pledged or securitized again rather than arriving already committed.

⚠️ The honest characterisation is a swap rather than a deterioration. A master repurchase agreement is margin-callable and collateral-dependent — it is the kind of funding that gets re-margined or pulled precisely when asset values fall. Retiring it removes a run-risk channel that is correlated with stress. The notes run to 2031 with no rollover, against a repo line that rolls constantly. What Figure has taken on instead is chronic and predictable: roughly $51M a year of coupon, payable in good conditions and bad.

⚠️ What is NOT established, stated rather than glossed: the 8-K does not disclose the SIZE of either retired facility. So the direction of the swap is known and its magnitude is not — settling that needs Kiavi’s pre-merger warehouse and repo capacity, which the filing does not carry. The offset is real; whether it is large enough to matter is unmeasured.

Two smaller details from the same filing. The notes closed 2026-07-14 and the merger 2026-09-01, so Figure carried the 8.500% coupon for about seven weeks before the money was deployed. And the consideration was funded “primarily” from the note proceeds — primarily, not solely, so some other source contributed and the filing does not say what.

Two contingent obligations matter. Figure may have to repurchase loans defaulting within 30 to 90 days of origination, an early underwriting signal. It also commits to replenish securitization reserves; the filings call that obligation “not limited”, although management does not expect material funding.

The group also discloses a history of losses and warns that it may not maintain profitability. This is not a near-term insolvency claim: it means volume alone does not prove durable earnings once funding costs, repurchases and reserve calls are counted.

The regulated subsidiary

FCC is the strongest node in the stack. It is a full SEC reporting registrant, and KPMG has audited its annual statements since 2024. The FY2025 opinion was clean. At 30 June 2026, FCC reported $561.910M of qualified assets against a $557.494M certificate reserve, and at 31 March it was $601.524M against $599.706M — about 100.3% coverage then.

⚠️ A dormant facility was drawn, which is a sharper fact than a counterparty changing. The second repo line was already documented at 31 March and recorded as unused; by 30 June it stood at $200,000K, while the UMB leg fell from $389.556M to $82,361K and the repo book as a whole shrank 27.5%. Someone switched a standing line on inside the quarter. FCC’s Q2 10-Q, Schedule I, lists two repurchase-agreement counterparties at 30 June 2026:

counterparty	amount	share
Hidden Road Partners CIV US LLC	$200,000K	70.8%
UMB Bank N.A.	$82,361K	29.2%

UMB still holds the investments on deposit — $555,722K under Schedule V, alongside $5,488K of cash at Lead Bank and $700K at Flagstar Bank, N.A. ⚠️ And the Lead Bank line is a custody fact rather than a cash balance. Schedule V footnote (2) reads: “The Lead Bank account is held by Figure Pay and includes amounts related to affiliated entities. The amount disclosed represents only the portion attributable to FCC and does not reflect the total balance maintained by Figure Pay.” So about 1% of a regulated issuer’s qualified assets sits in a commingled account held by an affiliate, with FCC’s share established by allocation rather than segregation — small in dollars, but the Investment Company Act §28 deposit requirement is an asset test, and an allocated share of someone else’s account is a different kind of asset from a segregated one. So custody and repo used to be the same name and are now two, and only custody is still UMB. ⚠️ $200M of a regulated retail certificate issuer’s qualified assets now sits in overnight repo with a counterparty that is not the bank holding its custody.

✅ Hidden Road Partners CIV US LLC is identifiable, and the answer runs the other way from what an unnamed counterparty implies. From its own audited SEC filing (CIK 0002029588, X-17A-5 filed 2026-02-27, audited by RSM US LLP): a Delaware LLC registered with the CFTC as a futures commission merchant, an NFA member and a CME clearing member. ⚠️ Its ultimate parent is Ripple Labs Inc. — the filing states it in terms: “On October 21, 2025, the Company’s ultimate parent changed as a result of the acquisition of Hidden Road Inc. by Ripple Labs Inc.” (Press dates that close to 10-24; the audited filing says 10-21.)

✅ And the balance sheet explains why the leverage headline is not the finding it looks like. At 2025-12-31: assets $39.019B against member’s equity $520.4M — about 75:1 — but reverse repo $38.183B against repo $37.770B. That is a matched book: offsetting financing legs, which is what a repo dealer is, not a directional position. ✅ FCC’s $200M is 0.51% of that book, and the collateral is overnight Treasuries.

⚠️ What does survive, and it is about correlation rather than credit. UMB is a bank — it has discount-window access and FDIC insurance. A broker-dealer has neither, and SIPC membership does not cover repo exposure. ⚠️ And the parent is a digital-asset company, so a blockchain-native certificate issuer now finances about 36% of its reserve through a crypto-owned dealer: the issuer and its largest reserve counterparty are exposed to the same sector shock, which is precisely the diversification UMB was supplying. ✅ Not a related party of Figure — Ripple is unconnected to FTS.

⚠️ Two limits on the above, stated rather than left implicit. Those financials are 2025-12-31 against a 2026-06-30 exposure, and the firm registered as a security-based swap entity on 2026-09-10 (SBSE-BD and SBSE-C), which is a change in what it does. ⚠️ The same filing also discloses $100M of repo outstanding with Ripple itself at 2025-12-31 — small against a $37.77B book, but its own repo counterparty list includes its parent. The assets were $211.3M of Treasuries and money-market funds plus $389.556M of overnight Treasury-collateralized repo, with no Level 3 holdings.

That quarterly information is unaudited. FCC’s Q1 filing explicitly says its interim financial information has not been audited; the KPMG opinion applies to the annual statements. The distinction matters because “SEC-filed” and “audited” are not interchangeable.

Three less obvious facts qualify the strong reserve. First, the certificates are explicitly unsecured obligations of FCC, backed solely by FCC’s assets. The Investment Company Act deposit rule is a regulatory asset test, not a security interest or bankruptcy-remote trust for certificate holders.

Second, at 31 March UMB Bank N.A. was both counterparty to the entire $389.556M repo book—64.8% of qualified assets—and securities custodian. A second repo arrangement was unused. Treasury collateral is strong, but one institution spans execution and custody.

Third, FCC’s interim “Liquidity and Going Concern” note concludes it has twelve months of funding because its parent committed to pay invoices and periodically forgive intercompany balances. This is management’s conditional interim assessment, not an auditor’s going-concern qualification; the annual opinion is clean.

FCC gives wYLDS high-quality assets and recurring disclosure. The wrapper remains one legal layer removed: Hastra controls account mapping and redemption, without a demonstrated bankruptcy-remote retail reserve. For PRIME, FCC is only the first step before certificates enter Figure Lending’s facility.

Governance and related parties

Governance is weakest because dependencies overlap across a listed parent, lender, regulated certificate issuer, advisers and transfer agents. Interchangeable use of “Parent” can assign an obligation to the wrong entity.

Hastra is a related party, not a subsidiary, with no public-company disclosure obligation. It licenses nontransferable Figure software and pays a 0.50% royalty on transaction revenue. The initial term runs to December 2028 and renews annually. Figure records Hastra’s YLDS as “Debt, current to related parties,” but the specific filing remains unresolved.

At 30 June 2026, $477.536M of $556.033M in certificates — 85.9% — was held by Figure’s parent and controlled entities, up from 79% at 31 March. ⚠️ The concentration rose because outside holders left, not because affiliates grew — and the single-quarter move is the finding. Three dates, because a change between two of them hides it:

	related party	third party	total	related share
31 Dec 2025	$252.282M	$75.911M	$328.193M	76.9%
31 Mar 2026	$472.990M	$125.057M	$598.047M	79.1%
30 Jun 2026	$477.536M	$78.497M	$556.033M	85.9%

⚠️⚠️ Across Q2 the third-party leg fell 37.2% while the related-party leg grew 1.0%, and the whole certificate book shrank 7.0%. Outside holders redeemed more than a third of their position in a single quarter, and that is why the affiliate share rose. ⚠️ The Dec-to-Jun window the 10-Q prints as its comparative shows third-party holdings up 3.4% — arithmetically true and the opposite of the quarter it contains. The filings do not put Hastra in that set. This is entity concentration, not director or executive ownership.

Figure issues the certificate, originates and services loans, runs the facility, supports FCC and licenses Hastra’s software. That reduces unknown-party risk but correlates failures. The 5.0 score reflects an unusually analyzable company whose integration makes legal boundaries and independent checks essential.

Bottom line

Figure is a real lender with real institutional plumbing, and by the standards of anything else backing a crypto token it is remarkably well documented. Most of what this report says comes from its own SEC filings rather than from marketing material, which is not something that can be said of most issuers in this space.

The concern is direction rather than solvency. The loan book is deteriorating, it has done so for three periods running, and the earliest-stage arrears are growing fastest — which usually means the trend continues before it turns. At the same time, the clean statistics that are easiest to find describe the pools that had their problem loans removed, so the reassuring number and the worrying number are not measuring the same thing.

For a token holder the practical reading is this. The certificate side, which backs wYLDS, is the strong part and is largely insulated from the lending risk. The lending side, which is where PRIME’s yield and much of its backing sit, is exposed to how these loans perform — and the protection at that layer is thinner than the headline loan-to-value figures imply. Neither is a reason to avoid the stack, and neither is a reason to treat it as a cash equivalent.

This report uses Figure Technology Solutions and Figure Certificate Company filings available through EDGAR, Figure’s published facility terms, and public protocol and account records. It is counterparty analysis for downstream token holders, not equity investment research. Corrections welcome to info@tidresearch.com.

Revision history: 2026-09-13 — whole-body verification against the Q2 10-Qs and the 09-01 8-K; all four axes and the overall held. ⚠️ FCC’s related-party concentration is 85.9%, not 79% — and it rose because outside holders redeemed, not because affiliates grew. Across Q2 the third-party leg fell 37.2% ($125.057M → $78.497M) while the related-party leg grew 1.0% and the whole certificate book shrank 7.0%. ⚠️ The 10-Q’s own Dec-to-Jun comparative shows third-party holdings UP 3.4%, which is true of that window and the opposite of the quarter inside it. ⚠️ Securitization collateral is $7.731B at 30 June 2026, and maximum exposure to loss grew faster than the collateral behind it — $347.8M → $444.8M (+27.9%) against collateral +14.2%. ⚠️ The securitized 60-plus line is 1.1%, and the mechanism is what makes that serious: the Democratized Prime pool is continuously cleaned of loans more than 60 days delinquent — the exact loans the metric counts — and the metric rose anyway. The two delinquency proxies moved in opposite directions this quarter and the one that worsened is the one with the favourable selection bias. ⚠️ The credit-quality trigger is restated per bucket, because the aggregate form was too coarse to see this quarter: the headline improvement is entirely 30–59 falling 53.4%, while 60–89 — the leading indicator — is worse on both dollars (+86.7%) and ratio (0.464% → 0.735%). ✅ 90-plus dollars rose only 1.7% and improved as a ratio, 3.90% → 3.37%, which on a book growing 17.8% reads as new originations performing. Nonaccrual $2.4M → $14.1M is restated as a six-month move rather than a quarterly one. Democratized Prime’s $351.8M lent against $521.2M supplied is dated to 2026-03-31 and the facility is 88.1% related-party funded, leaving a third-party leg near $41.4M.

*Revision history: 2026-09-09 — the four axes beneath the overall are published for the first time; no score changed. ⚠️ This report had carried an Overall of 5.0 with no breakdown at all, so the composite could not be argued with — the reader saw a number and no way to test it. Balance sheet 5.5 · Credit quality 5.0 · Funding model 5.5 · Governance 4.5 are the marks the analysis has been carrying, now shown against the sections that argue them, matching the convention the companion MSTR report has always followed. ✅ The Q2-2026 content was already current and was checked rather than assumed: headline delinquency 6.61% → 5.05%, the 90-plus bucket essentially flat at +1.7%, the 60–89 bucket at 2,285 → 4,267, and nonaccrual UPB $2.4M → $14.1M all present and correctly based. Overall holds at 5.0.

*Revision history: 2026-08-23 — refreshed to Figure Technology Solutions’ Q2-2026 10-Q (filed 2026-08-14; unaudited interim); Overall held at 5.0. The delinquency series this report’s thesis rests on gained its fourth data point. Headline held-for-sale delinquency fell 6.61% → 5.05%, and the improvement is genuine rather than a denominator effect: the book grew 17.8% — the exact shape that manufactures a false recovery — so the numerator was tested separately and delinquent dollars fell 9.9%. Both legs moved favourably, and the 30–59 day bucket fell 53%, reversing the 365% spike recorded for Q1. Method control: reconstructing 31 December 2025 from the new filing’s comparative column returns 5.457% against the 5.46% published here, so the new figure sits on the basis used here.

⚠️ The score is held anyway, and the reasoning is on the page rather than implied — a held score against a favourable headline reads as refusing to update unless the argument is shown. The trend did not reverse, it matured, exactly as the 30–59 spike implied — it “feeds the 60–89 and 90-plus day buckets over the next few quarters”: 90-plus dollars were essentially flat, $19.216M → $19.543M (+1.7%), while the 60–89 bucket nearly doubled, 2,285 → 4,267 loans — a wave that has aged out of the earliest bucket and is now sitting one step short of 90-plus — and nonaccrual UPB went $2.4M → $14.1M, taking the share of 90-plus loans on nonaccrual from about 17% to about 72%. Most importantly, the securitized weighted-average 60+ figure broke its flat line at 0.8% (December and March) and printed 1.1%. That is the selection-bias mechanism turning: Democratized Prime forces 60+ loans out of the collateral pool, which is why that number had stayed clean — and the clean pool is now less clean. The improvement is at the front of the pipeline and the deterioration is at the back, where losses are realised.

The hold is made falsifiable rather than left as judgement: what would move it is a second consecutive quarter of falling headline delinquency with 90-plus dollars flat or falling and securitized 60+ back at or below 0.8%. Q3 10-Q due around 2026-11-14.

Also added, as a reading habit rather than a per-report caveat: a favourable headline can move for reasons that do not mean what they look like. Here a rate fell partly because its denominator grew and its worst dollars aged into a slower bucket; wYLDS shows the inverse, a coverage ratio improving because its denominator shrank; thUSD shows matched dollars flattering a sub-100% ratio. When a ratio moves in your favour, check which side of it moved. last_verified is not bumped — only the Figure filing layer was refreshed, and the facility terms and on-chain material in this body still date from the earlier pass.*