Archegos made prime-broker concentration and margin escalation

Summary

Archegos obtained economic exposure to a concentrated group of public equities mainly through total-return swaps with several prime brokers. Each dealer could see the positions on its own books but did not automatically see the client's combined exposure across dealers. The structure did not eliminate counterparty risk; it transformed visible share ownership into a network of bilateral credit exposures whose aggregate scale depended on information the client supplied and the banks verified.
Public records now support several distinct propositions. A jury convicted Hwang and Halligan after trial. Hwang was later sentenced to 18 years in prison and Halligan to eight years. By contrast, SEC and CFTC complaints contain allegations unless a court, admitted settlement or other adjudicative record supplies a different status; the CFTC's case against Archegos and Halligan was later dismissed on jurisdictional grounds. Regulatory consent documents and bank-commissioned reviews have their own stated boundaries.
The SEC alleged that Archegos's capital rose from roughly $1.5 billion to more than $36 billion while gross exposure rose from about $10 billion to more than $160 billion. Those complaint figures describe the agency's case, not an independent civil merits judgment. The criminal verdict established charged offenses, but it does not turn every sentence in every parallel complaint into a jury finding.
Credit Suisse's board-commissioned review described a relationship in which static swap margin, concentrated positions, stale or missing information, recurring exposure-limit breaches and repeatedly deferred dynamic collateral changes coexisted for months. It also documented warning signals that reached credit and counterparty committees without producing timely de-risking. The report is a detailed bank record, not a judicial judgment.
Margin was the decisive accountability mechanism because disclosure alone would not have absorbed liquidation loss. A dealer that lacks complete cross-bank visibility can still require enough independent collateral, impose concentration and liquidity add-ons, shorten cure periods, limit new trades, and reserve close-out rights. Those protections lose force when commercial exceptions become persistent, limits are raised after breaches, or revised terms remain proposals.
On March 24 and 25, 2021, price declines generated calls that Archegos could not satisfy. Dealers then learned more of the aggregate book and faced a coordination problem: a managed unwind could reduce market impact only if rivals waited, while an early unilateral sale could move loss to slower banks. That incentive conflict made weak pre-default controls expensive at precisely the moment they were least repairable.
Loss measures are not perfectly comparable. Credit Suisse reported about $5.5 billion; Nomura reported losses and estimates spanning two financial years; Morgan Stanley disclosed a $644 million credit loss plus $267 million of trading losses from a single client event; UBS reported a $434 million effect on net profit attributable to shareholders; and MUFG Securities initially disclosed an approximately $270 million potential loss. Different bases, dates, taxes and recoveries prevent a clean league table.
The post-collapse record shifted the control standard. The Federal Reserve told supervised firms with large derivatives and investment-fund relationships to obtain and verify enough information to understand a counterparty's size, leverage, largest positions and other prime-broker relationships; set risk-sensitive margin; monitor limits; and prepare for rapid close-out. The January 9, 2026 revision removed references to reputational risk, but did not erase the counterparty-credit expectations.
The lasting accountability question is not whether a bank had a policy document. It is whether a named control owner could stop exposure growth when data were unverified, whether exceptions expired automatically, whether concentration and liquidation costs changed collateral before distress, and whether senior leaders received decision-ready information rather than indicators whose breaches could persist without consequence.

The event was a control test, not only a client default

Archegos was a family office founded and controlled by Hwang. It dealt with multiple global banks through prime-brokerage and equity-derivatives relationships. A total-return swap allowed Archegos to receive the economic gain or loss on a referenced security without buying the shares in its own name. The dealer commonly hedged by purchasing the reference shares. Archegos posted collateral and paid financing; the dealer carried market hedges and the credit risk that the client might fail after an adverse move.

The arrangement was bilateral, but the risk was systemic in the practical sense that several dealers financed overlapping positions and would sell the same or correlated shares after default.

That structure divided knowledge. A bank could observe its contracts, collateral, stress measures and hedge inventory. It could ask Archegos about net asset value, leverage, concentration, liquidity and positions held elsewhere. It could compare answers over time and insist on documentary evidence. But no dealer possessed a complete, continuously reconciled map of the client's cross-prime-broker book merely because it was a swap counterparty. The control challenge was therefore to price uncertainty, not to pretend uncertainty did not exist.

The Federal Reserve later framed the problem across firms rather than as a single-bank anomaly. Its counterparty-credit-risk supervisory letter, issued in 2021 and revised on January 9, 2026, identifies weak due diligence, limited public evidence information, poor margin practices, deficient risk measurement and escalation failures as recurring lessons from Archegos. The letter applies to Federal Reserve-supervised firms with large derivatives portfolios and relationships with investment funds; it is guidance about supervisory expectations, not a retroactive ruling that any particular person violated law.

The 2026 revision removed references to reputational risk. The central expectations concerning leverage, concentration, margin, limits and close-out remained.

That distinction matters. A client can be accountable for deception and manipulation while a bank remains accountable for the amount of unsecured uncertainty it accepted. Fraud is not a defense to weak controls when the purpose of those controls is to withstand incomplete or false information. Conversely, a bank control failure does not excuse a client's proven criminal conduct. Institutional legitimacy requires both propositions to remain visible at the same time.

Total-return swaps converted ownership opacity into bilateral credit risk

The swaps gave Archegos synthetic long exposure: if the referenced share price rose, the dealer owed the gain; if it fell, Archegos owed the decline. The dealer's hedge purchase could transmit Archegos's demand into the cash market even though the family office did not appear as registered owner of the hedge shares. Multiple dealers could hold hedges linked to economically similar Archegos positions, each believing its own exposure was collateralized according to bilateral terms.

This structure had three consequences for control design. First, gross exposure mattered alongside net asset value. A counterparty with $10 billion of capital and $100 billion of gross positions could suffer a rapid equity impairment from a much smaller percentage move than an unlevered investor. Second, name concentration mattered alongside portfolio-level direction. A short index position might reduce ordinary market beta but would not necessarily offset a collapse in a few large, relatively illiquid long positions. Third, liquidation horizon mattered alongside daily volatility.

A dealer could calculate a one-day move and still underestimate the discount required to sell billions of dollars of the same shares after other dealers began selling.

Credit Suisse's Special Committee report on the Archegos losses, published by the bank as an exhibit to a Form 6-K, supplies the most detailed public chronology of one dealer relationship. The committee retained outside counsel, reviewed documents and interviewed employees. Its conclusions are consequential because the bank's board published them and management accepted remedial action. They remain conclusions of a bank-commissioned investigation, however, not findings returned by a court.

The report describes Archegos's Credit Suisse portfolio changing from a more balanced book into a heavily long and concentrated one. As that happened, the economic meaning of existing margin changed. Collateral terms justified when long positions were offset by short swaps and prime-brokerage holdings became less protective when the short side diminished and the longs appreciated. A static percentage can fall in risk terms even when the percentage itself does not move: gains expand notional exposure, crowded positions require longer disposal, and stress loss grows faster than collateral.

The accountability implication is concrete. The owner of a swap relationship must not treat contract execution as the end of underwriting. Counterparty credit is a continuously renewed decision. Every increase in notional, every concentration increase, every weakening offset and every refusal to provide information should alter either capacity, collateral or permission to trade. A system that measures those changes but cannot force a decision creates records without control.

The scale and concentration claims require source labels

The SEC's April 2022 enforcement announcement said the agency sued Archegos, Hwang, Halligan, head trader William Tomita and chief risk officer Scott Becker. It described a parallel civil case involving an alleged fraudulent scheme and manipulative trading. The announcement is a regulator's description of its complaint. It is not itself a judgment, and its allegations against each defendant must not be recast as established civil liability.

In its filed civil complaint, the SEC alleged that Archegos's capital expanded from approximately $1.5 billion in March 2020 to more than $36 billion in March 2021, while gross exposure increased from about $10 billion to more than $160 billion. It alleged that swap exposure tied to some issuers became extraordinarily concentrated: at points, more than 50 percent of ViacomCBS's outstanding shares, more than 45 percent of Tencent Music Entertainment's, and more than 30 percent of a class of Discovery shares.

It also alleged that Archegos trading often represented more than 20 percent of daily volume in selected names and sometimes exceeded 30 or 40 percent. Every figure in this paragraph is presented as an SEC allegation from a pleading.

Those alleged ratios illuminate the prime-broker question even without presuming the SEC's final success. Concentration should be measured against several denominators: the client's capital, the dealer's limit, the issuer's free float, ordinary daily volume, stressed days to liquidate, and aggregate exposure inferred from client disclosures. A position can appear tolerable as a percentage of a bank's capital while still being dangerous relative to the market's capacity to absorb a synchronized sale.

The criminal record later changed the status of important overlapping conduct. The jury's verdict established offenses after trial for Hwang and Halligan; it did not adjudicate the SEC complaint count by count. Scott Becker and William Tomita entered guilty pleas in the criminal matter and also resolved CFTC proceedings with stated admissions. Parallel records can corroborate a narrative while retaining different defendants, elements, burdens and remedies. A responsible reconstruction states which record proves what.

A district-court September 19, 2023 SEC opinion reinforces that separation. The court denied Archegos's and Hwang's motions to dismiss, granted Halligan's motion in part and denied it in part, and stayed the SEC case pending the criminal matter. That ruling tested whether the SEC had pleaded claims capable of proceeding; it was not a final civil liability judgment and did not convert the complaint's factual assertions into findings.

Counterparty representations were a control input, not a control substitute

The alleged and later partly proven misrepresentations concerned facts a dealer would use to decide credit: portfolio composition, concentration, liquidity, leverage and the scale of business with other banks. The initial DOJ charging announcement expressly said the indictment's charges were accusations and that Hwang and Halligan were presumed innocent unless proven guilty. That was the correct status on April 27, 2022. The later trial verdict changed the criminal status of the offenses on which the jury convicted them; it did not erase the need to label the charging document historically.

For a prime broker, representations should feed a verification process with consequences. A dealer can request position files, financing totals, largest-name exposures, liquidity assumptions and the number of other providers. It can compare reported concentration with market activity, hedge flows, financing requests and prior statements. It can condition capacity on timely delivery, add collateral when data are stale, and stop new risk when answers are incomplete or inconsistent. None of those measures requires perfect market-wide transparency.

The CFTC's complaint against Archegos and Halligan alleged commodity-interest fraud involving swaps and counterparty communications. Among other things, it alleged that Archegos sometimes described its largest position as roughly 35 percent of net asset value when the actual share was approximately 70 percent, and that swap counterparties issued aggregate margin calls exceeding $13 billion as the firm collapsed. Those are allegations in the agency's pleading. The complaint should not be used to state that Archegos or Halligan admitted every asserted fact.

The CFTC's accompanying release and orders have a mixed evidentiary character. The case against Archegos and Halligan was filed as a complaint. Separate orders settling with Becker and Tomita contained commission findings and recorded admissions by those respondents, together with cooperation obligations and monetary sanctions. A single press release therefore contains both unproven allegations and settled administrative findings. The boundary follows the document and the respondent, not the convenience of a unified story.

That boundary later became sharper. In a September 19, 2023 opinion and order, the district court granted Archegos's and Halligan's motions to dismiss the CFTC amended complaint, denied the government's stay motion as moot, denied leave to replead on futility grounds, and directed the clerk to close the CFTC case. That dismissal is a procedural and jurisdictional outcome about the CFTC case, not a factual exoneration of the conduct alleged, not a reversal of the Becker and Tomita administrative orders, and not a limit on the later criminal verdict.

Credit Suisse's margin terms drifted away from the risk

The Credit Suisse committee chronology shows how an apparently collateralized relationship can become under-secured through commercial concessions and portfolio change. In May 2019, standard swap margin averaging about 20 percent was reduced to 7.5 percent. The rationale included offsetting short swaps and long positions in prime brokerage. Liquidity add-ons were discussed: five percent for positions exceeding two days of ordinary trading volume, with a further 8.5 percent beyond five days. The report says the add-ons were not implemented or properly recorded in the governing terms.

By September 2020, according to the report, the Credit Suisse portfolio had gross exposure of about $9.45 billion, comprising roughly $7.18 billion long and $2.27 billion short. The book was therefore no longer the substantially offset configuration used to justify the earlier concession. Average swap margin had fallen to about 5.9 percent as appreciated notional outgrew static collateral, while prime-brokerage margin averaged around 15 percent. The key failure was not one low percentage in isolation. It was the absence of a binding mechanism that recalibrated collateral when direction, concentration and liquidity changed.

In July 2020, a stress exposure measure exceeded its $250 million limit by large multiples. The report records approximately $600 million on July 16 and $828 million on July 22. Breaches then continued. Potential exposure also exceeded its limit. Instead of reducing positions or collecting enough collateral to restore compliance, the bank repeatedly allowed temporary excesses, raised limits or pursued a transfer of the relationship to another internal unit.

Limit increases are not inherently improper. A larger limit can be justified by new capital, improved collateral, better information or a demonstrably lower-risk portfolio. But raising a threshold after a breach, without changing the underlying exposure, converts an alarm into a revised tolerance. The control owner should have to show what new fact makes the higher number prudent, who approved it, how long it lasts, and what happens if promised remediation is not delivered.

Credit Suisse considered dynamic margin in September 2020. The proposal did not become an enforceable change with a deadline. A February 2021 calculation would have increased average margin to about 16.74 percent and required approximately $1.27 billion of additional initial margin, still below an estimated amount near $3 billion under a fuller approach described in the report. On March 8, the bank sought either dynamic terms or $250 million by March 15. Neither outcome was completed before the crisis.

Archegos did post $500 million of additional initial margin in February. That collection reduced exposure but did not cure the structural mismatch. Meanwhile, the bank's potential-exposure calculation moved from roughly $517 million on February 12 to $72 million a week later and then to nearly zero at the end of February, before rising again in mid-March. A calculation that can report negligible future credit exposure shortly before a multibillion-dollar loss demands challenge. Automation can increase speed, but speed amplifies whichever assumptions, netting rules, horizons and data feeds it is given.

The committee report also describes $2.4 billion of variation margin returned to Archegos between March 11 and March 19 as positions appreciated. Variation margin settles current mark-to-market value; holding it without contractual basis would not be a legitimate substitute for initial margin. The relevant failure was that the bank had not secured stronger independent collateral terms while it had negotiating power. It had a contractual right to demand discretionary initial margin on short notice, but personnel treated that step as commercially extreme.

A right that decision-makers believe cannot be used is not equivalent to usable protection.

Exception management became the real risk appetite

Formal limits tell only part of a bank's risk appetite. The effective appetite is visible in what happens after a breach. If a breach causes immediate escalation, a time-limited waiver, compensating collateral and a forced decision, the limit constrains behavior. If it produces recurring emails, temporary approvals and deadline extensions while new trades continue, the exception process becomes the actual policy.

Archegos's Credit Suisse relationship displayed several forms of drift. Stress breaches persisted. Potential-exposure limits were raised. The portfolio transfer was treated as a route to improved control even though migration did not itself reduce market risk. Dynamic margin remained under negotiation. Commercial teams continued adding exposure while risk personnel sought better terms. Responsibilities were spread across prime services, credit risk, counterparty oversight, legal documentation and senior committees, creating many points of awareness but no single point that reliably stopped growth.

An effective exception register needs more than an entry and an approver. It needs an original limit, current exposure, quantitative rationale, independent challenge, compensating action, expiry date, cumulative extensions, owner, senior escalation threshold and automatic trading restriction at expiry. Extensions should never erase the history of prior breaches. A committee should see the age and recurrence of an exception, not merely its latest approved state.

Concentration also needs its own hard consequence. Portfolio-wide stress can be diluted by offsets that disappear in a disorderly unwind. A name representing several days of ordinary volume should attract a liquidity charge that rises nonlinearly. A client that refuses to disclose cross-dealer exposure should face a conservative assumption, not a neutral blank. The dealer does not need certainty to act; it needs a rule for uncertainty.

The central accountability question is who possessed stop authority. A relationship manager can request more collateral, but may be measured on revenue. A credit officer can flag a breach, but may lack authority over trading systems. A committee can discuss the issue, but may meet after exposure changes. A credible design assigns one accountable executive the power and duty to block incremental exposure, with independent risk able to impose the block directly and senior management required to approve any override in writing.

March 2021 compressed months of deferred decisions into days

The final week did not introduce the underlying weaknesses. It converted them into cash demands. ViacomCBS announced a large equity offering on March 22. Its shares and other names in the Archegos portfolio declined. The Credit Suisse report states that its gross ViacomCBS holding reached about $5.1 billion that day. Archegos continued buying through dealers as prices weakened, increasing exposure during the period when liquidity and financing capacity were deteriorating.

On March 24, Tencent Music shares fell sharply. By the next day, Credit Suisse calculated a margin call above $2.5 billion. On March 25, it issued calls totaling more than $2.8 billion and learned that Archegos could not pay. The precise calls and times differ among dealer records, but the institutional fact is clear: once the client could no longer transfer enough cash, each bank's protection depended on existing collateral and the price at which it could liquidate hedge positions.

The dealer call on March 25 revealed the rough aggregate scale. According to the Credit Suisse report, Archegos described approximately $120 billion of gross exposure, split around $70 billion long and $50 billion short, against roughly $9 billion to $10 billion of equity. The client asked the prime brokers to avoid immediate unilateral sales. That disclosure arrived when the book was already in distress, demonstrating why cross-provider information collected only at default has limited preventive value.

Credit Suisse declared default and began unwinding on March 26. It sold more than $3 billion that day, including participation in blocks led by another dealer. Yet some competitors moved more quickly and sold larger amounts before prices fell further. On March 27 and 28, Credit Suisse, UBS and Nomura explored a coordinated liquidation, while other dealers were unwilling to delay. Credit Suisse later executed large blocks on April 5 and April 14 and had disposed of about 97 percent of its exposure by April 22, according to its committee report.

One operational event sharpened the governance concern. On March 12, the bank extended the maturity of more than $13 billion of swaps, many with 7.5 percent margin, despite an instruction to hold changes while risk terms were under review. The committee concluded the extension did not materially alter the ultimate loss because the original maturities were after the default. Even so, the event showed that booking and documentation processes could move a large portfolio in a direction contrary to a risk instruction. A control should prevent an unauthorized change before execution, not discover it after reconciliation.

Between March 12 and March 26, the bank also permitted approximately $1.48 billion of additional net long exposure, at average margin described near 21.2 percent for those trades. Higher margin on new trades did not cure the legacy book. This is a recurring control trap: a firm tightens terms prospectively while leaving the dominant stock of old exposure under weaker terms. Risk governance must evaluate the combined portfolio after every change.

By March 19, a severe stress view indicated a potential initial-margin loss near $1 billion. That signal did not immediately produce a full collateral demand or position reduction. The delay cannot be explained only by a lack of data. The bank had enough internal evidence to know that the relationship was large, concentrated, outside limits and dependent on unfinished margin negotiations. Imperfect information should have increased caution; it instead coexisted with continued trading.

Liquidation exposed a predictable coordination conflict

After a common counterparty defaults, dealers may prefer a coordinated sale because pacing can reduce market impact. But coordination is unstable. Each dealer knows that another can sell first, preserve its own collateral cushion and leave later sellers with lower prices. Unless there is a binding arrangement, shared incentive or credible legal structure, waiting can transfer value to rivals.

Archegos presented exactly that conflict. The dealers held hedges in overlapping names. They had different collateral levels, internal loss estimates, legal rights and decision speeds. A bank with enough collateral could tolerate a slower sale; an under-margined bank could not. A bank that believed others would liquidate had reason to accelerate. The proposed standstill therefore asked each dealer to bear market risk for a collective outcome it could not enforce.

This does not mean a fire sale was the only conceivable response. Banks should prepare default playbooks, validate authority, map positions to market liquidity, stage block-trade capacity and identify communication rules before distress. But liquidation planning is a last line of defense. It cannot restore collateral that was never collected or diversify a book after prices gap down.

The March outcome also shows why close-out readiness belongs in pre-trade credit. Expected loss should include the cost of exiting a concentrated hedge while peers sell the same asset. The proper horizon is not the time required to send an order; it is the time required to complete an economically plausible disposal under stress. If that horizon grows, allowable notional should fall or independent collateral should rise.

Bank losses must be compared on their own accounting bases

Credit Suisse suffered the largest publicly reported dealer loss, approximately $5.5 billion. The Federal Reserve's 2023 enforcement announcement used that figure and imposed a $268.5 million penalty on Credit Suisse for unsafe and unsound counterparty-credit-risk practices. The announcement summarizes an agreed enforcement action. The underlying consent document, discussed below, expressly avoided a formal adjudication of disputed fact or law.

Nomura's April 27, 2021 results update reported a ¥245.7 billion loss connected with transactions with a U.S. client for the year ended March 31, plus an estimated approximately $570 million impact in the following financial year. It said more than 97 percent of the positions had been unwound by April 23. The release did not name Archegos in the quoted loss section, so attribution rests on the event context and other public records. The two amounts span reporting periods and should not be added to another bank's single-period number without explanation.

Morgan Stanley's first-quarter 2021 Form 10-Q disclosed a $644 million credit loss and $267 million of related trading losses arising from a single prime-brokerage client event. The filing did not identify the client by name in that disclosure. Those components total $911 million on the firm's stated bases, but the distinction between credit and trading loss should be retained because it explains how disposal timing affected the result.

UBS said in its first-quarter 2021 results that the default of a U.S.-based client of its prime-brokerage business reduced net profit attributable to shareholders by $434 million. That after-tax or attributable-profit measure is not directly comparable with a gross trading loss or a pre-tax credit charge. The official release did not name Archegos in that sentence, so it is best described as the relevant single-client default in context rather than as a direct naming by the bank.

MUFG Securities disclosed in a March 31, 2021 loss update a potential loss of approximately $270 million arising from transactions with a U.S. client and said recovery efforts could reduce the amount collected from the client. The notice did not name Archegos. A potential-loss disclosure is different from a finalized booked loss.

The Credit Suisse committee's comparative account placed Nomura near $2.9 billion, Morgan Stanley near $1 billion and UBS at a $774 million pre-tax effect, while describing losses at Goldman Sachs, Deutsche Bank and Wells Fargo as immaterial or absent in public reporting. That comparison is useful for liquidation outcomes but does not replace each firm's accounting statement. The sharp dispersion supports a control conclusion, not a moral ranking: collateral, speed, legal terms, hedge composition and decision authority changed who absorbed the common shock.

Different outcomes reveal the value of usable escalation

Some prime brokers escaped with little reported loss despite financing the same client. That does not prove every one of their pre-default controls was adequate. An institution can make weak underwriting decisions and still exit early enough to avoid loss; another can identify risk but act too slowly. Loss is an important outcome measure, not a complete evaluation of process.

Still, the dispersion is informative. Faster banks could calculate exposure, obtain authority and execute blocks while liquidity remained. Slower banks carried positions through further declines. Dealers with stronger margin started with a larger buffer. Dealers whose legal and operational teams were ready could exercise close-out rights without prolonged internal debate. Those capabilities are parts of counterparty risk, not merely trading execution.

Credit Suisse's losses were not caused by one person or one missed call. The committee report describes failures across first-line business ownership, credit challenge, risk committees, limit governance, data and senior oversight. Responsibility should therefore be allocated by decision: who approved the original concession, who owned the recurring breach, who could demand collateral, who permitted new exposure, who validated the exposure calculation, who received escalation and who controlled the default sale.

That decision map avoids two accountability errors. The first is diffuse blame, in which an entire culture is faulted but no duty is assigned. The second is single-person blame, in which removal of one executive substitutes for repair. A defensible review links each consequential decision to authority, information available at the time, policy requirement, documented rationale and outcome.

The civil, criminal and administrative records prove different things

The DOJ's verdict notice records the jury result in the criminal prosecution. On July 10, 2024, the jury convicted Hwang on 10 of 11 counts and Halligan on all three counts against him after a nine-week trial. Hwang's convictions included racketeering conspiracy, securities fraud, market manipulation and wire fraud; Halligan's included racketeering conspiracy, securities fraud and wire fraud. Those verdicts establish guilt on the convicted counts under the criminal burden of proof. They do not establish guilt on an acquitted count, liability for every civil allegation, or responsibility of non-defendants.

The SEC civil case is separate. Its complaint alleged manipulative trading, false representations and related securities-law violations by specified defendants. The September 2023 district-court opinion allowed major parts of the SEC case to proceed and stayed it pending the criminal matter. The public district-court January 12, 2026 status order reflects procedural management after the criminal case and directed another update. Neither a motion-to-dismiss ruling nor a status order resolves the merits. As of July 17, 2026, the evidence used here does not support describing the SEC complaint as a final civil judgment against all defendants.

The CFTC record also separates respondents and instruments. Its filed complaint against Archegos and Halligan remained allegations at filing and was later dismissed by the district court. The commission's settled orders concerning Becker and Tomita stated findings and admissions by those respondents. Those admissions cannot automatically be imputed to every defendant, while the criminal guilty pleas and later jury verdicts stand on their own records.

Credit Suisse's internal report is neither a pleading nor an agency adjudication. It is a board-commissioned account that the bank released. Its detailed emails, committee chronology and financial figures are strong evidence of internal process, subject to the report's scope and methodology. It should not be described as a regulator's independent proof of every event.

The Federal Reserve consent action is similarly bounded. The signed 2023 consent order recites deficient counterparty-credit practices, including inadequate margin, unresolved limit breaches, weak data and failures to respond to warning signs, and requires remediation. It also states that the parties entered the order to settle the matter without formal proceedings and before adjudication or a finding on disputed issues of fact or law. Agreement to the order, waiver of hearing and compliance obligations are real; they are not a jury verdict or an unrestricted admission of every recital.

The Prudential Regulation Authority took a different form of settled action against Credit Suisse International and Credit Suisse Securities (Europe) Limited. Its Final Notice found breaches of fundamental governance and risk requirements and imposed an £87.082 million penalty after a 30 percent settlement discount from £124.403 million. The notice attributes about $5.1 billion of the Credit Suisse loss to the UK firms. It expressly limits its findings to those firms and says it makes no finding or criticism concerning third parties. That limitation should travel with any quotation or paraphrase of the notice.

FINMA's July 2023 Archegos proceeding concluded that Credit Suisse had seriously and systematically breached financial-market law in risk management. It ordered remedial measures to continue under UBS following the takeover and opened an enforcement proceeding against a former Credit Suisse manager. Opening an individual proceeding is not a finding of individual liability. FINMA's institutional conclusion and the unresolved status of a separate person must not be collapsed.

Finally, a separate investor action produced a 2025 appellate decision. The Second Circuit's opinion in the Archegos-related securities litigation affirmed dismissal of investor claims under the legal theories presented there, including asserted duties involving prime brokers. That ruling resolves those claims on their pleaded grounds. It does not declare the banks' risk controls sound, reverse regulatory findings, or alter the criminal verdict.

Sentencing and appeal status must be dated

Hwang's December 2024 sentencing record states that the district judge imposed 18 years in prison, three years of supervised release and more than $9 billion in restitution. The DOJ's current fraud-prosecution summary records that Halligan was sentenced on January 27, 2025 to eight years in prison. These are imposed criminal sentences, not requested penalties or statutory maximums.

The appellate status is not final. A public Second Circuit docket for one of the consolidated Hwang appeals shows that on May 19, 2026 the court consolidated docket numbers 25-1941, 25-2088, 26-872 and 26-874. It set June 19 for appellants' supplemental opening brief, August 18 for the government's supplemental opposition, and September 8 for the supplemental reply. Because July 17 falls before the government's stated deadline, the appeals were unresolved on the publication date. The convictions and sentences remained operative; their appellate review was pending.

This dated formulation prevents two opposite errors. Calling the criminal case merely an allegation after the verdict understates adjudicated guilt. Calling every aspect finally exhausted while appeals remain open overstates finality. Accountability reporting should update procedural verbs as a matter moves from charge, to verdict, to judgment, to appeal.

Supervisors converted the collapse into control expectations

The Federal Reserve letter asks supervised institutions to understand a fund counterparty's strategy, risk profile, leverage, largest positions, concentration and relationships with other market entities. If a counterparty will not provide enough information, the firm should reconsider the relationship or use more conservative limits and terms. That expectation is important because it rejects the idea that opacity is a fixed condition to be accepted without price.

Due diligence must be recurring. Information that was adequate when a relationship was small can become inadequate after rapid growth. Verification should be proportionate to exposure and should combine client submissions with independent market and transaction data. The objective is not to reconstruct every trade at every competing bank. It is to reach a prudent decision under uncertainty.

Margin should be risk-sensitive and capable of changing before default. A robust approach incorporates volatility, direction, name concentration, wrong-way risk, liquidity, gap risk and the expected time to close. It distinguishes variation margin, which settles current gains and losses, from independent collateral that protects against future movement during cure and liquidation. It includes authority to call additional collateral and a process that can actually use that authority.

Limits need escalation linked to action. A breach should trigger a trade block or a documented temporary exception, not merely visibility. Senior committees need gross and net exposure, stress loss, collateral, top concentrations, days to liquidate, stale-data flags, exception age and cross-provider uncertainty in one decision view. The relevant question is whether the committee can choose among collect, reduce, hedge, stop or exit while there is still time.

Default planning should connect legal rights with executable operations. Firms need current contact trees, tested authority, reconciled positions, collateral location, block-sale channels, market-abuse safeguards and predefined governance for communications with other dealers. They should rehearse a scenario in which peers refuse a coordinated sale, because that is the economically rational stress case.

The 2026 revision to the Federal Reserve letter is easy to misstate. It did not issue a new Archegos enforcement judgment or certify that industry repairs were complete. It revised supervisory guidance by removing reputational-risk references. Counterparty-credit expectations in the letter continued to address due diligence, information, margin, exposure measurement, limit governance and close-out preparedness.

Remediation must be proved through changed decisions

A remediation plan is not evidence that risk has been repaired. Proof requires observable changes in live decisions. For Archegos-type exposure, the first test is whether the firm can aggregate legal-entity, product and hedge data into a counterparty view fast enough to act. The aggregate should reconcile prime brokerage, swaps, collateral, financing, issuer concentration and stress data without relying on manual adjustments that cannot be reproduced.

The second test is whether missing information has an encoded consequence. A stale net-asset-value report, unexplained variance, incomplete provider list or refused position detail should raise a flag with a deadline. At the deadline, capacity should fall, margin should rise or trading should stop according to approved rules. Human review remains necessary, but review cannot mean that every consequence is optional.

The third test is whether limits bind. Historical data should show how often limits were breached, how long breaches lasted, how many times exceptions were extended, who approved them and whether exposure grew during the exception. A healthy control environment should be able to produce examples where profitable business was constrained because verification or collateral was limited public evidence.

The fourth test is whether margin tracks liquidation risk. Back-testing should compare assumed disposal horizons and discounts with stressed market evidence, including episodes when multiple dealers sold the same names. Concentration add-ons should increase before a position dominates ordinary volume. Independent collateral should not fall merely because recent price appreciation creates a benign short history.

The fifth test is whether calculations are challenged. A sudden movement from hundreds of millions of potential exposure to nearly zero should generate a traceable review of inputs, netting, horizons and exclusions. Control staff need access to underlying data, not only a dashboard result. Changes to calculation logic should have independent approval and retrospective comparison.

The sixth test is whether close-out can happen at institutional speed. Simulations should measure the time from default signal to legal confirmation, trading authority, position reconciliation and first sale. They should include a failed standstill, conflicting prices, unavailable decision-makers and a rapidly falling market. Results should go to the same executives who approve counterparty appetite.

The Federal Reserve order required Credit Suisse to strengthen board oversight, counterparty-credit governance, data quality, concentration controls, margin practices, scenario analysis and independent review. The Board later announced that the cease-and-desist order had been terminated on May 12, 2026. That notice establishes the end of the formal action; it does not withdraw the historical recitals or penalty, and its short text does not make a detailed public finding that every control across the combined institution or wider industry was repaired. Completed, tested controls remain the relevant operating evidence.

A project milestone can close while the same economic exception survives under a new label.

Enterprise automation should shorten accountability, not obscure it

Counterparty risk spans contracts, trades, collateral, market prices, limits and client information. Software is essential at that scale. It can reconcile positions, calculate gross and net exposure, detect concentration, age exceptions, preserve approvals and block orders. The governance value comes from shortening the distance between a warning and an accountable decision.

Automation can also create false assurance. A green status may reflect a raised limit rather than reduced exposure. A low potential-exposure number may depend on offsets that cannot be liquidated together. A completed task may record that someone reviewed a breach without recording what action followed. A migration may move data while leaving decision rights unchanged.

Every automated indicator should therefore expose its denominator, source time, exclusions and consequence. Users should be able to move from a portfolio total to the legal entities, trades and collateral behind it. Exception logic should preserve the original breach and all extensions. Order controls should read the current risk state before accepting incremental exposure. Senior reports should distinguish measured facts, client assertions and conservative estimates.

The aim is not to remove judgment. It is to make judgment attributable. A named person should decide whether evidence justifies an override, and the system should record what was known, what policy required, what compensating measure applied and when the decision expires. That record supports both faster action during stress and fairer accountability afterward.

An accountability map for prime-broker concentration

The client owns truthful, timely representations and performance of collateral obligations. Senior client officers own statements about net asset value, leverage, concentration and liquidity within their authority. Traders own accurate transaction instructions and compliance with applicable market rules. Those duties are not diluted because banks also have controls.

The relationship business owns initial and continuing underwriting, commercial terms and escalation of adverse information. It should not be able to neutralize independent risk through revenue arguments alone. Credit risk owns limits, collateral sufficiency, data challenge and the authority to restrict exposure. Market risk contributes liquidation and concentration analysis. Operations owns accurate booking and collateral movement. Legal owns enforceable terms and usable default rights.

Senior management owns aggregate appetite and unresolved exceptions. The board owns oversight of material concentration and assurance that independent control functions have stature, resources and direct access. Internal audit owns testing whether rules operate in live cases, including whether exceptions are routinely renewed. Supervisors own clear expectations and credible follow-through, while respecting the legal limits of guidance and consent actions.

No one actor has complete control, but that does not make accountability collective and vague. Each handoff should have an input, deadline, decision right and consequence. If client information is missing, credit decides the conservative treatment. If collateral is not delivered, trading capacity changes. If a limit is breached, an approver with sufficient seniority accepts a dated exception or exposure falls. If default occurs, a designated leader activates the close-out plan.

This map also clarifies institutional legitimacy. Banks are licensed and supported because they are expected to transform risk with disciplined controls, not merely pass client representations through to balance sheets. A credible institution can explain why it financed a concentrated counterparty, what uncertainty it priced, when its tolerance was exceeded, who acted, and how evidence now demonstrates repair.

What would have changed the loss trajectory

No single counterfactual can be proved, because dealer actions affect market prices and one bank's exit changes another's outcome. Still, the record identifies several controls that would have reduced exposure before the final week. Earlier dynamic margin would have transferred more cash while Archegos had liquidity. Enforced stress limits would have capped growth. Concentration add-ons tied to days of volume would have priced disposal cost. A trade block after missed information or expired exceptions would have prevented incremental notional.

The strongest counterfactual is cumulative. Smaller exposure, more independent collateral, fewer concentrated names and tested close-out authority would have improved the bank's position under any plausible liquidation path. The objective of counterparty control is not to forecast the exact trigger. It is to keep an adverse but credible trigger from becoming a franchise-scale loss.

That is also why fraud and bank failure should not be treated as competing explanations. Deception can defeat a control that depends entirely on truthful answers. A resilient control uses verification, conservative assumptions, collateral and limits to reduce dependence on those answers. The criminal verdict addresses the defendants' conduct; the supervisory and bank records address whether institutional defenses were proportionate to the risk.

Conclusion

Archegos turned prime brokerage into an accountability test because the decisive failures occurred before the margin calls. Economic exposure grew across dealers, concentration deepened, collateral terms lagged, representations became more important as verification became less adequate, and warnings accumulated without a binding reduction in risk. When prices fell, the remaining question was who could sell first.

The public record now supports a disciplined allocation of claims. The jury verdict and sentences establish criminal outcomes for Hwang and Halligan, subject to pending appellate review as of July 17, 2026. SEC and unresolved CFTC complaint assertions remain allegations except where separate pleas, admissions, findings or judgments apply. Bank reviews, supervisory guidance, consent terms and final notices each speak with their own authority and limits.

The repair standard is operational. A prime broker should be able to show that it aggregates exposure, verifies what matters, prices what it cannot verify, makes concentration expensive, forces exceptions to expire, blocks growth when limits fail and can close out without waiting for commercial consensus. Accountability becomes credible when those controls change decisions before loss, not when institutions explain afterward that many people saw the warning.