The payment queue is the product
A large bank can owe another bank a substantial dollar payment while expecting an almost equally large payment in the opposite direction. Paying both separately requires cash to be in the right place at the right moment. Offsetting the obligations can reduce that funding burden. CHIPS, the Clearing House Interbank Payments System, builds this idea into a high-value U.S. dollar payment network: it continuously looks for payments that can be released and settled together, while also permitting individual release. The Clearing House operates the system. [1, 3]
The resulting service is neither merely a messaging channel nor a promise to settle everything later. A submitted instruction may wait in a queue; once it is released under the system's rules, it settles finally. This difference between waiting for settlement and reversing a completed settlement is the central design feature. The account below focuses on that mechanism, rather than on the wider question of how a bank funds all its intraday obligations.
Why the distinction matters at scale
The operator's annual statistics show $503.382 trillion of payment value across 158.105 million transactions in 2025, or about $2.014 trillion and 632,419 transactions per business day. The table reports dollar amounts in thousands, a unit that must be converted before calling them trillions. Its 2026 row covers only the year through August: approximately $2.245 trillion per day over 168 business days. That is an interim activity measure, not a full-year result or operator revenue. [5]
CHIPS is a bank-level service. The September 28, 2026 participant directory includes U.S. institutions and international banks, and distinguishes participant identifiers rather than merely counting global banking groups. Businesses and banks outside the direct network generally reach it through participant relationships. A customer's payment can therefore have operational steps before entry and after release. Finality between CHIPS participants does not establish that a corporate beneficiary has already reconciled its receivable, or that every bank in an international route has completed its own posting. [1, 6]
Prefunding makes the position real
Participants supply opening funds to the CHIPS prefunded balance account at the Federal Reserve Bank of New York. CHIPS records positions attributable to participants and adjusts those positions as payments settle. Supplemental transfers can add funding during the day. These are settlement resources, rather than an unfunded promise that every bank will supply whatever money is needed at closing. The June 4, 2026 rules govern opening requirements, primary and supplemental positions, release, and closing procedures. [3]
Economically, prefunding establishes an inventory of settlement capacity. A bank that receives a released payment gains capacity to send other payments. It does not need to ship fresh dollars into the network for every outgoing instruction. Conversely, a bank with persistent net outflows cannot manufacture the missing cash merely by submitting more instructions. Reuse and offsetting help with the timing and shape of flows; they do not eliminate the need to finance a genuine net obligation.
A participant's CHIPS position should not be confused with a customer's deposit balance. They sit at different points in the payment chain and are governed by different relationships. Customer balances, internal limits, sanctions screening and instruction validation remain meaningful even when the settlement network has sufficient resources.
Individual release, bilateral offsets and multilateral cycles
There are three useful ways to think about release. An individual payment uses the sender's available position and increases the receiver's position. Bilateral netting groups payments going in opposite directions between two participants. Multilateral netting considers a group spanning at least three participants. In each case, the system must satisfy its position constraints; it does not choose a combination solely because its gross value is large. The rules also preserve the aggregate positions when a batch is released. [3]
The operator's October 2024 public disclosure describes CHIPS as a hybrid system: instructions wait in storage while a -saving algorithm seeks appropriate release opportunities. It distinguishes the system from a deferred net settlement arrangement that waits for a single settlement event. The mechanism allows finality to occur throughout the day. The disclosure is useful for architecture, while the newer rulebook is the reference for the rules in force examined here. [4]
A useful conceptual distinction is between netting and recycling. Netting settles a compatible set using its residual obligations. Recycling allows funds already received to support later payments. Both reduce the need for fresh funding, but they are not the same operation. A bank with balanced total flows may still require money if its outgoing payments arrive much earlier than the incoming ones.
A hypothetical three-bank queue
Suppose Bank A instructs a $100 million payment to B, B instructs $90 million to C, and C instructs $80 million to A. Assume these are the only instructions, that all participants have already met their opening requirements, and that the relevant system constraints permit the batch. This is a simplified illustration, not a reconstruction of the actual algorithm or a claim about minimum required CHIPS prefunding.
The gross payment value is $270 million. A sends $100 million and receives $80 million, so its net position falls by $20 million. B receives $100 million and sends $90 million, so it gains $10 million. C receives $90 million and sends $80 million, so it also gains $10 million. The three changes sum to zero. If A begins with $20 million of usable capacity and B and C with zero incremental capacity beyond the assumed opening conditions, the illustrated batch can consume $20 million at A and leave $10 million at each of B and C.
Compare that with a rigid sequence requiring A's entire $100 million payment to go first. A would need $100 million at that instant. But it would be misleading to call $270 million the unavoidable gross-system funding requirement: the same money can be reused as B and C pay. The example demonstrates a sequencing benefit, not a universal percentage saving.
Scroll horizontally to see all columns.
| Hypothetical bank | Outgoing | Incoming | Net position change |
|---|---|---|---|
| A | $100 million | $80 million | −$20 million |
| B | $90 million | $100 million | +$10 million |
| C | $80 million | $90 million | +$10 million |
| Total | $270 million | $270 million | $0 |
Priority has an opportunity cost
A payment may be urgent because its beneficiary has a deadline. Another payment may tolerate a short wait for an offset. CHIPS permits priority designations and mechanisms for reserving eligible supplemental resources. Priority affects which payments the release methodology tries to select; it is not a guarantee that an unfunded instruction will settle immediately. The public product description explicitly qualifies speed by payment value, funding conditions and participant controls. [1, 3]
There is an economic tradeoff. More money committed to the network can reduce waiting. More willingness to wait can create opportunities to offset. An individual bank's preference is not always identical to the network's most cash-efficient combination. A sender may care more about one important payment than about maximizing the value of all payments released per dollar of funding.
This also explains why an average network statistic is a poor service promise for a particular customer. A route dominated by net outgoing payments, a very large instruction or a deadline outside a downstream bank's posting cycle can behave differently from a typical transaction. A high percentage of fast releases can coexist with a small group of consequential delayed payments.
Finality belongs to released payments
The rulebook ties final settlement to the recorded position changes accompanying release. The October 2024 disclosure explains the legal and risk-management framework supporting that result, including its relationship to the Federal Reserve's supervision of a designated financial market utility. Systemic importance brings scrutiny; it is not a government guarantee of every customer's claim. [3, 4]
Before release, an instruction is still pending in the network's queue. After final settlement, the original participant-to-participant obligation has been discharged. An operational request to recover a mistaken payment therefore should not be described as simply erasing the completed settlement. Recovering value and determining responsibility are distinct questions involving the applicable rules and parties.
End-of-day procedures matter for the residual queue, but they do not reopen every payment that settled earlier. There are three distinct states: a queued instruction, a released and settled instruction, and an instruction still unresolved at closing. Treating all three as 'sent' hides exactly the information needed to understand timing risk. The same distinction helps explain why a queue can support efficiency without making completed payments provisional.
What liquidity savings are worth
The Clearing House's strategic-role paper reports a 2024 -efficiency ratio of roughly 29 dollars settled per dollar of funding and estimates $5.14 billion of annualized economic savings. Its current product page reports approximately 26:1 for 2025. These are operator measures and estimates with different periods. They are not independently measured increases in participants' profits, and the ratio is not a return on invested capital. [1, 2]
A transparent hypothetical illustrates why the valuation depends on assumptions. If a bank can avoid carrying $50 million of extra average funding and assigns that funding an annual marginal cost of 4%, the simple annual cost is $2 million before service fees, technology costs, liquidity-buffer requirements or behavioral changes. If the genuinely avoidable funding is only $10 million, the same calculation is $400,000. The value depends on the counterfactual, not just the impressive volume passing through the system.
Nor does freeing settlement liquidity automatically create the same amount of regulatory capital. Cash, capital and balance-sheet capacity are related but different constraints. A bank may conserve cash without changing its binding capital requirement, or retain the released capacity as a safety buffer rather than put it into additional lending.
The limit is the flow, not only the software
The mechanism works best when usable offsetting instructions are available at compatible times. If many participants delay submission or one large receiver absorbs funds without sending them onward, opportunities change. A disruption can also reduce the usefulness of historical funding patterns. The operator's disclosure discusses -efficiency and operational risks precisely because reliable software alone cannot make every day's pattern identical. [4]
CHIPS also provides an alternative high-value dollar processing path, but redundancy is not the same as complete independence. Funding arrangements, participants' own systems and downstream customer processing remain dependencies. Network availability and usable beneficiary funds are different outcomes.
The durable insight is narrower and more powerful than 'netting makes payments cheap.' By combining funded positions with continuous selection of compatible obligations, CHIPS can economize on liquidity while giving released payments finality. How much it saves, how quickly a specific payment exits the queue and when the recipient can use the proceeds depend on the flow of instructions and the surrounding banks. That is the settlement bargain the network actually offers.
Sources
- The Clearing House, CHIPS product overview; checked October 4, 2026SourceBack to text: ↑1↑2↑3↑4
- The Clearing House, Strategic Role of CHIPS; 2024 efficiency estimatesSourceBack to text: ↑
- CHIPS Rules and Administrative Procedures, June 4, 2026Source · PDFBack to text: ↑1↑2↑3↑4↑5
- CHIPS public disclosure, October 2024; operator-hosted January 2025 fileSource · PDFBack to text: ↑1↑2↑3↑4
- CHIPS annual statistics; 2026 row through AugustSourceBack to text: ↑
- CHIPS participant directory; September 28, 2026SourceBack to text: ↑