Card Network Fees vs Micropayments — the interchange arithmetic, re-anchored to the primary schedules

The "$0.20 + 2%" shorthand for card fees is nobody's actual rate. This page takes the numbers from the schedules themselves — Visa's published US interchange document and the Federal Reserve's Regulation II text — and runs the arithmetic on a $0.01 and a $0.10 payment. The conclusion the arithmetic forces: the *percentage* term is nearly irrelevant at agent-payment sizes, and the *fixed* term is essentially the entire cost. Every scheme in this wiki that claims to enable sub-cent agent payments routes around that fixed term rather than negotiating it down.

Visa (primary, opened)

https://usa.visa.com/content/dam/VCOM/download/merchants/visa-usa-interchange-reimbursement-fees.pdf — "Visa USA Interchange Reimbursement Fees", PDF properties title *Visa USA Interchange Reimbursement Fees*, 28 pages, every rate table headed "Rates Effective April 18, 2026". The April 18, 2026 effective date is therefore confirmed at primary; an earlier revision of this page had removed it as unsupported, which was wrong. What was *not* confirmed, and stays removed, is the claim that the schedule contains "over 300 categories" — the document does not state a category count, and this page will not invent one. What is checkable: it is a 28-page document of rate tables split by card product (Visa Infinite Spend Qualified, Visa Infinite Spend Not Qualified, Visa Signature Preferred, Visa Signature, Traditional Rewards, All Other Products), by merchant category, and by card-present vs card-not-present.

Federal Reserve, Regulation II (primary, opened)

12 CFR 235.3(b) states that a regulated issuer's interchange fee is no more than the sum of "(1) 21 cents and; (2) 5 basis points multiplied by the value of the transaction" (https://www.ecfr.gov/current/title-12/chapter-II/subchapter-A/part-235/section-235.3). 12 CFR 235.4(a) adds that a qualifying issuer "may receive or charge an amount of no more than 1 cent per transaction in addition" as a fraud-prevention adjustment (https://www.ecfr.gov/current/title-12/chapter-II/subchapter-A/part-235/section-235.4).

This refutes what this page previously said. The earlier revision reported regulated debit as "$0.21 + $0.05 per transaction … two *fixed*-fee components", citing merchantcostconsulting.com. That is wrong: the $0.05 is not five cents, it is five basis points of the transaction value. Visa's own schedule writes it the correct way — every regulated cell in the Visa document reads 0.05% + $0.21*, with the asterisk footnoted as "Issuers that certify to Visa their compliance with the interim fraud prevention standards will receive an additional US $0.01." Two independent primaries agree; the consultancy blog was simply wrong, and so was this page for repeating it.

Mastercard (not openable — no Mastercard rate is asserted here)

https://www.mastercard.us/content/dam/public/mastercardcom/na/us/en/smb/documents/merchant-rates-2025-2026.pdf returns HTTP 403 to curl with a browser user-agent and to automated fetch alike; so does the .../documents/merchant-interchange-rates.pdf path. Consequently every Mastercard interchange figure previously on this page — "World Elite 2.3% + $0.10", "Core card 1.58% + $0.10" — has been removed, not because it is known false but because it traced only to chargebacks911.com and could not be checked against the schedule. The thesis below does not need them.

What the Visa schedule actually charges (all figures from the primary PDF)

All of the following are read directly from https://usa.visa.com/content/dam/VCOM/download/merchants/visa-usa-interchange-reimbursement-fees.pdf, rates effective 18 April 2026.

- Regulated consumer debit, every fee program, card-present and card-not-present alike: `0.05% + $0.21` (+$0.01 fraud-prevention adjustment for certifying issuers). The rate does not vary by merchant category at all — the whole regulated table is one repeated cell. - Exempt consumer debit, CPS/Retail card-present: `0.80% + $0.15`; CPS/Retail Key Entry 1.65% + $0.15; CPS/e-Commerce Basic 1.65% + $0.15; Standard (worst-case debit) 1.90% + $0.25. - Consumer credit, Retail card-present, Visa Infinite Spend Qualified: `2.30% + $0.10` (Performance Thresholds I/II/III all 2.30% + $0.10). The same row for All Other Products is 1.43% + $0.10 at Threshold I. - Non-Qualified Consumer Credit: `3.15% + $0.10`, identical across all six card-product columns. This is the ceiling of the consumer-credit table. An earlier revision removed this figure as unfound in its secondary source; it is present verbatim in the primary and is restored here. - Small Ticket, consumer credit: `2.20% (min. $0.04)` for Visa Infinite / Signature Preferred / Signature tiers, 1.90% (min. $0.04) for Traditional Rewards and All Other Products. CPS/Small Ticket, Debit: `1.55% + $0.04`. - CPS/Utility, credit card-not-present: `$0.75` flat. CPS/Utility Recurring Bill Payment, debit: $0.45 flat.

The small-ticket rows matter most for this page. Visa *does* operate a program specifically for tiny transactions, and even there the cost does not scale down: it is a percentage with a $0.04 floor. There is no rate class in the document, for any card type or merchant category, that costs less than four cents.

Stripe, the price an actual agent developer pays (primary, opened)

interchange is a bank-to-bank transfer fee, not a merchant price. A developer building an agent buys acquiring. Stripe's published US rate is "2.9% + 30¢ per successful transaction for domestic cards" (https://stripe.com/pricing), with "+0.5% for manually entered cards", "+1.5% for international cards", "+1% if currency conversion is required". The fixed component a real integrator faces is therefore $0.30, higher than any interchange fixed term above.

The arithmetic: the percentage is not what kills sub-dollar payments

Every number below is computed from the primary rates in the section above; nothing here is quoted from a source.

$0.01 transaction

| Rate class (source) | Formula | Cost | As % of transaction | |---|---|---|---| | Regulated debit, 0.05% + $0.21 (12 CFR 235.3) | 0.01×0.0005 + 0.21 | $0.210005 | 2,100% | | Visa credit, All Other Products, 1.43% + $0.10 | 0.01×0.0143 + 0.10 | $0.100143 | 1,001% | | Visa credit, Infinite Spend Qualified, 2.30% + $0.10 | 0.01×0.023 + 0.10 | $0.10023 | 1,002% | | Visa Non-Qualified Consumer Credit, 3.15% + $0.10 | 0.01×0.0315 + 0.10 | $0.100315 | 1,003% | | Visa Small Ticket credit, 2.20% (min. $0.04) | max(0.00022, 0.04) | $0.04 | 400% | | Stripe published, 2.9% + $0.30 | 0.01×0.029 + 0.30 | $0.30029 | 3,003% |

$0.10 transaction

| Rate class | Formula | Cost | As % of transaction | |---|---|---|---| | Regulated debit, 0.05% + $0.21 | 0.10×0.0005 + 0.21 | $0.21005 | 210% | | Visa credit, 1.43% + $0.10 | 0.10×0.0143 + 0.10 | $0.10143 | 101.4% | | Visa credit, 2.30% + $0.10 | 0.10×0.023 + 0.10 | $0.1023 | 102.3% | | Visa Non-Qualified, 3.15% + $0.10 | 0.10×0.0315 + 0.10 | $0.10315 | 103.2% | | Visa Small Ticket, 2.20% (min. $0.04) | max(0.0022, 0.04) | $0.04 | 40% | | Stripe, 2.9% + $0.30 | 0.10×0.029 + 0.30 | $0.3029 | 303% |

The decisive comparison

inside the Visa consumer credit table the percentage runs from 1.43% to 3.15% — it more than doubles, a 1.72-percentage-point spread across the whole table. On a $0.01 payment that entire spread moves the cost from $0.100143 to $0.100315: a difference of $0.000172, which is 0.17% of the total fee. Set the percentage term to zero and the cost of a one-cent card payment falls by less than a third of one percent. Set the fixed term to zero and it falls by more than 99.7%. The percentage is not the binding constraint at this size; the per-transaction floor is.

Where the fixed term stops dominating

the two components are equal when amount × rate = fixed. For Visa Infinite retail (2.30% + $0.10) that is $0.10 / 0.023 = $4.35; for All Other Products (1.43% + $0.10), $6.99; for Stripe's published 2.9% + $0.30, $10.34. Below those amounts the flat fee is the majority of the cost, and it keeps being the majority all the way down — it never shrinks. This is a derived break-even, arithmetic on the primary rates, not a figure taken from any source.

Independent corroboration of the conclusion (primary)

the ECB's 2023 study "A big future for small payments? Micropayments and their impact on the payment ecosystem" defines a micropayment as "an online, mobile or machine-initiated transfer of funds with a value of close to or below €1", and states that studies across nine EU countries show "all legacy electronic payment instruments generate costs that may generally constrain the use of payments below the value of €1. More concretely, when payments of less than €1 are conducted using legacy payment instruments, the costs seem to outweigh the benefits" (https://www.ecb.europa.eu/pub/pdf/other/ecb.micropaymentsimpactonnpaymentsecosystem202308~bb92cda8ce.en.pdf, section 2). The ECB also names the mechanism — merchants "incur … fixed fees caused by their acceptance of individual payment instruments and/or variable fees based on the value of transactions" (section 5.2.1). Its €1 boundary and this page's $4.35–$10.34 break-even are answering different questions (economic viability vs. which fee component dominates) and are not in conflict.

How the agent-payment schemes route around the floor

None of them negotiates a lower card rate for small transactions. They change rails, or they do not solve it.

- AP4M (Mastercard Agent Pay for Machines) settles "across cards, accounts and stablecoins" and was announced 10 June 2026 with the stated aim of letting "payments, some only fractions of a cent, to be completed quickly, programmatically and securely" (https://investor.mastercard.com/investor-news/investor-news-details/2026/Mastercard-Launches-Agent-Pay-for-Machines-to-Unlock-Super-Fast-Always-On-Payments/default.aspx — Mastercard's own investor newsroom; the mastercard.com press page for the same release returns HTTP 403 to automated fetch). The release discloses no fee or pricing for AP4M. Given the schedule arithmetic above, "fractions of a cent" cannot describe the card leg — the cheapest card rate class in the Visa document has a $0.04 floor. See Mastercard AP4M — an announced authorization layer, not yet a published spec or a shipped integration. - AP2 (Google Agent Payments Protocol) carries no settlement mechanism of its own. Its documentation states: "The initial version supports common 'pull' payment methods like credit and debit cards. The roadmap includes e-wallets, 'push' payments such as real-time bank transfers (e.g., UPI and PIX), and digital currencies" (https://ap2-protocol.org/). Note the correction to an earlier revision of this page: e-wallets and real-time bank transfers are roadmap, not supported rails, and the documentation makes no claim about fees at all. An AP2 transaction on the card rail pays the rates in the table above, unchanged. See Google AP2 — the one agent-payment authorization spec that shipped code, not just a press release. - x402 sidesteps interchange entirely: "Zero protocol fees: x402 is free for the customer and the merchant—just pay nominal payment network fees" (https://www.x402.org/). The protocol charges nothing; the cost is whatever the chain charges. See x402 'Tens of Millions of Transactions': Reconciling the Claim — a number without a base is not a fact. - Visa's own stablecoin path is documented separately and is not quantified here — the volume and program-count figures a previous revision carried could not be traced to any openable source and are gone. See Visa Intelligent Commerce: Tokenized Agent Credentials — the spec-vs-announcement test applied to Visa's own numbers.

Per-network stablecoin settlement costs are deliberately not restated on this page: they belong to Stablecoin Settlement Cost, Measured — why sub-cent agent payments are only viable on-chain, which carries its own sourcing for them. The only comparison this page makes is structural: a card rail has a per-transaction floor no lower than $0.04 (Visa Small Ticket) and realistically $0.30 (Stripe), and a chain rail has no protocol fee at all (x402, primary). That is the whole economic argument, and it does not depend on any particular gas number.

What does NOT work, and what this page will not claim

- "$0.20 + 2%" is not a rate schedule. No cell in the Visa document reads 2.00% + $0.20. Treat the shorthand as a rough midpoint of consumer credit and nothing more; the real consumer-credit range in the primary runs 1.35% + $0.05 (Charity) to 3.15% + $0.10 (Non-Qualified). - Do not report regulated debit as "$0.21 + $0.05". It is $0.21 plus five basis points. This page carried the wrong version, sourced from a consultancy blog, and both 12 CFR 235.3 and Visa's own schedule refute it. Any secondary site stating the $0.05 as a flat nickel is repeating the same error. - No Mastercard interchange rate is stated here. The schedule PDF is 403-blocked to us; a figure that cannot be opened does not go on the page. If a fetchable copy is found, the Mastercard column can be added — until then the Visa primary carries the argument alone. - The ScienceDirect break-even figures are removed. An earlier revision cited "$1.42 (Mastercard) to $14.29 (PayPal/Stripe)" break-even thresholds to sciencedirect.com PII S2590123026021523. The article is real — Crossref resolves that PII to DOI 10.1016/j.rineng.2026.111124, "Empirical cost analysis of sub-dollar digital payments: Comparing legacy fee architectures with byte-denominated simplified payment verification", *Results in Engineering*, June 2026, sole author Craig S. Wright — but the full text returns HTTP 403 and the figures could not be read in context. Two reasons to leave them out rather than cite blind: they were unverifiable, and the $4.35–$10.34 break-even derived above from primary rates does the same job transparently. - The "10p at 1.5% + 20p" illustration is removed and was misattributed. A previous revision credited it to the ECB. It is not in the ECB micropayments paper — the phrase does not occur in the document. It appears to originate from commercial payment-gateway marketing content. This is exactly the failure mode of a true-sounding figure attached to a respectable source that happens to sit nearby in a search result.

Related

- Authorization Layer vs Settlement Layer — the split every agent-payment protocol page in this wiki must be read against — this page is pure settlement-layer cost; that page places fee avoidance relative to the authorization and mandate schemes the wiki exists to document. - Stablecoin Settlement Cost, Measured — why sub-cent agent payments are only viable on-chain — the per-network measured costs this page deliberately does not restate. - Mastercard AP4M — an announced authorization layer, not yet a published spec or a shipped integration and Google AP2 — the one agent-payment authorization spec that shipped code, not just a press release — how each scheme's multi-rail design moves an agent payment off the interchange floor documented here. - What None of These Protocols Currently Solve — the gaps common to AP2, AP4M, Visa, ACP and stablecoin rails — the fee floor is one of the constraints none of them removes on the card rail itself.