A
Anvesan
Stablecoin Think Tank
March 2026 · Research Brief
State of the Market

The State of Agentic Payments

A March 2026 view of what is live, who is leading, what the numbers actually say, and how the market is splitting between card-led agentic checkout and internet-native machine payments.
Live consumer surface
700M+
weekly ChatGPT users, giving agentic commerce a giant discovery interface.
Machine rail signal
75.41M
x402 transactions over the last 30 days at time of writing
Protocol breadth
60+
organizations Google says are collaborating on AP2.
Market map
Commercial readiness vs long-term strategic weight
March 2026
Long-term strategic weight
Commercial readiness
Longer arc
Scaling now
Experimental
Operational
Early live Machine-native payments
Live now Card-led agentic checkout
The first lane is closer to scale today. The second may matter more over the longer run.
What is live right now Who is leading What the numbers say What comes next
The market in one frame
Live Bounded consumer commerce

Card-led agentic checkout

OpenAI, Stripe, Mastercard, Visa, and Google are building governed purchase flows where AI can discover, compare, and transact under explicit user authorization.

Early live Machine payments

Internet-native agent payments

x402, Stripe MPP, and Circle Nanopayments are building rails for agents paying APIs, software, data, and compute programmatically.

What matters most

Agentic payment as a concept is well established now. The real breakthrough is that the industry is finally building the trust, liability, and orchestration layers needed to make those payments governable.

Bottom line

The long-term winners will prioritize cross-rail presence. They will have to control authorization, trust, merchant integration, and settlement connectivity across multiple payment environments.

A
Anvesan
Stablecoin Think Tank
March 2026 · Executive overview
Executive summary

Agentic payments are real, but maturity is uneven

In March 2026, the category has split into two distinct markets: bounded agentic checkout on card and merchant rails, and machine-native payments on internet and stablecoin rails. The first is commercially further along. The second is architecturally cleaner and may matter more over the long run.
Core reading

What exists today is not unconstrained autonomous spending. It is controlled delegation, trusted authentication, scoped credentials, and early machine-to-machine monetization.

Market shape

Bounded checkout is ahead. Machine-native payments are earlier, but fit APIs, software, data, and compute much better than traditional checkout flows.

Anvesan view

The real moat is in trust and orchestration, the sooner companies realize that the further ahead they'll be.

Market structure
LiveMerchant rails
Card-led agentic checkout

OpenAI Instant Checkout, Stripe’s ACP stack, Mastercard Agent Pay, Visa Intelligent Commerce, and Google’s AP2/UCP work are pushing governed consumer transactions into production.

Commercial signal
Closer to scale
Existing merchant, network, and dispute infrastructure gives this lane a meaningful adoption advantage.
Strategic edge
Distribution
The first wave rewards access to merchants, issuers, and trusted payment credentials.
Early liveInternet rails
Machine-native payments

x402, Stripe MPP, Visa’s MPP card spec, and Circle Nanopayments are shaping software-to-software payment flows for APIs, real-time data, content, and compute.

Technical fit
Architecturally cleaner
Better fit for micropayments, machine timing, and global programmable settlement.
Current constraint
Demand is earlier
The rail is compelling. The durable economic traffic is still forming.
Executive summary

What is happening right now

A category is forming, but the maturity is uneven.
1 · Commercial reality
Bounded checkout is ahead

OpenAI Instant Checkout is live in bounded form. Stripe has productized ACP and expanded agentic payment methods. Mastercard has now announced live authorized transactions in Europe and Latin America. This is the most commercially credible segment.

2 · Protocol direction
Standards are multiplying

ACP, AP2, UCP, Trusted Agent Protocol, Agent Pay, x402, and MPP all point to the same conclusion: the central market problem is no longer payment movement alone, but secure delegation, agent identity, and post-transaction accountability.

3 · Machine economy
Stablecoins fit the use case better

For API calls, compute, real-time data, and software-to-software transactions, programmable stablecoin settlement is a better technical fit than traditional checkout. x402 is the clearest proof of traction, though measurement still varies materially across sources.

4 · Forecast
Trust layers decide the market

Over the next 12–24 months, the market is likely to expand through more controlled retail deployments, more merchant integration, and more machine payment protocols. The real moat will sit in authorization, orchestration, and interoperability across rails.

The market split

Two agentic-payments markets now exist side by side

Live nowMerchant rails

Card-led agentic checkout

What it is: AI helps discover and buy products using existing merchant and card infrastructure.

  • OpenAI + Stripe Instant Checkout and ACP
  • Mastercard Agent Pay and network-issued agentic tokens
  • Visa Intelligent Commerce + Trusted Agent Protocol
  • Google UCP / AP2 to standardize the commerce workflow
Closer to scale

Why: Existing merchant relationships, mature dispute processes, known authorization primitives, and consumer familiarity.

Early liveInternet rails

Machine-native payments

What it is: Agents pay for services, APIs, content, data, or compute directly, without human checkout flows.

  • x402 for HTTP-native stablecoin payments
  • Stripe MPP for programmable machine payments
  • Visa card spec for MPP
  • Circle Nanopayments for sub-cent transfers on testnet
Architecturally superior

Why: Better fit for micropayments, global software usage, machine timing, and low-friction API monetization.

Numbers

The metrics that matter in March 2026

The strongest current signal points.
ChatGPT weekly users
700M+
OpenAI says more than 700M people turn to ChatGPT each week, giving agentic commerce a giant discovery surface.
Shopify merchants in queue
1M+
OpenAI says over a million Shopify merchants are coming soon to Instant Checkout after Etsy went live first.
x402 last 30 days
75.41M
Official x402 site shows 75.41M transactions over the last 30 days at time of writing
x402 last 30 days volume
$24.24M
Volume remains modest relative to hype, but it is enough to prove non-trivial usage.
x402 buyers
94.06K
Proof that machine-native payments are attracting a real usage base, even if still early.
x402 sellers
22K
A meaningful count for internet-native monetization, though not yet evidence of mainstream adoption.
AP2 collaborators
60+
Google says more than 60 organizations are shaping AP2, reflecting broad interest in common payment rules.
Mastercard LAC participants
17
Processors and issuers participated in Mastercard-led live transactions across Latin America and the Caribbean.
Maturity ladder

Where the market stands today

Agentic retail checkout
82
Network-authenticated card flows
74
Merchant protocol adoption
61
Machine payments over MPP
46
Stablecoin-native API payments
42
Sub-cent nanopayments
19

Indicative Anvesan scoring based on what is live, measurable, and commercially usable in March 2026. It is not a market-size forecast.

The hidden giant

Visa’s existing distribution still dwarfs the current agentic market

4.8B

Payment credentials

150M+

Merchant locations

300B+

Transactions processed every year

Near zero

Fraud rate Stripe says it is seeing on early agentic commerce flows

The point is simple: the current agentic market is tiny next to the installed payment base. This is why card-network participation matters. Even modest agentic adoption riding existing networks can become enormous very quickly.

Leaders

Who is leading, and where

The market does not have one winner. It has leaders by layer.
Leadership map
Card rail scale
Cross-rail control
Niche or protocol first
Machine-native frontier
Merchant / consumer reach →
Machine-native readiness →
Stripe
Mastercard
Visa
OpenAI
Google
x402
Tempo
Circle
How to read the field

Leadership by segment

CompanyWhy it matters now
OpenAITurned agentic shopping from concept into live consumer behavior through ChatGPT Instant Checkout and ACP.
StripeThe strongest cross-segment operator: co-built ACP, launched MPP, runs the token layer, supports network-led and BNPL agentic flows.
MastercardStrongest public proof of controlled live card-network transactions in Europe and Latin America.
VisaOwns enormous existing distribution and is pushing the trust stack via Intelligent Commerce and Trusted Agent Protocol.
GoogleTrying to standardize the broader commerce workflow through AP2 and UCP, especially around carts, catalog data, and identity linking.
x402Most visible stablecoin-native machine-payment standard in the market today.
CircleDefines the sub-cent endpoint with Nanopayments, but remains in testnet territory.

Anvesan take: Stripe is the most strategically interesting company in the field because it is active in both emerging markets at the same time: consumer checkout and machine-native payments.

Strongest retail signal

OpenAI + Stripe show the clearest merchant-facing consumer use case. Distribution plus payment rails matters more than protocol purity in the first phase.

Strongest network signal

Mastercard currently has the sharpest “real transactions happened” narrative. That matters because most of the market is still selling roadmaps.

Strongest machine signal

x402 + MPP make the strongest case that the second half of agentic payments will be API-native, software-driven, and much more global than retail checkout.

What is live

The timeline that created the current market

The category accelerated fast between September 2025 and March 2026.
September 2025

Google launches AP2. More than 60 organizations join the effort to create a payment-agnostic trust and authorization framework for agents, merchants, and payment providers.

September 2025

OpenAI launches Instant Checkout. U.S. users can buy from U.S. Etsy sellers inside ChatGPT, with more than one million Shopify merchants coming soon.

October 2025

Visa debuts Trusted Agent Protocol. The market begins explicitly treating agent identity and cryptographic proof of authorization as first-class infrastructure.

January 2026

Google and partners introduce UCP for shopping. The push shifts from abstract protocol design to the messy operational layer of carts, catalogs, and retail surfaces.

March 2, 2026

Mastercard + Santander complete Europe’s first live end-to-end agentic payment in a regulated banking framework. Important caveat: Mastercard says it is still a pilot, not a commercial rollout.

March 10, 2026

Circle launches Nanopayments on testnet. The market’s sub-cent payment thesis becomes more concrete, but not yet production-grade.

March 18, 2026

Stripe launches MPP. Visa simultaneously ships a card specification and SDK for MPP, a sign that machine payments are becoming a cross-rail category rather than a crypto-only niche.

March 19–24, 2026

Google updates UCP and Mastercard scales live LAC activity. The market broadens from isolated launches into ecosystem building and regional proof points.

Status classification

Not all “launches” mean the same thing

Commercially live
4
Controlled live pilots
3
Open protocol rollouts
5
Testnet / frontier
1

The market is advancing fast, but it is still easy to confuse an open specification, a branded initiative, a regulated pilot, and genuine commercial usage.

Anvesan interpretation

This is an inflection point for agentic payments. We're seeing companies going from theoretical discourse to taking decisive steps in building and integrating payments infrastructure.

The implications are different. Infrastructure markets are won slowly, through trust, compatibility, merchant onboarding, and operational reliability, not by a single flashy partnership headline.

The stack

How the agentic-payments stack is forming

No company owns the full stack yet.
Interface layer

Where intent starts. ChatGPT, Gemini, retailer assistants, and API-first agent interfaces capture user requests and convert them into transaction-ready instructions.

OpenAI · Google
Commerce orchestration

Where shopping logic lives. Carts, catalog data, real-time pricing, shipping state, returns, discounts, and post-purchase details must all move correctly.

ACP · UCP
Authorization & trust

Where the market is really being won. Mandates, verifiable credentials, shared payment tokens, network-issued agentic tokens, and trusted agent identity sit here.

AP2 · Stripe · Visa · Mastercard
Machine payment layer

Where software pays software. MPP and x402 let services coordinate payment programmatically without forcing human checkout flows into machine behavior.

MPP · x402
Settlement layer

Where value actually moves. Cards, bank rails, stablecoins, and eventually other programmable settlement systems coexist rather than fully replace one another.

Cards · Stablecoins
What this means

Why the stack matters

The visible part of the market is the buy button. The invisible part is harder and more valuable: identity, credential scoping, authorization history, merchant readiness, refund logic, and liability routing. That is why every serious player is moving deeper into protocol and control layers, not just UX.

What the first phase rewards
LayerNear-term moat
InterfaceDistribution and attention
CommerceMerchant integrations and catalog fidelity
TrustScoped credentials and verifiable intent
Machine paymentsDeveloper simplicity and low-friction pricing
SettlementGlobal connectivity and compliance
Anvesan thesis

The stack is converging toward a hybrid model

Card rails will remain dominant where consumer protection, merchant familiarity, and issuer controls matter most. Stablecoin rails will increasingly dominate where payment has to behave like software: tiny, global, instant, and programmatic. The connective tissue between those worlds is where the most defensible infrastructure value will sit.

Bottlenecks

What still blocks mass adoption

The rails exist. The constraints are elsewhere.
Trust

Agent identity

Merchants and issuers need cryptographic proof that an agent is legitimate, acting within permission, and traceable if something goes wrong.

Control

Liability routing

When a transaction is incorrect, fraudulent, or disputed, responsibility must be legible across agent, merchant, PSP, network, and issuer.

Operations

Commerce-state complexity

Prices, inventory, returns, taxes, shipping, loyalty, cancellations, and support all need reliable machine-readable workflows.

Adoption

Merchant readiness

Most merchants are not staffed to chase every protocol shift. This gives orchestration layers a meaningful role.

Economics

Micropayment fit

Traditional payment rails are structurally awkward for tiny, high-frequency machine payments. That is where stablecoin-native methods have an edge.

Measurement

Noisy data

One of the clearest signs of immaturity is that reported activity levels for crypto-native agent payments still vary materially depending on the source.

Constraint Card-led checkout Machine-native payments Why it matters
Authorization proof Medium High Card ecosystems already know how to authorize people. They do not yet fully know how to authorize software agents.
Merchant acceptance Medium High Retailers can be onboarded through existing PSPs; API sellers and internet services are only beginning to expose payment-required endpoints.
Dispute handling Low High Card rails have decades of consumer and issuer processes. Machine-native payments still lack comparable standardization.
Micropayment economics High Low Card economics are poorly matched to tiny machine payments; stablecoin-native flows are naturally better here.
Compliance legibility Medium Medium Both worlds need clearer rules for autonomous action, but card rails start from a more familiar compliance posture.

Anvesan conclusion: the core bottleneck is not whether AI systems can execute payments, but whether such payments can be trusted, appropriately bounded, audited, reversed when necessary, and clearly understood by all participants involved in the transaction flow.

Competitive snapshot

How the main players differ

Same category. Very different strategic positions.
Player What they lead Best evidence Main strength Main constraint Status
OpenAI Consumer discovery and in-chat buying Instant Checkout live with Etsy; 1M+ Shopify merchants coming Distribution and interface control Still bounded and merchant-curated Live
Stripe Cross-rail orchestration ACP, SPTs, Agentic Commerce Suite, MPP, network token support Touches both retail and machine payments Depends on broader ecosystem adoption Live
Mastercard Network-authenticated agentic card flows Europe pilot + live LAC transactions across 17 participants Issuer and network trust Still controlled environments Pilot/live
Visa Trust layer and global distribution Intelligent Commerce, Trusted Agent Protocol, MPP card spec Scale of installed network Less public live transaction evidence than Mastercard Infrastructure
Google Protocol and commerce coordination AP2 + UCP with carts, catalogs, and identity linking Standard-setting leverage Merchant conversion proof still early Protocol
x402 Stablecoin-native machine payments 75.41M tx, $24.24M 30-day volume on official site Best live internet-native signal Measurement quality and economic maturity Early live
Circle Sub-cent payment frontier Nanopayments launch on testnet Clear thesis for machine-scale pricing Not yet proven in production Testnet
Best current positioning

Stripe

The company with the broadest exposure to where the category is going, because it is present in both the visible checkout layer and the invisible machine-payment layer.

Most underrated fact

While Google is competing on AI shopping surfaces, it is trying to define the interoperability grammar of agentic commerce.

Most important warning

Stablecoin-native volume is the cleanest proof of future-fit architecture, but it is still too early to treat today’s numbers as evidence of mass behavior.

Forward view

What the future most likely looks like

Three time horizons, one broad direction.
Next 12 months

Controlled expansion

  • More bounded merchant rollouts
  • More network-token launches
  • More UCP / ACP / AP2 convergence pressure
  • More machine-payment experiments tied to real services
  • More measurement and benchmarking tools
Next 2–3 years

Standardization fight

  • Trust layers become table stakes
  • Machine agents begin paying for software routinely
  • Merchant tools absorb protocol complexity
  • Stablecoin settlement grows underneath invisible experiences
  • More explicit regulatory language around autonomous payment authority
Long term

Payments behave like software

  • Agents transact continuously, not episodically
  • Card and stablecoin worlds coexist rather than replace one another
  • Pricing models shrink from subscriptions toward usage-based economics
  • Trust and orchestration providers capture the most durable value
Most likely path

The next big goal is to be invisible.

Agentic payments become important when they stop feeling special. The market wins once payment is folded invisibly into tasks such as replenishment, business procurement, software access, dynamic shipping, travel coordination, data retrieval, and machine-managed subscriptions.

That shift requires trusted rails, but it also requires cultural normalization. Consumers and enterprises will accept bounded delegated payments long before they accept unconstrained agent spending.

Scenario view
ScenarioWhat drives itOdds
Base caseControlled checkout scales first; machine payments grow steadily in API and SaaS nichesHigh
Bull caseProtocol convergence reduces fragmentation and merchant adoption accelerates sharplyMedium
Bear caseFraud, liability disputes, or poor user experience slow trust and keep adoption siloedMedium

Anvesan forecast: the category expands meaningfully over the next two years, but the shape of the expansion matters. Consumer shopping agents likely grow through existing merchant and card systems. Machine-native payments likely grow through APIs, software, cloud, data, and service endpoints where stablecoin settlement has real structural advantages.

Strategic conclusion

The Anvesan view

The market is no longer asking whether agents can pay.

AI Agents moving money is no longer a theoretical thought exercise; it's well established. The defining question is which trust models, payment rails, and orchestration layers will govern that behavior at scale.

1

Card-led agentic checkout is ahead today. It has clearer merchant economics, better dispute structures, and far deeper existing acceptance infrastructure.

2

Stablecoin-native machine payments fit the long-term software use case better. That does not mean they win retail first. It means they may own the part of the market that looks the most like the internet.

3

Trust infrastructure is becoming the real product. Mandates, scoped credentials, trusted-agent signatures, and verifiable intent are the category’s new core primitives.

4

Protocol wars will matter, but orchestration layers may matter more. Most merchants and enterprises will not want to rebuild each time standards change.

5

The eventual winners will span worlds. They will connect card rails, bank logic, stablecoin settlement, merchant systems, and software-native pricing models.

What to watch next
  • More live issuer and merchant launches
  • Protocol convergence or fragmentation
  • Growth in real machine-payment endpoints
  • How fraud and liability are handled
  • Whether stablecoins remain backend rails or become visible user rails
Method

This brief emphasizes official primary sources and current public company statements, not broad speculation. Numbers and claims are intentionally biased toward what is verifiable right now in March 2026.

Use

This report is designed as a market-state note: to frame the field, identify current leaders, isolate the strongest numerical signals, and sketch the likely direction of travel without overstating maturity.

Selected sources

Primary references used in this report

All recorded in March 2026.
  1. OpenAI, Buy it in ChatGPT: Instant Checkout and the Agentic Commerce Protocol. openai.com/index/buy-it-in-chatgpt/
  2. Stripe, 10 things we learned building for the first generation of agentic commerce. stripe.com/blog/10-lessons
  3. Stripe, Introducing the Machine Payments Protocol. stripe.com/blog/machine-payments-protocol
  4. Stripe, Supporting additional payment methods for agentic commerce. stripe.com/blog/supporting-additional-payment-methods-for-agentic-commerce
  5. Visa, Enabling AI agents to buy securely and seamlessly. corporate.visa.com/en/products/intelligent-commerce.html
  6. Visa, Trusted Agent Protocol. developer.visa.com/use-cases/trusted-agent-protocol
  1. Visa, Card specification and SDK for Machine Payments Protocol. corporate.visa.com/.../machine-payments-protocol.html
  2. Google Cloud, Announcing Agent Payments Protocol (AP2). cloud.google.com/.../announcing-agents-to-payments-ap2-protocol
  3. Google, Universal Commerce Protocol updates improve AI shopping for retailers. blog.google/.../shopping/ucp-updates/
  4. Mastercard, Santander and Mastercard complete Europe’s first live end-to-end payment executed by an AI agent. mastercard.com/.../europe...
  5. Mastercard, Agentic payments in Latin America and the Caribbean with live transactions completed across the region. mastercard.com/.../latin-america...
  6. x402, Payment Required. x402.org
  7. Circle, Circle Nanopayments Launches on Testnet as the Core Primitive for Agentic Economic Activity. circle.com/blog/...
Source note

x402 measurement remains early and can vary across dashboards. This report uses figures displayed on the official x402 website at time of writing, while treating them as proof of activity rather than a final market-measurement standard.