硅谷Fintech观察⑥｜Agent 钱包大战：四个协议抢标准，Cloudflare 却在收地租Silicon Valley Fintech Watch ⑥ | The Agent Wallet Wars: Four Protocols Fight Over the Standard, While Cloudflare Collects the Ground Rent

Silicon Valley Fintech Watch series · Part Six

In brief. Roughly one in five websites on earth sits behind the same company. On August 4, 2026, that company handed every AI agent a wallet, for free. Meanwhile four payment protocols and nine giants are fighting over the standard for how agents pay: Stripe is betting on everything, Google donated its own protocol, OpenAI's shopping button died five months after launch, and even the agent programs at Visa and Mastercard stand on identity bricks laid by that same company. My read: this war looks like a fight over protocols. It is actually a fight over where settlement happens. Read on.

Updated August 8, 2026. Stripe is in talks to acquire OpenRouter for about $10 billion, reaching over to the side where agents spend. The same logic, replayed on another flow. I unpack that deal in Part Seven.

Before we start

On August 4, 2026, Cloudflare launched something new: cloudflare.pay.

You read that right. Not .com. Dot pay. The page is minimal: an orange card, one input box, one line: Reserve your Cloudflare Wallet. You type a name, and it holds @yourname, plus an address at yourname.cloudflare.pay, for you. Free, first come first served. I typed a four-letter name and had it in under a minute.

This wallet is not for people. It is for agents. Cloudflare's pitch: when your AI agent goes out to work, it pays for APIs, content, and data out of this wallet. You set the allowance and the whitelist, it spends inside the cage, you don't have to watch it, and it can't max out your card.

A company that grew up selling CDN and firewalls is suddenly issuing wallets. To understand this, zoom out. Over the past year and a half, Silicon Valley has produced four protocols and nine contenders around one question: how do agents pay? A coverage chart has been making the rounds that lays the table out neatly. Four protocols across the top (ACP, MPP, AP2, x402), nine players down the side (Stripe, Visa, Mastercard, Coinbase, Google, OpenAI, PayPal, Cloudflare, Circle), with each cell marked creator or partner.

Two things jump out of that chart. Stripe's coverage bar is the longest: it touches all four protocols. Cloudflare's is nearly the shortest: one cell, as creator of x402.

Weakest hand at the table? My read is the opposite. This piece walks through three things: what this wallet actually is; what the four protocols and five major players are each fighting for; and why the most lopsided player on the chart may be holding the most valuable thing in the game.

1. Take the wallet apart first

Plain language. Cloudflare Wallets has two layers.

The first is the Account Wallet. This is the wallet you hold as a human. You can fund it and withdraw from it, in dollar-pegged stablecoins. It hangs off your Cloudflare account, and the @name is its street address.

The second is the Virtual Wallet: an allowance account you carve out of your Account Wallet and hand to a specific agent, operated through an API key. You can put three locks on it: a total balance cap, a per-transaction cap, and a merchant whitelist. The agent can only spend inside those three locks.

An analogy: this is not binding your bank card to an agent. This is giving your intern a corporate card with a limit, valid only at a set list of stores. However much the intern screws up, the damage is capped at the allowance.

There is a quieter layer under the wallet: identity. Every wallet sits on a verified Cloudflare account. When an agent pays, it can optionally disclose who it represents, and the receiving side can verify who authorized the request. It can also stay anonymous. In one move this half-answers the two most painful questions in the agent economy: whose bot is this, and who stands behind its spending?

The underlying protocol is x402. The name comes from HTTP status code 402, officially Payment Required. That code was written into the HTTP standard back in 1997 to hold the slot for paywalls, and then sat unused for more than two decades. What x402 does is wake the dormant code up: a server answers a request with a 402, meaning pay first; the agent pays on the spot in stablecoins and retries. The whole exchange happens inside HTTP. No account opening, no card on file, no human in the loop.

Cold water first, though: as of this writing, the wallet only lets you reserve a name. Funding, withdrawals, issuing virtual wallets to agents: all of that is, officially, coming in the next few months. What you grabbed today is a street address.

But the ambition behind the street address only makes sense at the card table.

2. What the four protocols are fighting over: a map

The names are a mouthful, so here is an anchor you can keep: the three players who actually decide this game are Stripe, Coinbase, and Cloudflare. Everyone else is picking sides among their routes. Protocol by protocol:

ACP: buy inside the chat box. Agentic Commerce Protocol, maintained by OpenAI and Stripe. It covers the flow where a person says "I want this" in ChatGPT and the agent places the order: merchants hand their catalog to OpenAI, payments run through Stripe, the user taps confirm in chat. The protocol carries a famous footnote. It debuted alongside ChatGPT's Instant Checkout, and in March 2026 OpenAI quietly shut Instant Checkout down. Multiple industry analyses say the flagship consumer surface ran about five months with next to no sales. The protocol survives, but it proved something: putting a buy button inside a chat box does not mean people buy inside a chat box.

AP2: who authorized this, and then Google gave it away. Agent Payments Protocol, launched by Google in September 2025 with sixty organizations behind it, Visa and Mastercard included. It governs not how money moves but how blame is assigned: the user signs a cryptographic credential for the agent (the protocol calls it a Mandate, think of it as a power of attorney), the agent carries it on every purchase, and when a dispute comes up you can verify whether the user really authorized this or the agent went rogue. On April 28, 2026, Google donated AP2 wholesale to the FIDO Alliance (the body behind fingerprint and passkey standards), with sixty organizations co-signing. FIDO stood up two working groups: the authentication group chaired by CVS Health, Google, and OpenAI, and the payments group chaired by Mastercard and Visa. Translation: Google stepped down from dealer to referee, because at the authorization-and-liability layer, whoever monopolizes it can't get it adopted. And the card networks grabbed the payments chairs because "who pays when it goes wrong" has been their family estate for fifty years.

MPP: Stripe's checkout counter for machines. Machine Payments Protocol, released March 18, 2026, co-authored by Stripe and Tempo. It covers the merchant side: drop in a snippet of code and your site can take agent money over three rails, stablecoins on the Tempo chain, the Lightning Network, and credit or debit cards via Stripe. Visa built the first card SDK for it; Anthropic, OpenAI, Mastercard, and Shopify have all announced integrations.

x402: make payment part of HTTP itself. Proposed by Coinbase in May 2025, with Cloudflare as the earliest co-builder; the two together launched the x402 Foundation. The scoreboard deserves a close look. By Coinbase's mid-2026 count, the protocol handled more than 169 million payments in its first year, across 590,000 buyers and 100,000 sellers. Note the structure: most of those 590,000 buyers are not people. They are machines. Another industry figure from March 2026: roughly $600 million in annualized volume, with zero protocol-level fees. What is $600 million? Visa clears somewhere north of ten trillion dollars a year. x402 today is a rounding error on a rounding error. But what happened on July 14, 2026 shows nobody dares ignore it: the Linux Foundation took over the x402 Foundation, and the first forty members included Google, Stripe, Visa, Mastercard, Cloudflare, and Coinbase. Full attendance.

Put the four together and the map is clear. AP2 covers who authorized it. ACP, and Google's separate UCP, cover how the order gets placed. MPP and x402 cover which rail the money rides. Entry, liability, rails: three layers. And the two rail-layer protocols carry two different worldviews. MPP says machine payments should stay compatible with humanity's old rails: cards, banks, chains, plug them all in. x402 says machine payments don't need humanity's rails at all: HTTP is the rail.

That is the real fork in the road. Everything else is seat-picking. And there is a fourth layer hiding off the chart, which we will hold for now: before any transaction, someone at the door has to ask, who is this bot? That layer isn't drawn on the coverage chart, and it is the key to Cloudflare.

3. The research desk: five majors, hand by hand

Protocols aren't enough. This section walks the five real heavyweights one by one: their cards, their play, their weak spot, and which layer they camp on.

Stripe: the full-stack combo, betting that whoever wins, the register is theirs

Stripe has the widest spread on the table, and it isn't just signing partnerships. It is pouring concrete.

Look at what it unloaded at Sessions 2026 in one go. The Agentic Commerce Suite: merchants upload a catalog to the Stripe dashboard and sell directly inside AI surfaces, plus announced partnerships with Meta and Google, so orders placed in Gemini and Google's AI Mode can settle through Stripe. Issuing for agents: programmatically issue single-use virtual cards with real-time authorization and spend controls; the agent swipes once and the card burns. And Link's agent wallet, which lets its consumer fast-checkout accounts, more than a hundred million of them, be delegated to agents. Entry, issuing, wallet, acquiring: the whole combo meal.

Add the two aces already on the table: ACP (with OpenAI) and MPP (with Tempo). Tempo deserves its own sentence. It is a payments-only Layer 1 incubated by Stripe and Paradigm: no native token, fees paid directly in stablecoins, roughly 0.6-second finality. In early 2026, per The Block and others, it raised $500 million at a $5 billion valuation, led by Thrive and Greenoaks; mainnet went live in March, and the first outside validators arrived in April: Stripe, Visa, and Standard Chartered's Zodia.

Stripe's worldview is plain: don't bet on a rail, bet that merchants will always need a register. If cards win, it has acquiring. If chains win, it has Tempo. If chat boxes win, it has ACP. It even holds a seat in rival Coinbase's x402 foundation. Its bar on the coverage chart is longest not because it knows which road wins, but precisely because it doesn't, so it bought a ticket for every road.

That is also the weak spot: at every layer of the combo, there is a more focused opponent. It has more enemies than anyone else at the table.

And its heaviest bet is not in payments at all. Around July 23, 2026, the Wall Street Journal reported that Stripe was in talks to acquire OpenRouter, a routing layer that fans requests across some seventy providers and four hundred-plus models, for roughly $10 billion. It is the largest acquisition Stripe has ever attempted, and the only move at this table that reaches over to the side where agents spend. That deal gets its own piece: Part Seven.

The card networks: the rulekeepers' counterattack

Visa and Mastercard are near mirror images, so take them together.

Visa launched Intelligent Commerce in April 2025 (tokenized card credentials handed to agents), then the Trusted Agent Protocol in October, a verification framework that helps merchants tell malicious bots from legitimate agents. It plans to merge the two into one SDK, AP2-compatible by default, and has said publicly it is aligning x402 interoperability with Coinbase. Mastercard launched Agent Pay in late April 2025, built on Agentic Tokens, dedicated payment credentials for agents. Then in March 2026 it swung big: agreeing to acquire stablecoin infrastructure firm BVNK for up to $1.8 billion.

The key to reading all this is knowing what the card networks' real asset is. It is not the clearing network; new rails can route around clearing. It is the rulebook that has run for fifty years: chargebacks, dispute arbitration, fraud liability. Every move above is about carrying that rulebook into the new world: push card tokens into every protocol, chair the FIDO payments group, buy a stablecoin firm to cover the new rail. This is not offense. It is a counterattack in defense: you can lose battles, but you cannot lose the rulebook.

The weak spot is just as obvious: the biggest increment in the agent economy is machine-to-machine micropayments, API calls at a few cents each, and card economics simply don't clear at that size. Which is why both firms talk cards and buy chains.

And one detail I most want you to keep from this entire piece: Visa's Trusted Agent Protocol and Mastercard's Agent Pay, at the layer that verifies which agent is knocking, both use the same thing: Web Bot Auth. Who drove that standard? It was drafted by Cloudflare engineers and taken to the IETF, and Cloudflare's own blog spells it out: we are working with Visa and Mastercard to help agents transact safely.

The identity brick at the front door of both card networks' agent programs was laid by Cloudflare.

Google: eat the entrance, give away the rules

Google plays both ends.

At the entrance, it bets heavy and keeps the meal: in January 2026, at the retail industry's big annual show, it launched UCP (Universal Commerce Protocol) with Shopify, covering product discovery, carts, and checkout, compatible with A2A, AP2, and MCP. The same month it began rolling out direct checkout inside AI Mode and Gemini with selected US merchants. You buy without leaving Google's surface, but the merchant remains the merchant of record. Google absorbs the journey, not the ledger.

At the rules end, it gives the family silver away: AP2 donated to FIDO, Google reduced to one seat among many.

That push-and-pull exposes Google's real position. Its fear is not missing payments. It is agents gutting the ad tax base. Search advertising is built on humans browsing: the longer you wander and the more you click, the more the ads are worth. An agent browses for you and goes straight to the answer, and the ad slots hang in the air. So the entrance must be defended in person, checkout kept inside its own walls, while the liability layer, which no single owner could ever get adopted anyway, is donated in exchange for a referee's chair.

Coinbase: the Stripe of the on-chain half, and its wallet shipped first

Coinbase is the full-stack player of the on-chain half, and it moves faster than most people register.

Protocol layer: x402 is its proposal. Wallet layer, note the date: on February 11, 2026, Coinbase shipped Agentic Wallets, MPC wallets for agents (key shards held in custody), with session caps, per-transaction limits, gasless settlement on Base, and native x402 support. Nearly half a year ahead of Cloudflare's wallet. Acquiring layer: its developer platform has an SDK that adds agent payments to any API or MCP service in three lines of code, and since July 23, 2026, Coinbase Business accounts accept USDC paid directly by agents. Clearing layer: roughly 85% of x402 transactions run on its own Base chain.

Protocol, wallet, acquiring, clearing: four layers deep. That is why I call it the Stripe of the on-chain half. The difference between the two Stripes is worldview: Stripe expands from the card half toward chains, Coinbase expands from the chain half toward merchants, and they will meet in the middle with swords out.

As for Coinbase and Cloudflare: honeymoon allies for now; x402 is their shared child. But the overlap at the wallet and settlement layers is already sitting on the table. The difference, for the moment, is audience. Coinbase's wallet is for on-chain developers; the keys are private keys and gas. Cloudflare's wallet is for ordinary developers and site owners; the key is an account name. Whoever captures the people who don't understand chains wins the second half.

While we're here, the water seller: Circle. USDC is the default settlement currency of nearly every protocol. Per DeFiLlama as of late April 2026, USDC's circulation is about $77.3 billion, and together with USDT they make up over eighty percent of a roughly $318 billion stablecoin float. The harder the protocols fight, the more water gets sold. Circle is the steadiest winner in this war. With one caveat: the day Cloudflare's own NET Dollar goes live is the day this partner becomes an opponent.

Cloudflare: fighting for no layer, laid underneath every layer

Now pull the camera back to Cloudflare and connect the moves of the past year and a half.

May 2025: it drives Web Bot Auth, the agent identity standard that Visa and Mastercard would later stand on. July 2025: Pay Per Crawl, blocking AI crawlers by default and letting site owners allow, refuse, or charge. The first tollgate on the supply side. Its own data shows why it had no choice: by June 2026, 52% of crawler requests on its network were for AI training, up from 22% a year earlier, and more than half of AI crawler traffic was re-fetching pages that hadn't changed. September 2025: it announces its own dollar stablecoin, NET Dollar, fully dollar-collateralized (note: to this day it hasn't even named the chain). July 1, 2026: the Monetization Gateway. The tollgate upgraded: any resource behind Cloudflare, APIs, datasets, MCP tool calls, can now charge per use, settled on the spot in stablecoins over x402. August 4, 2026: Wallets. The paying side gets its purse.

The identity standard, the supply-side tollgate, the demand-side wallet, the settlement currency, the accounting rules (x402). Five pieces. And under them, the hole card: roughly a fifth of the world's website traffic passes through its network.

Why is the ground worth so much? Because in the machine economy, for the first time, payment and delivery live inside the same act. When a human shops online, payment happens at Stripe and delivery happens on a truck. The two are separate, so the payment layer is a vendor you can swap at will. When a machine buys an API or a piece of content, the goods are the HTTP response itself. Payment and delivery happen at the same door, in the same round trip. Protocols can be swapped freely; swapping a protocol is swapping a snippet of code. But where your API is deployed, and whose network shields it, that door does not move. What can be easily replaced collects no rent. What cannot move, does.

That is why the coverage chart lies. Count cells and Stripe has four to Cloudflare's one. Count assets and Stripe's four cells are four lottery tickets, while Cloudflare's one cell has a fifth of the internet underneath it. While everyone argues over whose POS terminal the toll plaza should use, Cloudflare built the toll plaza on its own highway, and wrote the national standard for the ID check at the gate while it was at it.

Even the business model is off the shelf. Cloudflare's most famous move in history is turning other people's paid products into its own free features: in 2014 it launched Universal SSL and gave HTTPS certificates away, flattening the certificate business overnight. Giving wallets away for free is pure muscle memory. The wallet is not there to make money. It is there to turn a fifth of web traffic into a fifth of payment traffic.

The macro seating chart

See it? Five layers from the top down, and every one is a bloodbath, except the bottom one, which for now has a single occupant. Stripe and Coinbase are the full-stack schools of their respective halves. The card networks are the rulekeepers. Google is the entrance-keeper doubling as referee. Cloudflare is the landlord. The full-stack schools have to win every layer. The landlord doesn't need to win any. If the machine economy happens at all, it is standing at the till.

4. Devil's Advocate: five buckets of cold water for Cloudflare

As always, make the opposing case at full strength first, then come back to the judgment.

Bucket one: demand for agent payments has not been proven. The loudest evidence is the death of Instant Checkout: backed by the biggest consumer AI entrance on earth, gone in five months. x402's 169 million payments sound loud, but a $600 million annualized pot at a few dollars or cents a ticket is machine pocket change. Calling this the next Visa is premature.

Bucket two: Cloudflare is not first, and maybe not best prepared. Coinbase's agent wallet shipped half a year earlier and actually works: MPC custody, gasless, native x402. Stripe's single-use virtual cards sit on a full issuing stack and risk engine. Cloudflare's wallet today is a page that holds your name; funding and withdrawal are months out. A first-mover street address versus first-mover products: unsettled.

Bucket three: pipe companies have a terrible record in finance. Telecom carriers sat on billions of users and spent twenty years trying payments. Outside of special cases like Kenya's M-Pesa, they mostly failed. Pipes have traffic; finance takes licenses, risk management, compliance, and trust, and Cloudflare has proven none of them. NET Dollar was announced almost a year ago and still hasn't named a chain. That is not the tempo of confidence.

Bucket four: the wallet is locked inside its own account system. Cloudflare accounts belong to developers and site owners. Ordinary consumers don't have one. It can cover machine-to-machine, but the human shopping half, agents booking your flights and ordering your groceries, runs through Apple, Google, and OpenAI at the entrance and Visa and Mastercard on the rails. None of that is Cloudflare's.

Bucket five: the regulators haven't really looked yet. Since the US stablecoin law took effect in 2025, issuing a stablecoin is a licensed business. And nobody's regulator anywhere has answered who eats the fraud when an agent spends stablecoins autonomously. The day the first big agent-fraud case lands, the whole industry goes back to school.

Now back to the judgment, narrowed on purpose:

Cloudflare doesn't need to win the wallet war. It doesn't even need NET Dollar to work. Its real bet is one sentence: machine-to-machine payment, for APIs, content, data, compute, becomes a real market, and that traffic already flows through its network. Lose the wallet to Coinbase, fail the currency, fine. As long as machines start paying, identity runs on its standard, traffic runs on its network, and the tollgate sits on its edge nodes. It is at the till either way.

So the narrowed conclusion: in the human shopping half, Cloudflare is a supporting actor. In the machine payment half, it is the only player holding both the ground and the gate. The upside is not "the next payments giant." The upside is ground rent on the machine economy.

5. Bay Area startups: who gets steamrolled, who picks up a weapon

Finally, what this night means for startups.

Over the past two years the Bay Area assembled a small cohort at the agent payments-and-identity layer: Skyfire, building agent payment abstraction, raised $9.5 million from backers including a16z's crypto accelerator and Coinbase Ventures; Nekuda, agent identity verification, raised a $5 million seed in May 2025 with Amex Ventures and Visa Ventures participating; Basis Theory, payment credential vaulting, raised a $33 million Series B in October 2025. One industry tally puts the layer's total disclosed funding at just over $50 million.

Now watch the last three months: February, Coinbase ships an agent wallet. May, Stripe ships single-use virtual cards and the Link agent wallet. August, Cloudflare ships a free wallet. A layer of business built on $50 million of venture money, and three platforms lined up to give it away.

That is why I called it extinction-grade. Platform free-ization has one iron law: anything that can be turned into an attribute of an account will end up as an attribute of an account. A wallet is an account attribute. Identity is an account attribute. Spending limits are an account attribute. The SSL certificate industry already demonstrated this with its life.

What paths remain? I can see four.

One: burrow into compliance and licensing. None of the three platforms will touch KYC, money transmitter licenses, or large-value clearing any time soon, and that layer gets more valuable the more regulated it becomes.

Two: build liability and dispute resolution. The x402 world has no chargebacks; a stablecoin paid on-chain is paid. Whoever builds the layer where agent transactions can be traced, refunded, and arbitrated becomes the embryo of the card network of the agent era. The incumbents chairing FIDO and writing $1.8 billion checks for stablecoin firms tells you the entry fee at the center is out of startup range, but the vertical niches still have cracks.

Three: cross-protocol routing. None of the four protocols will swallow the others soon, and both merchants and agents will need a layer that automatically picks the cheapest, best-fit rail. The messier the protocols, the more the router is worth. One caveat, though: this path has already been priced, and the price is far beyond any startup's reach. Part Seven runs that arithmetic.

Four: change seats entirely. Don't build infrastructure; use it. Payment infrastructure being bulldozed to free is pure upside for companies building agents, a cost line going to zero. For content and API owners it conjures a new revenue line: selling to machines by the call. In the age of the rent-collecting landlord, the best position may not be next to the landlord. It may be running a shop on the landlord's land.

Closing

Three sentences for the whole piece.

One: there will be many protocols and many standards, but the place where settlement happens does not move, and whoever holds the place collects the rent.

Two: wallets will be free, identity will be free, and the two things left worth owning are the chokepoint traffic must pass and the answer to "who pays when it breaks."

Three: the entire agent economy today wouldn't round the corner of a card network's ledger. Every judgment above rests on one premise: that machines will truly start spending. When that premise gets proven true or false matters more than everything else in this piece.

(Appendix: two personal notes)

First, on name-grabbing. The day I wrote this piece, I reserved my own four-letter handle on cloudflare.pay. One minute, free.

My habit: on any new platform, if the name is free and takes a minute, claim it first. It is the cheapest option money can't buy. If you never use it, you lose nothing; if the platform takes off and you didn't, your name belongs to someone else. One reminder while we're here: the .pay domain suffix itself is run by a different company (Amazon's registry) and doesn't open to public registration until early 2027. That will be another land rush. Don't confuse the two.

Second, on position. This piece is about agents and protocols, but the logic underneath is the same thing my earlier series kept saying: in a gold rush, the steadiest business was never panning for gold. It is selling shovels and collecting tolls. The agent economy is still at the "is there even gold" stage, but the road builders, gate builders, wallet issuers, and rulemakers are all in position, and they've already had two rounds of public arguments. You don't have to start a company. But you should see one thing clearly: where the next wave of money will flow. Wherever the money flows, the opportunities settle along that road.

(Fact notes: Cloudflare Wallets and cloudflare.pay launch details, the Account/Virtual Wallet mechanics, spending caps and merchant whitelists are from Cloudflare's official blog announcement of August 4, 2026 and same-day coverage including Fortune. The Web Bot Auth collaboration with Visa/Mastercard is from Cloudflare's "Securing agentic commerce" blog post. Pay Per Crawl (July 2025), the Monetization Gateway (July 1, 2026), and crawler statistics (52% vs 22%) are from Cloudflare's blog and public reporting. NET Dollar was announced in September 2025 and remains unlaunched with no chain disclosed as of this writing. Visa Intelligent Commerce (April 2025) and the Trusted Agent Protocol (October 2025, built with Cloudflare and Akamai, to be unified in one SDK with default AP2 compatibility) are from Visa's official announcements; Mastercard Agent Pay (April 2025) and the BVNK acquisition (March 2026, up to $1.8 billion) are from public reporting. Stripe Sessions 2026 items (Agentic Commerce Suite, Issuing for agents, Link agent wallet, Meta and Google partnerships) are from Stripe's official blog; Tempo's raise ($500 million at a $5 billion valuation) is per The Block and others. AP2's launch (September 2025, 60 organizations) and FIDO donation (April 28, 2026) are from Google and FIDO announcements; UCP (January 2026) and the AI Mode/Gemini checkout rollout are from Google's blog and merchant documentation. x402 figures (169 million payments, 590,000 buyers, 100,000 sellers in year one, ~85% on Base) are Coinbase's mid-2026 count; the ~$600 million annualized volume is a March 2026 industry figure. Coinbase Agentic Wallets (February 11, 2026) and Coinbase Business USDC acceptance (July 23, 2026) are from Coinbase announcements. USDC circulation of ~$77.3 billion is per DeFiLlama as of April 29, 2026. The Instant Checkout shutdown timing and "near-zero sales" characterization come from multiple industry analyses, not OpenAI. Skyfire, Nekuda, and Basis Theory funding figures are from Tracxn and public reporting. Market-size forecasts were deliberately excluded due to divergent methodologies. Where the circulating coverage chart's creator/partner labels differ from official accounts (e.g., PayPal is officially an AP2 partner), this piece follows the official protocol announcements. The image is the wallet card as rendered by the cloudflare.pay page itself.)