Year-End Payments Recap

Summary: B2B Stablecoin and The End of the Interface Era

As we close the books on 2025, the payments industry finds itself at  a moment that future historians will likely designate as the end of the “Interface Era” and the dawn of the “Agentic Era.” For the past three decades, the digitization of payments has been defined by the migration of human intent from POS to digital screens. From the first e-commerce transaction to the ubiquity of mobile wallets, the fundamental atomic unit of the economy remained the same: a human being, interacting with a graphical user interface (GUI), making a conscious decision to exchange value for goods or services.

The events of 2025 have irrevocably shattered this paradigm. The rapid maturation of Agentic Commerce, the strategic bifurcation of blockchain architectures, and the hardening of identity around device-centric roots of trust have collectively dismantled the economic assumptions that underpinned the internet for thirty years. Merchants may no longer invest to further optimize for human eyes. Today, all major players are racing to build infrastructure for a machine-to-machine economy where the “purchase order” replaces the shopping cart, and where the “marketing funnel” is rendered obsolete.

Today I’ll attempt to synthesize this year’s blogs covering: Stablecoins, Agentic Commerce, “Closed Chains bigger than On Chain” thesis, Identity Wars, and the structural evolution of Visa and Mastercard. As Payment Geeks you can see that the metrics of the past (GDV, MAU) may become lagging indicators.

1. Stablecoins in B2B

The single most significant economic development of 2025 was the graduation of the digital asset ecosystem from a decade of speculative experimentation to a decisive phase of infrastructure modernization. For fifteen years, the discourse surrounding blockchain was dominated by the volatility of crypto-assets, effectively obscuring the underlying utility of the technology. That era has concluded.

In 2025, Stablecoins emerged not as a “new form of money” for consumers, but as a fundamental settlement innovation for B2B commerce. The rails are being replaced while the train is moving, and those who understand the mechanics of these new tracks are now determining the destination of global capital.

The catalyst for this shift was the passage of the GENIUS Act. This legislative milestone provided the regulatory clarity required to transition stablecoins from the periphery of finance to its very core.

• Codifying Trust: The Act codified “trust” in the instrument by reducing settlement risk strictly to the stablecoin issuer’s balance sheet.  This legal certainty allowed treasurers and CFOs to finally approve the use of digital dollars for wholesale settlement.
• The Euro-Response: The legislation triggered a geopolitical shockwave. European finance ministers have since declared U.S. stablecoins a greater threat to their monetary sovereignty than trade tariffs. They recognize that a regulated, friction-free digital dollar is the ultimate export, capable of dollarizing B2B supply chains globally at the expense of the Euro.

Crucially, this B2B transformation is an evolution that upgrades the banking system, not a revolution that overthrows it. While the GENIUS Act solved the asset risk, it did not address the broader governance issues—disputes, AML compliance, and participant management. This gap has reinforced the need for established financial institutions to manage the “governance layer” on top of the new settlement rails. While not as pronounced, DLT is beginning to impact core bank internal General Ledger. 

2. The Shift from “Shopping” to “Shop for Me”

Technically, the most disruptive force of 2025 was the transition of Generative AI from a content creation tool to a transactional execution layer: the birth of Agentic Commerce. This is not merely an automation of the checkout button; it is a fundamental restructuring of commercial intent.

In the legacy model (1995–2024), a consumer navigated a “funnel.” They were exposed to awareness marketing (Top of Funnel), engaged in consideration via search and comparison (Middle of Funnel), and then wento the a merchant’s side for selection and checkout  (Bottom of Funnel). The entire global advertising industry (and Google) was built to lubricate this slide.

In 2025, the funnel collapsed. When a consumer delegates an objective to an AI agent—”Restock my pantry with organic staples at the best price”—the agent does not “shop.” It does not browse Instagram for inspiration; it does not click on Google Ads; it does not get swayed by end-cap displays. Instead, it executes a Purchase Order with a Payment and consumer authorization. 

The Walmart-OpenAI Paradigm

The definitive signal of this shift was the strategic alignment between #2 online retailer Walmart and OpenAI. This partnership is the “biggest deal to track” because it represents the fusion of massive, structured inventory data, scaled consumer demand (behavior) with the reasoning engine capable of navigating it, and a brand capable of owning the risk (Walmart).

In this model, Walmart ceases to be a destination for humans and becomes a “headless” inventory cloud for agents. To be clear, WMT is not betting all of its existing consumers will start their search on OpenAI, quite the opposite as it bets that its own agent (Sparky) will drive 50% of GMV by 2030. WMT’s bet is on new demand (given Amazon can’t make the OpenAI bet). 

In this future, OpenAI acts as the procurement officer for the household. This fundamentally threatens the search-advertising dominance of Google. If the agent goes directly to the inventory source (Walmart) via a reasoning API, the “Ten Blue Links” of search become irrelevant. The implications for marketing spend are catastrophic for platforms relying on human attention. Billions in infrastructure investment designed to capture human eyeballs must now be redirected toward “Agent Optimization”—ensuring that product data is semantically rich, accurate, and accessible to the machine buyers of the future. See blog Agentic Commerce Revolution.

Agentic Reality – Not 2026.. But 2027

Despite the technological breakthroughs, reoganizing 30 yrs of infrastructure takes time , as does consumer behavior (and trust), thus there is uncertainty when agentic volume will materialize (see Inevitable or Unworkable)

• The Economic Gordian Knot: The monetization of Agentic Commerce remains an unsolved “Gordian Knot.” The core issue is linking a general mandate (e.g., “Plan my travel”) to a specific financial action in a way that is legally binding and monetizable. No commercial framework currently exists to handle the liability of an agent that overspends or misinterprets an instruction.
• Infrastructure Lag: There is currently “more hype than reality” regarding immediate volume because the infrastructure to manage risk and identity is still developing.2 Until we solve the “Identity” problem—ensuring a specific agent is bound to a specific human and authorized for a specific scope—banks will effectively block these transactions to prevent fraud.
• Governance.

Existing Fraud Infrastructure and Bots

The removal of the human from the loop has triggered a crisis for merchants and fraud services industry. For fifteen years, the defense against digital fraud has relied on device ID and Behavioral Biometrics. Risk engines device IDs, IP addresses, time of day, location …etc.  when a consumer clicking “Buy,”. These “human” signals were the primary differentiator between a legitimate user and a scripted bot attack.

Agentic Commerce invalidates this entire stack. A legitimate AI agent, executing a transaction on behalf of a verified user, behaves exactly like a malicious bot. It moves with superhuman speed. It does not hesitate. It does not mistype. It does not scroll randomly.

• The Blind Spot: Banks and networks are currently blind. They cannot distinguish between a “Trusted Agent” authorized by a user and a “Rogue Agent” draining an account.
• Infrastructure Obsolescence: Billions of dollars invested in behavioral analytics are effectively depreciating to zero. The industry must pivot from probabilistic behavior analysis to deterministic cryptographic identity.
• Future Outlook: We are moving toward a model of “Intent Binding,” where the user cryptographically signs the intent (“Spend $50 at Walmart”), and the agent carries this signature as a bearer token of authorization. While AP2 is the technical spec to make this happen, it requires V/MA for governance and bank “trust” in the digital signatures present (see Trust and Governance, and AP2). 

Purchase Order and RFQ

The “Purchase Order with Payment” model is not new; it is the standard for B2B commerce. However, 2025 saw this logic invade consumer retail. In an agentic world, the payment is the final step in a complex chain of logic.

• Conditionality: “Pay $100 IF delivery is by Tuesday AND product is organic.”
• Negotiation: Agents are “chatty.” A single transaction may involve 50 API calls to verify stock, compare shipping rates, and check warranties.
• Negotiation may involve both pricing (request for quote) and payment terms (ex who owns the risk) and request for customer information (for loyalty pricing)
• This “chatter” drives the growth of Network Value Added Services (VAS). The payment networks are no longer just moving money; they are validating the logic of the trade.

3: The Identity Wars – Silicon vs. Sovereignty

Identity is the bedrock of the eCommerce and the Agentic economy. For years the networks have worked to improve the flow of information from merchant to bank for risk decisioning (see 3DS). For liability to shift to banks, the “device” signal must be very strong (see M2M Governance Gaps). 2025 marked the year Google finally aligned its security philosophy with Apple, shifting decisively to device-bound identity anchored by the Titan M2 architecture (see blog).

For years, the industry operated on a bifurcated security model. Apple insisted on the Secure Enclave (hardware isolation), while Android allowed for a more permissive, software-heavy Host Card Emulation (HCE) environment. This fragmentation held back the deployment of high-value digital credentials.

• Google’s Titan M2: By enforcing a hardware root of trust, Google has admitted that software keys are insufficient for the next generation of financial risk. The Titan M2 acts as a vault, physically isolating private keys from the operating system. If the Android OS is compromised by malware, the keys in the Titan M2 remain secure.
• Implication for Payments: This convergence means that 90%+ of global smartphones will possess a standardized, hardware-grade secure element (note Samsung has Knox). This paves the way for the mobile device to formally replace the physical smart card, not just as a payment instrument, but as a government ID and a regulatory passport.

Of course Google will continue to support FIDO2, but this bi-lateral exchange of credentials for authentication is different than a federated wallet for identification and assertions (see payments and identity)

AP2 – 160+ Partners but Still 12mo+ Away

The launch of Google’s Agentic Payments Protocol (AP2) on September 16, 2025, was the tactical manifestation of this new identity capability. However, it also exposed the deep strategic gaps between Big Tech and Financial Services

AP2 is an engineers dream, leveraging W3C’s Verifiable Credentials and extending them into commerce contexts. A federated architecture designed to create a “Trusted Agent Economy” with no central orchestrator. 

• Challenges: What credential do I use, how was it provisioned and who is managing the governance around the entire process? 
• The Liability Gap: If an automated agent makes a mistake via AP2, sending USDC to the wrong wallet, who makes the consumer whole? In the card network model, the bank absorbs the fraud. In the AP2 “bearer instrument” model, the consumer is on their own. For a liability shift to happen it must happen with bank support and within a well defined card network rule set (ex DAF/TAF).

EU’s eIDAS: The Failed Siege of Fortress America

While American tech giants consolidated control over identity via silicon (Titan/Secure Enclave), Europe is attempting to build a set of digital keys to access anyone’s platform in order to regain sovereignty via regulation (see . eIDAS 2.0 Siege Attempt).

The European vision is a world where a digital Euro identity works seamlessly on a browser, a phone, or a distinct hardware token, without paying the “Apple Tax” or feeding the Google data machine.

• The Reality: The initiative is destined to fail in the near term because it lacks Economic Incentive. 
• The Compliance Trap: eIDAS is a compliance mandate, not a business product. Merchants and banks will do the bare minimum to comply, but without a revenue driver, they will not invest in the seamless user experience required to unseat Apple/Google. Consumers want security and privacy (which Big Tech delivers); they do not care about “sovereignty.” The “Siege” will likely result in a parallel, clunky bureaucratic infrastructure that sees little genuine commercial volume compared to the friction-free rails of the US tech giants.
• Regulatory complexity. At the EU level, eIDAS is mandatory for any business that requires SCA, but local country banking regulators have not mandated compliance for banks. Banks themselve are actively resisting as it is hard for them to own regulatory compliance for KYC and transactional risk when they don’t have any say on the credentials used for authentication and authorization (see EU Bank Complaint). 

4: The Stablecoin Reality Check – Closed vs. On-Chain

The crypto-native ethos of “Code is Law” and “Trustless” architecture has been rejected by the global financial system. Banks deal in liability, not just liquidity.

• Trust vs. Tech: Trust in institutional finance is managed through multi-party legal agreements and regulatory enforcement, not by the consensus algorithm of a blockchain. A smart contract cannot be sued; a counterparty can.
• Privacy: Financial institutions cannot operate on public chains (like Ethereum or Solana) because the metadata of their transactions (volume, timing, counterparty) is proprietary. “Privacy” on a public chain is a contradiction in terms for a bank that relies on informational asymmetry for trading margins.

In this current time of flux, walled gardens and closed networks like J.P. Morgan’s Kinexys (formerly Onyx) possess a significant strategic advantage (see Last Week’s blog – Commercial Frameworks will Define Success).

• Centralized Governance: While the open ecosystem struggles with the “Gordian Knot” of liability and diverse standards, closed networks can unilaterally enforce standards, trust, and governance. J.P. Morgan acts as the “Switch,” arbitrating the trust differential between participants.
• Tokenized Deposits (TDs): By utilizing Tokenized Deposits on permissioned ledgers, these networks create a “clean” environment where every wallet is KYC’d and every transaction is compliant by default.
• Monetization Control: Closed networks can effectively monetize their rails because they control the access points. Unlike public chains where value leaks to gas fees and miners, closed chains capture value through commercial agreements and liquidity provisions.
• The “Switching” Role: Banks are evolving into “Switching” entities. They sit between the public stablecoin world (which they may custody but not settle on) and the private TD world, arbitraging the trust differential.

The “Closed Chain” victory is secured by Commercial Agreements. Complex financial transactions require governance frameworks—AML checks, sanction screening, and dispute resolution.

• Consortium Dominance: We are seeing the formation of bank consortiums that utilize “Orchid Compatible Ledgers” (a nod to Singapore’s Project Orchid) and private chains like hyperLedger . These ledgers bake compliance into the protocol and permission every participant and action. For example, A wallet cannot hold a token unless it has passed KYC. This is hard to enforce retroactively on a permissionless chain (ex Solana Token 2022 Permanent Delegate).

5: Visa and Mastercard – The Network of Networks

As the “Universal Cart” of human commerce fractures into the “Swarm” of agentic commerce, Visa and Mastercard (V/MA) are executing a brilliant strategic pivot. They are positioning themselves not as mere transaction pipes, but as the Governance Layer of the Agentic Economy: A switch for trust, identity, payment, data exchange and optionality. 

In 2024, one transaction equaled one authorization. In 2029, one consumer intent (e.g., “Plan my vacation”) may trigger 20+ distinct agent interactions before a final payment is made.

• The Opportunity: If the web remains “Open,” agents will need to constantly authenticate each other. “Are you a verified travel agent?” “Is this a verified hotel API?”
• New VAS

• • Agent Authentication/Authorization
  • Agent/Platform registration
  • Token/Agent Binding

• Consumer Authentication/Authorization

The headwinds for these VAS are walled gardens, If Amazon and Walmart win, the agentic chatter happens inside their private servers. V/MA only sees the final settlement.

The tailwind is if Google/Apple/OpenAI, …. win and the market remains fragmented. In this future, V/MA becomes the indispensable “Network of Networks,” providing the trust infrastructure for billions of agent interactions.

Strategic Outlook & Recommendations

As we exit 2025, the new hierarchy of the payments ecosystem is clear. It is defined by who controls the Agency and who controls the Liability.

• The Device Layer (Apple/Google): Controls the keys (Titan/Enclave) and the biometrics. They own the “What You Have” and “Who You Are.”
• The Governance Layer (V/MA): Controls the rules of engagement. They monetize the friction between agents.
• The Settlement Layer (Banks/Closed Chains): Controls the capital via permissioned ledgers (Kinexys) and B2B Stablecoins (GENIUS Act compliance).

Areas to Watch in 2026

• • Stablecoin in B2B. Transformational with largest economic impact. 

• Closed Networks and Walled Gardens.

• • OpenAI transaction volume with Walmart

• Google’s AP2 Volume – Aligning 160+ partners in the most strategic long term effort to define how Agentic operates.

• Agentic Fraud Events: Today, Stripe’s ACP dominates because it retains the existing fraud systems investment. However, we anticipate the first major “Flash Crash” of fraud, where rogue agents exploit a logic loop to drain funds at high speed. This will force a crackdown , with hardware-based identity (Titan M2 standards) providing the best alternative.
• The Fall of “Generic” SEO: Marketing budgets will shift almost entirely to data structuring. If your product is not readable by an OpenAI agent, it does not exist. Google is best positioned here. 
• The Protocol Wars: ACP vs AP2 vs ?PayPal? Vs. ??

The era of evolution is over. The “Purchase Order with Payment” is the new unit of commerce. The “Closed Chain” is the new rail of finance. And the “Agent” is the new consumer. The winners of the next decade will be those who stop building for eyes and thumbs, and start building for logic gates and cryptographic proofs.