↓ Skip to main content
Expert LinkedIn

SEA Weekly: The AI Agent Arrives at the Checkout

#Fintech #Digital Banking
10 min read
Chloe Tan
Chloe Tan Fintech Product Leader & Digital Banking Strategist

Mastercard went live with authenticated AI-agent transactions in Singapore and Malaysia this week. Vietnam’s most successful consumer fintech — MoMo, profitable since 2024, thirty million users, once targeting a 2025 IPO — is simultaneously shopping itself around at roughly the same valuation it raised at five years ago.

The Mechanics of Agent Pay
#

On April 7, Mastercard launched the first wave of its Agent Pay system in Southeast Asia, in partnership with UOB. The system is built around a concept Mastercard developed with Google called “verifiable intent” — a tamper-resistant record of what a consumer authorised when an AI agent acted on their behalf. The framework includes Mastercard Agentic Tokens, Payment Passkeys, and an audit trail that issuers, merchants, and consumers can all reference.

The pitch is straightforward: if an AI agent buys a flight on your behalf, everyone in the payment chain should be able to verify that you actually instructed it to do that, at that price, with that card. The verifiable intent record is the answer to the obvious question — “who said the AI could do this?” — rendered in a form that is cryptographically verifiable and legally traceable.

UOB is the regional testing partner; local bank deployments are rolling in each market. Mastercard plans to open a regional AI Centre of Excellence in Singapore later in 2026, framed as its largest innovation space in Asia-Pacific and bringing together cybersecurity, payments research, and real-time risk.

The launch is technically a pilot, not a full rollout. But the signal it sends is strategic, not incremental. The card network has declared that it intends to be the trust infrastructure for AI-initiated commerce in Southeast Asia — and it is doing so before any of the region’s super-apps, neobanks, or digital wallets have positioned themselves for the same role. The Mastercard press release frames this as the region “embracing secure, AI-enabled commerce.” What it doesn’t say is that none of the other candidates got there first.

The Five-Year Plateau
#

Now consider what was announced in Hanoi on the same day.

Reuters reported on April 7 that MoMo, Vietnam’s dominant mobile payment platform, is weighing options including new strategic investors, with Jefferies and Morgan Stanley hired to manage the process. The target valuation: above US$2 billion. MoMo’s last major fundraise was US$200 million in 2021, led by Mizuho Bank, at a valuation of approximately US$2 billion.

Read that twice. Five years. Profitable since 2024. Thirty million users. A market that has grown from US$150 billion in annual digital payment transaction value toward a projected US$300–400 billion by 2030, per Bain estimates. Revenue of US$482 million in 2024. And as Fintech News Singapore noted, the IPO that was supposed to close the value gap by 2025 is no longer on the near-term agenda.

The question worth sitting with is not what MoMo is worth. It is what the market is telling you about the category of company MoMo represents.

The obvious framing is that a profitable fintech with 30 million users in a fast-growing market should be worth significantly more than $2 billion today. MoMo’s 2024 revenue of US$482 million at a 4-5x revenue multiple would imply a $2-2.5 billion valuation anyway — which is where they seem to be sitting. The market is not pricing in growth. It is pricing in the current run rate.

Why? Vietnam’s regulatory architecture offers a hint. The five licensed exchanges I covered in my April 5 column — TCEX (Techcombank), CAEX (VPBank), LPEX (LPBank), VIX Securities, and Sun Group — are bank-owned entities with state proximity that the government is building into a parallel financial infrastructure. VietQR, the national real-time payment standard, runs beneath every major wallet. When the central bank and the commercial banking sector are both actively building competing payment infrastructure, the competitive moat around an independent consumer wallet is narrower than it looks. MoMo is not a platform the government built, which means it is a platform the government is building around.

The IPO window closing is the confirming signal. A public market prices future cash flows. Strategic investors price something else — distribution, user base, regulatory relationships. MoMo is apparently more valuable as a distribution asset than as a standalone public growth company. That is the honest summary of this week’s news.

The Rails Are Almost Done
#

While MoMo sought backers and Mastercard launched pilots, Singapore’s cross-border payment infrastructure quietly crossed another threshold.

On April 10, Technode Global reported that Thunes has joined Circle Payments Network Managed Payments — enabling stablecoin-powered settlement for Thunes’ network of 140-plus countries and twelve billion financial endpoints. The partnership builds on a 2024 integration, when Thunes first incorporated USDC into its Direct Global Network to manage liquidity. The new arrangement goes further: Thunes customers — banks, fintechs, gig economy platforms — can now access stablecoin settlement while maintaining their existing fiat workflows. They do not need to “adopt crypto.” USDC handles the back-end; fiat appears at the endpoints.

I have been tracking the assembly of this stablecoin settlement layer since January. My March 22 column covered the Thunes-Swift integration and Triple-A joining Circle. My March 29 column tracked Tazapay’s US$36 million Series B (Circle Ventures-led) and the emergence of Circle Payments Network as the operating spine of this infrastructure. This week’s Thunes announcement is the largest single addition yet: a 140-country payment router joining the network means that any business connected to Thunes — which is most of the cross-border payment stack in Southeast Asia — can now settle in real-time, around the clock, in USDC, without any changes to their existing systems.

The implication for agentic commerce is not abstract. AI agents transacting cross-border — paying a Vietnamese developer, subscribing to a Malaysian SaaS, tipping a Filipino content creator — need settlement infrastructure that is always on, cheap, and global. The Thunes-Circle architecture is now, theoretically, that infrastructure. The Mastercard Agent Pay trust framework verifies the intent. The Thunes-Circle rail settles the value. The two halves of an agentic payment system now exist in the same region.

What is still missing is the consumer interface — the wallet or app through which the AI agent actually acts. Which brings us back to MoMo’s five-year plateau.

What Western Union Paid For
#

One additional data point on the consumer layer: in the week of April 3, Western Union completed its acquisition of Dash, the Singapore digital wallet that Singtel had operated since 2014. Western Union’s first digital wallet acquisition in Asia-Pacific; Dash’s 1.4 million users now sit inside a global remittance network spanning 200 countries.

The detail worth noting is Singtel’s exit. A major regional telco — backed by Singapore’s sovereign wealth infrastructure, with distribution across Southeast Asia — spent more than a decade building a financial services wallet and concluded the right outcome was selling it to a global incumbent payment company rather than growing it further. Singtel’s retreat from Dash is the same logic that drove Kredivo’s acquisition of Timo and that is driving the Revolut acquisition search I covered last week. Building digital wallet distribution at scale, without a full banking license, is harder than it looked in 2018. The banking layer is the competitive moat; the payment interface is increasingly commoditised.

Western Union gets Dash’s Singapore distribution and a launchpad for embedding its global send-and-receive network into everyday local financial activity. What Singtel gets is confirmation that the telco-wallet model requires a partner with global payment scale to realise its value. The deal was not a failure — it was a correct read of where Dash’s value was actually highest.

Whose Checkout?
#

The question that Money20/20 Bangkok (April 21–23) will probably not ask directly, but that every product strategy team in Southeast Asia is now running models on: if AI agents are going to be significant consumers of digital services by 2027–28, which wallets and platforms will they transact through?

The honest current answer is: whichever ones are Mastercard Agent Pay-compatible and have verifiable intent infrastructure. Today, that means UOB and Mastercard’s bank partners. Not GrabPay, which is still building consumer credit product from scratch — the GrabX Cash Loan launch this week in the Philippines, using behavioural signals from ride frequency and food orders to assess consumer creditworthiness, is a genuinely interesting product but reveals that Grab is, in 2026, still at the “teach the risk model” phase of consumer lending. Not MoMo, which is profitable and large and valued at its 2021 price. Not the ShopBack cashback layer, which is reporting a headline profit of US$208 million built almost entirely on preference share revaluation gains while its core revenue declined 2.8% — that kind of accounting alchemy is a distraction from the operational reality.

The uncomfortable truth is that Southeast Asia’s consumer fintech layer — the wallets and apps that a billion people actually use — is powerful in adoption and weak in the kind of institutional trust infrastructure that agentic commerce requires. The institutional layer (Mastercard, card networks, UOB-class banks) is being positioned for the next era before the consumer layer has finished solving the current one.

Near-term, the Money20/20 Intersection Stage will produce a lot of conference conversation about TradFi-DeFi convergence. The sharper conversation will be about TradFi-AI convergence — specifically, whether the verifiable intent framework that Mastercard built with Google becomes the industry standard for agentic authorisation, or whether the super-apps build their own.

Medium-term, MoMo’s investor search is the most important corporate development in Vietnam fintech this year. If a Chinese strategic investor — Tencent, Ant International, or a domestic e-commerce giant — acquires a stake, it reframes MoMo as a Belt-and-Road-adjacent digital financial infrastructure asset. That is a different risk profile than “Vietnamese fintech growth story,” and it is exactly the kind of strategic positioning that the tariff year would rationally incentivise. If a Japanese bank takes the stake instead, you get a vanilla yield-seeking investment and MoMo continues on its current path. The identity of MoMo’s next investor will tell you more about the medium-term direction of Southeast Asian digital finance than any conference keynote.

Longer-term, the architecture now exists: Mastercard authenticates the agent’s intent, Thunes-Circle settles the value in real-time across 140 countries, and the consumer wallet serves as the interface. The question is which wallets are still standing when that architecture hits scale. MoMo, Grab, and the region’s digital banks have eighteen months, approximately, to decide whether they are building toward that architecture or building around it.

References
#

AI-Generated Content Notice

This article was created using artificial intelligence technology. While we strive for accuracy and provide valuable insights, readers should independently verify information and use their own judgment when making business decisions. The content may not reflect real-time market conditions or personal circumstances.

Related Articles

Tags

#Southeast Asia #agentic AI #Vietnam #digital payments #stablecoins #Mastercard #MoMo