21 March 2026

Visa Inc (NYSE: V) — The Chokepoint Under Siege

Applying the Chokepoint-to-Moat Framework to Visa’s Structural Position in an AI-Agent World

EQUITY RESEARCH · MOAT ASSESSMENT

Why I'm Looking at This

Visa has always struck me as one of those businesses that doesn’t need to be clever. It just needs to be in the room. Every room. And for thirty years, it was. The 2–3% interchange fee isn’t really a processing charge — it’s closer to a tax on the way people shop: the thrill of points, the comfort of fraud cover, the habit of tapping and walking away. Most of us never think about it. That’s the point.

What’s changed — and what prompted me to actually sit down and do this properly — is the arrival of a buyer who does think about it. AI agents don’t feel brand loyalty. They don’t want miles. They have a budget and instructions, and they’ll route to the cheapest rail every single time, at scale, without hesitation. That’s a genuinely new kind of threat, and Citrini Research laid out the sequence pretty compellingly back in February. The market has already repriced 19% of the risk. The stock is now at 22x forward earnings on $21.6B of free cash flow — which, for a business of this quality, is not expensive.

My read: the disruption is real, but slow. This isn’t a cliff — it’s a gradient. I’d hold here and watch the Q3 2027 earnings cycle as the first hard data point. That’s where we’ll start to see whether agents are actually diverting volume, or whether the fear was ahead of the reality.

Part I — The Framework

I came across RAISINI’s Chokepoint Framework a few weeks ago — originally written about the Strait of Hormuz — and it struck me as unusually useful for thinking about corporate moats. The core idea is simple: power doesn’t live in scale or visible strength. It lives in position. The chokepoint. The place where traffic has to pass through whether it wants to or not.

Applied to companies, the framework asks one question that most moat analyses quietly sidestep: is this moat actually robust, or has it just never been properly tested? Visa has seen off PayPal, Square, Apple Pay, bank-direct rails, buy-now-pay-later. It’s survived all of them. But here’s the thing — none of those challengers were perfectly rational. They still needed humans to choose them. The AI agent doesn’t need to be chosen. It just optimises. That’s a different kind of test, and Visa hasn’t faced it yet.

Part II — The Six-Dimension Moat Scorecard

Scored 1–5 per dimension. Total: 19 / 30.

Part III — Dimension-by-Dimension Analysis

D1 · Economic Flow Centrality — Score: 4 / 5

Visa is genuinely central to global commerce. It processed $15.4 trillion in payment volume in FY2025, across 4.4 billion cards and 130 million merchant acceptance points. For human commerce — which still constitutes effectively 100% of all meaningful consumer spending — Visa is a structural condition, not a participant. Removing Visa from the payments system overnight would cause severe degradation across retail, travel, e-commerce, and B2B procurement.

The score is 4, not 5, for two reasons. First, Visa occupies the network layer only — it does not own the issuing relationship or the consumer credit product. Mastercard is a near-perfect substitute for the network function, which constrains ultimate pricing power. Second, and more importantly: the economic flow question must now be asked separately for AI agent commerce. For agents, Visa is not yet a structural condition. Agent-driven commerce is choosing its rails now, and those rails may not yet be Visa’s.

D2 · Switching Barrier Architecture — Score: 3 / 5

For human consumers and merchants, the switching barrier is classically severe. The Visa network effect is a textbook two-sided coordination problem: merchants accept Visa because consumers carry Visa; consumers carry Visa because merchants accept it. Re-routing human commerce off Visa rails would require simultaneous coordination across billions of cardholders and 130 million merchants globally — the equivalent of re-routing supertankers away from Hormuz while the oil is already in the pipe.

For AI agents, the switching barrier is near-zero. An agent has no stored card credentials, no habitual checkout behaviour, no brand preference, and no psychological attachment to rewards. It carries a budget and instructions. The cheapest, fastest rail wins every transaction, every time, with no switching cost. This is the structural novelty that Citrini identified and the market has begun to price. This score reflects the dual-universe reality: very high for today’s commerce, structurally declining as agent share grows.

D3 · Pricing & Condition-Setting Power — Score: 4 / 5

A 60.0% operating margin and 50.1% net profit margin are among the most extraordinary financial statistics in the S&P 500. Visa does not just participate in commerce — it sets the conditions under which commerce happens. Its brand standards, network rules, interchange framework, and dispute resolution protocols define the operating environment for every merchant, every bank, and every consumer using its rail. This is genuine condition-setting power in the Raisini sense.

The score is 4, not 5, because this power is contingent on human psychology persisting as the basis of commerce. The interchange fee structure — which funds rewards, fraud protection, and dispute resolution — is justified by human needs. Remove the human from the transaction and the fee loses both its justification and its enforceability. An AI agent will not accept a 2.3% rake to get “purchase protection” it does not need. Pricing power today is excellent; pricing power in a world of agent-dominated commerce is unproven.

D4 · Customer Optimisation Depth — Score: 3 / 5

Banks have built entire product lines around Visa rails. Merchants have embedded Visa into POS infrastructure, e-commerce checkout flows, and auto-renewal billing. Corporate treasury departments have integrated Visa commercial card programmes into expense management and ERP systems. The depth of integration in the incumbent human commercial stack is real, and it creates genuine inertia.

However, the emerging commercial actor — the AI agent — has zero optimisation depth for Visa. Agents start as a clean slate. They will integrate deeply with whatever rail they choose at the outset of the agent commerce era. This creates a discontinuity risk: the legacy human integration layer is deep, but it does not transfer to agents. The first-mover advantage in the agent payment stack belongs to whoever builds the native agent rail, not to whoever owns the human card rail.

D5 · Moat Durability & Threat Lead Time — Score: 3 / 5

Building a replacement for Visa in human consumer commerce would take 10+ years and require regulatory approvals, issuer relationships, merchant acceptance infrastructure, and consumer trust at global scale. On this dimension, Visa’s moat is genuinely durable for its existing customer base.

The threat, however, is not to replace Visa in human commerce. The threat is to bypass Visa for the new category of commerce: machine-to-machine, agent-to-API, stablecoin-settled micropayments. This category is being built now. Tempo went live on mainnet in March 2026. Circle’s Nanopayments are on testnet. Mastercard acquired BVNK for $1.8 billion. The x402 protocol has crossed 50 million transactions. The bypass infrastructure is assembling.

Critically, Visa is not standing still. Visa CLI, its Tempo design partnership, and its stablecoin prefunding initiative reflect genuine understanding of the threat and active positioning inside the new rails. The score of 3 reflects the shortening threat lead time — not a verdict that the threat will materialise on the Citrini timeline.

D6 · Latent Fragility (Stress Test) — Score: 2 / 5

This is the dimension that matters most for long-term investors, and it is where Visa scores weakest. The framework’s core inversion question: is this moat genuinely robust, or has it simply never been tested? For Visa, the answer is honest: the moat has been tested against everything the pre-AI world could produce — PayPal, Square, Apple Pay, buy-now-pay-later, bank-direct rails — and it has survived all of them. That is a genuinely strong track record. It has never been tested against a rational, fee-minimising agent with no psychological attachments.

The Token Dispatch puts it precisely: “Visa is being disrupted by a better customer. One that has no use for what made Visa great.” The 2–3% interchange fee is, at its structural foundation, a tax on human irrationality — on the desire for rewards, the reliance on fraud protection, the friction of comparison-shopping. Remove irrationality from the buyer, and the justification for the tax dissolves.

Regulatory risk adds a second layer of latent fragility. Antitrust scrutiny of interchange is building globally. The EU’s Interchange Fee Regulation already caps consumer card interchange. Any forced extension to commercial cards or cross-border transactions hits Visa’s revenue line directly. The score of 2 is not a prediction of failure — it is a flag that this moat has significant untested exposure, and that the test may come sooner than consensus expects.

Part IV — Financial Quality Snapshot (FY2025)

The FCF yield sits around 5.5% on a $394B market cap. Not screaming cheap, but historically inexpensive for Visa specifically — and the cash flows are real. $21.6B in free cash flow last year, growing revenue 11.3%. Nothing about these numbers says distress. What they do say is that the market has taken a view on the terminal value and marked it down. Whether that markdown is too aggressive is the whole question.

Part V — Risk Register: Citrini / Token Dispatch Scenario Overlay

One distinction worth holding onto: Visa is not American Express. Visa is a network — it collects network fees and steps back. Amex owns the whole P&L, including the interchange, and its entire brand proposition is built around a human making a conscious, aspirational purchase. That human choosing to flex the gold card. An AI agent isn’t doing that. So if I had to pick the more vulnerable business in this scenario, it’s Amex, not Visa. Token Dispatch called it “the cleanest version of this problem.” I think that’s right.

Part VI — Where I Land

I don’t think this is a sell. The Citrini and Token Dispatch theses are genuinely interesting — I found myself reading both pieces twice — but the scenario they’re describing is 2028, not 2026. Right now you’re paying 22x forward earnings for a business doing $21.6B in free cash flow with 60% operating margins. The disruption is real. But disruption at this scale is slow. It doesn’t arrive on a schedule.

The market has already taken 19% off the peak. Some of that fear is justified. Some of it, I suspect, is consensus running ahead of the data. At current levels, I think the risk/reward has improved enough to hold, and to add more if it gets cheaper. The investment case essentially comes down to one question: can Visa embed itself in the new agent/stablecoin rails before those rails make the old ones irrelevant? They’re trying. Visa CLI, the Tempo partnership, stablecoin prefunding — these aren’t PR moves. But whether they’re enough, and whether they’re fast enough, I genuinely don’t know yet.

What Would Change My Mind

I’d reduce if…

Mastercard or Visa management start commenting on “agent-led price optimisation” at Q3 2027 earnings. If purchase volume in travel, subscriptions, or digital goods starts decelerating in a way that correlates with agent penetration, that’s the canary. I’d rather act on the first credible signal than wait for the headline.

I’d add more if…

Visa starts reporting meaningful transaction volume through its own agent and stablecoin rails. That would tell you the bear case — that new rails bypass Visa entirely — is losing. The bull case is that agents still settle on Visa’s tokenised network because Visa got there first. April earnings (Visa Q2 FY2026) is the first chance to hear management frame this directly.

On the Citrini Scenario

I want to be honest about this: the “2028 Global Intelligence Crisis” is a scenario exercise. Citrini say so themselves. The mechanism — agents routing around interchange via stablecoins — is structurally coherent. The timeline is aggressive. Consumer trust in agents for real discretionary spending is the bottleneck, and in my experience behavioural shifts like this always take longer than the technology allows for.

The part that actually keeps me up at night isn’t the consumer commerce disruption. It’s the micropayment category — fractions of a cent per API call, hundreds of calls per session. That was never on Visa rails. It’s being built natively on x402, Nanopayments, and Tempo right now. Visa isn’t losing revenue there. But if that category becomes the dominant pattern for agent commerce, it builds habits and infrastructure that could eventually come for the consumer commerce category too. That’s the longer-tailed risk.

How I’d Position It

Size: Not a full conviction holding given the uncertainty. Somewhere around 2–3% of a portfolio feels right — meaningful enough to matter if the bull case plays out, manageable if the disruption timeline accelerates.

Pairs idea: Long Visa, short American Express. If agents are the threat, Amex is structurally more exposed — its whole model is built on the aspirational human purchase. That’s a very specific thing agents don’t do.

Key dates to watch: Visa Q2 FY2026 earnings in April 2026. Mastercard Q3 2027 earnings is the first real data point on agent volume. Any stablecoin legislation in the US or EU would be a binary event for the whole thesis.

“How do I know this will matter? Because Visa spent $1.8 billion this week making sure it is not left out of the answer.”
— Token Dispatch, March 2026

That sentence is both the bull and the bear case in one line.