Agentic Commerce and Stablecoins: What Payment Giants Told Investors

The last week of July was the densest stretch of payments earnings this year: Visa on July 28, then Mastercard and Coinbase back to back on July 30. Listen to those three calls end to end and two phrases do most of the strategic heavy lifting: "agentic commerce" and "stablecoin."

Executives repeating a buzzword is not news. What is news — or at least measurable — is how many companies are saying it, how that number has moved over time, and who exactly is still talking once the hype cools. That is a full-text search problem, and it is exactly what a transcript corpus is for.

So we ran both phrases through the EarningsAPI corpus — 253,000+ transcripts from 12,000+ companies across 175 exchanges — and counted, per quarter, how many companies had at least one earnings call matching each phrase. One methodology note up front: these counts are calls matching a full-text search, so they are an upper bound. An analyst asking a skeptical question counts the same as a CEO announcing a product line. That caveat matters less than you might think, because the interesting signal here is the shape of each curve, not any single number.

Two themes, two very different curves

Here is the quarterly company count for each phrase since the start of 2024:

The late-July calls quoted below land in Q3 2026, so they are not in this table yet. But the table is the context that makes those calls worth reading closely: the two themes are at completely different points in their life cycle.

Agentic commerce is still diffusing

The agentic commerce curve is a classic diffusion pattern. From 89 companies in Q1 2024 to 350 in Q1 2026 — nearly a fourfold rise in nine quarters — with only a mild pullback to 325 in Q2 2026. Every quarter, more management teams decide they need a stated position, and more analysts push the ones who don't.

When a theme is still spreading like this, the mention count itself is noisy. Plenty of those 325 companies are name-checking the term because their peers did, or because an analyst forced the topic. The upper-bound caveat bites hardest here. The useful move at this stage is not to trust the aggregate but to filter for who is committing capital and shipping — which is where the actual calls come in below. A diffusion curve tells you when a theme has become table stakes; it cannot tell you who will monetize it.

Stablecoins already peaked — and that's the interesting part

The stablecoin curve tells a different story. It idled in the single digits through 2024 — 4, 5, 3, 9 companies per quarter — then broke out through 2025: 25 in Q1, 39 in Q2, and a spike to 115 in Q3 2025, the quarter after the GENIUS Act, the US stablecoin framework, was signed in July 2025. That was the peak. The three quarters since have printed 95, 104, and 89.

Falling mention counts usually read as a theme dying. Here the sector composition says otherwise. In Q2 2026, 77 of the 89 companies whose calls matched "stablecoin" were Financials — roughly 87 percent of the theme concentrated in one sector. Compare that with the Q3 2025 spike, when the regulatory headline pulled in commentary from everywhere.

That is what consolidation looks like in transcript data: the tourists left, and what remains is the sector actually building. If you ship fintech products, this second pattern is the more actionable one. It says the narrative phase is over and the infrastructure phase has started — the companies still talking are the ones you should be reading. Breadth-versus-composition shifts like this are invisible in headlines and obvious in a queryable corpus.

The networks: rails for agents, rails for stablecoins

Now to the calls themselves, starting with the two networks.

On Mastercard's July 30 call, agentic commerce was framed as a core thesis rather than a curiosity: "Agentic Commerce is the next evolution in payments where the importance of security, transparency and control only increase. Agentic Commerce creates a significant opportunity for Mastercard. It leads to incremental transactions."

Note the last phrase. "Incremental transactions" is network economics in three words: if software agents compare, negotiate, and reorder on a human's behalf, each unit of human intent fans out into more transactions — and a network takes a toll on every one. On stablecoins, Mastercard laid out the scaling criteria — "there are a few essential principles for it to scale: reliability, security and interoperability" — and pointed to its initiative launched with "more than 30 industry leaders participating."

Visa, two days earlier, was just as direct about priorities, naming stablecoins and agentic commerce as its two focus deployment areas. On the stablecoin side, the language was full-stack: "We are active and investing in each layer of the stablecoin stack from blockchain to issuance, wallets, infrastructure and orchestration and applications."

Read together, the network position is symmetrical and defensive in the best sense: whichever front-end wins — agent-initiated checkout or stablecoin settlement — both companies intend to be the rails underneath it, supplying the security, tokenization, and interoperability layer that the new flows will need.

The exchange and the merchant platform

The other two calls worth your time sit at different layers of the same stack.

On Coinbase's July 30 call, CEO Brian Armstrong described a distribution strategy, not a single-coin bet: "we're a multi stablecoin platform. We want to provide the stablecoins that all of our customers want to use. And where possible, we want to strike good economic arrangements with that."

That is an exchange planning for a multi-issuer world — carrying every stablecoin customers demand and negotiating economics with each issuer, the way a retailer stocks competing brands and takes margin on all of them. It fits the consolidation signal in the table: post-GENIUS Act, the question inside Financials has shifted from "will stablecoins matter" to "who captures the economics of distribution."

Fiserv, back on its May 28 call, showed what agentic commerce means at the merchant layer. The company talked about democratizing agentic commerce for the roughly 900,000 merchants on its Clover platform — and its example was pointedly small-scale: a single-location bootmaker in Wyoming should surface as an option when someone asks an AI assistant for cowboy boots. In other words, the merchant-side fight is over agent discoverability. When buying decisions route through assistants, being findable by an agent becomes the new being on page one — and merchant platforms want to sell that capability to the long tail.

Put the four calls side by side and you get a rough map of the emerging stack: networks positioning as rails (Visa, Mastercard), exchanges going multi-issuer on distribution (Coinbase), merchant platforms fighting for agent discoverability (Fiserv). Each layer is making a different bet, and each bet is falsifiable on future calls — which is what makes transcripts better source material than announcements. None of this resolution comes from press releases; it comes from reading what management actually committed to on the record, with analysts pushing back in the same room.

What to build with this

If the analysis above is the "what," here is the "how" — three concrete ways to wire this into your own tooling.

1. A payments-theme watchlist over the search endpoint. The chart in this post is one query per quarter. Automate it:

curl 'https://earningsapi.io/api/v1/search?q="agentic commerce"&type=transcripts&date_from=2026-04-01' -H "X-API-Key: $KEY"

Run it on a schedule for the phrases you care about, diff the results, and alert on new tickers entering the match set. Scope it to sector=Financials to track the consolidated stablecoin cohort specifically — per the Q2 2026 split, that filter now captures the bulk of the signal while cutting the noise.

2. Filter for analyst speech to see what's probed next. Every transcript in the corpus is split into speaker-tagged segments — 11.9M of them — with executive, analyst, and operator roles. Add speaker_type=analyst to the same search and you isolate what the sell side is probing rather than what management volunteers. Themes tend to show up in the Q&A before they get a slide in the prepared remarks; watching analyst questions inside Financials is a cheap early-warning system for whatever the next "agentic commerce" turns out to be. Phrase-tracking like this is one of the five text signals quants pull from earnings calls — the others compose with the same endpoint.

3. Drop the segments into an agent or RAG pipeline. Because segments come back as structured JSON — speaker, role, company, date, text — they slot directly into retrieval pipelines without a parsing layer. We've written up the pattern end to end: feeding earnings call transcripts into a RAG pipeline via MCP, or going a step further, building an earnings research agent with MCP and Claude that runs searches like the one above on its own. The entire API is also exposed as an MCP server, so the same watchlist works inside Claude, Cursor, or any MCP client with zero glue code.

The meta-lesson from this dataset: one full-text search endpoint, two phrases, and you can tell a spreading narrative apart from a consolidating one — and identify which four calls out of thousands are actually worth reading. That is the whole pitch for treating earnings calls as queryable data instead of PDFs.

Want to run these queries yourself? earningsapi.io gives you full-text search over 253,000+ speaker-tagged transcripts via REST API and MCP server — quarterly plans start at $145.