The Agent Economy in 2026: By the Numbers

The agent economy is real and it’s moving faster than most forecasts predicted. Here’s what the numbers actually say.

Market Size: What the Research Shows

Agentic AI market: Gartner projects the agentic AI market will exceed $47 billion by 2030, growing from near-zero in 2024. This represents one of the fastest growth curves Gartner has tracked for an enterprise technology category.

Broader AI agents market: Grand View Research estimates the AI agents market at $7.84 billion in 2025, growing at a 44.8% compound annual growth rate through 2030 — which would put it at ~$47 billion by 2030 (corroborating Gartner’s figures through a different methodology).

Agent commerce specifically: The subsegment of agent-to-service transactions — agents paying for things autonomously — is newer and harder to measure, but is growing faster than the broader category as the infrastructure matures.

Enterprise Deployment Scale

The largest deployed agent systems give a sense of scale:

Salesforce Agentforce: Reported over 1 billion autonomous actions in its first few months of broad availability. Salesforce has over 150,000 enterprise customers; even moderate adoption produces enormous transaction volume.

Microsoft Copilot: Hundreds of millions of monthly interactions across enterprise customers. GitHub Copilot alone has over 1.8 million paid subscribers (2024 figures, growing rapidly).

Anthropic’s enterprise deployments: Multiple Fortune 500 companies with thousands of concurrent Claude agents. The company cited substantial year-over-year revenue growth driven primarily by enterprise agent deployments.

OpenAI’s operator ecosystem: Thousands of companies have built agents on OpenAI’s API. The number of active agent deployments using OpenAI’s platform is in the hundreds of thousands.

What Agents Are Spending

Agent expenditures in 2026 fall into several categories:

LLM Inference Costs

The largest current category. A mid-scale enterprise might run 1-10 million LLM calls per day across its agent infrastructure. At average costs of $0.01-0.05 per call, this is $10,000-500,000/month per company at scale.

OpenAI reported annualized revenue of $4 billion in early 2025 and has grown substantially since. Most of that revenue comes from API usage — enterprises running agents at scale.

External Data and Research APIs

Agents that do research, verification, or data enrichment pay for data access. Market data APIs (Bloomberg, Refinitiv), web search APIs (Perplexity, Tavily), enrichment APIs (Clearbit, Clay) — these are purchased by agents in the same way they’ve been purchased by software systems for years, just at higher velocity.

SaaS Tool Access

Agents increasingly consume SaaS tools programmatically. Salesforce, HubSpot, Notion, Airtable — all have APIs that agents can use, and API usage contributes to SaaS vendor revenues.

Direct Purchases

Still early, but growing: agents purchasing compute credits, buying research reports, booking services, procuring software licenses. This is the “agent commerce” segment specifically — agents transacting in the way humans have historically transacted.

The Infrastructure Layer Is Being Built Now

The agent economy requires infrastructure that didn’t exist in 2023:

Payment protocols: Five major agent payment protocols emerged between 2024-2026: x402 (HTTP-native), Stripe ACP/MPP, Google AP2/UCP, Visa TAP, and IOU token systems. The rapid emergence of multiple competing standards signals that the market is real — the protocol question is which standards survive.

Agent identity: KYA (Know Your Agent) standards are being developed by Visa, Mastercard, and emerging players like ATXP. Financial institutions need to know whether they’re transacting with an authorized agent before processing payments.

Compliance infrastructure: Financial regulators in the US, EU, and UK are developing frameworks for agent financial activity. The EU AI Act has specific provisions for AI systems that take consequential actions; agent payments almost certainly qualify.

Where This Goes

Three trends will shape the agent economy over the next 24-36 months:

Protocol consolidation: Five payment protocols won’t all survive. Likely outcome: 2-3 dominant standards, with compatibility layers handling cross-protocol routing. The analogy is credit card networks — multiple coexist, but each merchant doesn’t need separate infrastructure for each.

Identity maturation: KYA will become a baseline requirement for agents transacting above a threshold value. This mirrors the KYC evolution in consumer finance — friction at first, then table stakes.

Autonomous procurement at scale: The most significant economic shift is enterprises delegating purchasing authority to agents. Routine procurement (software licenses, cloud compute, data access, vendor services) will increasingly be handled by agents within authorized parameters. This changes how B2B commerce works structurally.

The Investment Signal

Venture investment in agent infrastructure totaled over $2 billion in 2024-2025 (Anthropic, OpenAI, Cohere, and the layer of infrastructure companies building on top of them). This isn’t speculative money — these are infrastructure bets on a market that’s clearly forming.

The agent economy in 2026 is early but established. The interesting question isn’t whether it’s real — it clearly is — but which infrastructure layer will be most valuable as it scales.

Our bet: the payment and identity layer. Whoever controls how agents transact is in the position that Stripe occupies for human commerce. That’s a significant position to be building for.

ATXP is the payment and identity layer for the agent economy. For the protocol landscape that underpins it: agent payment protocols compared.