The pitch sounds like something from a sci-fi screenplay: autonomous software programs roaming the internet, buying access to servers, APIs, and cloud compute without human supervision. Yet Sapiom, a San Francisco startup building precisely that infrastructure, managed to convince Accel and a roster of heavyweight investors to back the vision—to the tune of $15.75 million.
The seed round, announced February 6, 2026, drew participation from Gradient Ventures, Menlo Ventures, Anthropic, Coinbase Ventures, Array Ventures, Okta Ventures, Formus Capital, and Operator Collective. Angels from Shopify, OpenAI, Vercel, GitHub, Circle, and Mercury joined as well. It's the kind of cap table that signals either prescient positioning or aggressive FOMO—possibly both.
"The next trillion buyers won't be human," Sapiom proclaims on its site. Hyperbolic, perhaps. But the company is wagering that as AI agents grow more capable and autonomous, they'll need a way to pay for things. Not occasionally. Constantly.
The Problem Nobody Knew They Had
Here's what Sapiom is actually building: a financial layer that lets AI agents access paid services without a human reaching for a corporate card every few minutes. Think of it as Stripe or Plaid, but for software that makes its own purchasing decisions.
The platform handles what Sapiom calls "Know Your Agent"—essentially, identity verification for bots. It manages programmable wallets, spend controls, risk assessment, billing, and settlement. An agent operating within the system gets access to over 400 AI models, web scraping tools, browser automation, compute sandboxes, image and audio generation, databases, storage, messaging, even domain management. All through a single integration.
The controls are granular, almost paranoid. Per-agent budgets. Per-tool budgets. Spend ceilings per run or within specific time windows. Automatic session termination when limits hit. Real-time alerts if something goes sideways.
Sapiom supports emerging technical standards that most people have never heard of: x402, which repurposes the dusty HTTP 402 "Payment Required" status code for agentic micropayments, and the Agent Payments Protocol (AP2) for authorization and mandates. Whether these become industrywide rails or footnotes in someone's GitHub archive remains to be seen.
A Market Taking Shape, Maybe
The timing of Sapiom's raise reflects a broader—if speculative—shift. Mastercard launched something called "Agent Pay for Machines" last June. Adyen rolled out "Adyen Agentic," billing it as a "universal translator" for machine-driven commerce. The IMF weighed in with a note flagging "Know Your Agent" concepts and evolving settlement infrastructure. Even J.P. Morgan's 2026 payments outlook name-checked agentic AI as a new driver of industry evolution.
Whether this represents genuine market signal or collective hallucination is difficult to say. The infrastructure exists before the use cases have fully materialized—a familiar dynamic in Silicon Valley, though not always a successful one.
Competitors are circling the same territory. Natural pulled in a $30 million Series A in July. Orthogonal secured $4.3 million in June. AIsa raised $6.5 million, also in July. All chasing variations on the same theme: payments infrastructure for software that shops.
Who's Behind It
Sapiom's CEO, Ilan Zerbib, spent roughly five years as Director of Engineering, Payments at Shopify—a credential that carries weight in fintech circles. The investor lineup is strategically coherent, if a bit on-the-nose. Anthropic builds the AI models that power many of these agents. Okta Ventures brings identity expertise and enterprise go-to-market muscle. Gradient focuses almost exclusively on AI infrastructure plays.
Following the seed close, Sapiom posted openings for a Staff Engineer, Developer Relations Lead, and Growth Marketing Manager on Accel's job board between April and June. The company describes itself as a "small team" and dangles "founder-level equity" for early hires willing to help build what it calls the "definitive economic substrate for machines."
That's a lot of swagger for a company that hasn't disclosed customer traction, enterprise logos, or much detail about how it plans to spend the $15.75 million beyond the standard "team expansion and platform development."
The bet, ultimately, is that AI agents will become prolific enough—and autonomous enough—that the infrastructure to facilitate their transactions becomes critical. If that world arrives, Sapiom wants to be the Visa or Mastercard of the machine economy.
If it doesn't? Well, at least the cap table will make for good networking.
