Sapiom Raises $35 Million Series A to Help AI Agents Reach Production

6 min read
Co-founder of JustAINews
Share
Key Points
  • Sapiom raised a $35 million Series A led by Dragonfly to build infrastructure that helps AI agents run reliably in production.
  • The round includes Accel, Gradient, Coinbase Ventures, Operator Collective, Formus Capital, VanEck Ventures, Okta Ventures, Menlo Ventures, Anthropic, and Array Ventures.
  • Sapiom is launching Router, Agent Studio, and Runtime to cut agent costs and give engineering teams full visibility over execution.
Sapiom removes the infrastructure barriers to shipping and scaling agents, starting with the cost of execution
Credits: Sapiom

Sapiom, a San Francisco startup founded in 2025 by Ilan Zerbib, builds infrastructure that helps engineering teams get AI agents into production. The problem it targets is specific: most AI agents work well as demos but fail when companies try to run them continuously, at scale, and within a budget. In six months of operation, the company says it has processed over 270 million transactions and now powers more than 100,000 agent runs per day.

That gap between demo and production is getting more attention as spending on AI agents accelerates. According to Gartner, more than 40% of agentic AI projects will be canceled by the end of 2027, with escalating costs cited as a primary reason. Building a capable agent has become relatively straightforward. Running one in production, where it must take real actions, handle unpredictable workloads, and stay within budget, is where most teams still struggle. The global AI agents market was valued at $7.6 billion in 2025 and is projected to grow to nearly $183 billion by 2033. As the market grows, so does the need for infrastructure that can keep agents reliable and affordable.

Sapiom has raised $35 million in a Series A led by Dragonfly, a crypto-focused investment fund. Accel, Gradient, Coinbase Ventures, Operator Collective, Formus Capital, and VanEck Ventures also participated. Existing investors Okta Ventures, Menlo Ventures, Anthropic, and Array Ventures returned for this round. The Series A brings the company's total funding to $50 million, following a $15 million seed round closed just six months earlier.

Why running AI agents in production is still expensive

Modern AI models can reason, plan, and carry out complex tasks. The difficulty is not building an agent that works once. It is keeping that agent running reliably, day after day, without costs spiraling out of control. A single agent run can trigger dozens of separate calls to language models, web services, and external tools, each with its own cost. Most teams pick one model and one set of infrastructure during development and rarely change those choices afterward. As prices shift and better models become available, that original setup becomes increasingly wasteful.

Sapiom sits at the point where agents execute their work. For each action an agent takes, the platform evaluates the available options and selects the most efficient path based on cost, quality, speed, reliability, and company policy. One customer reduced inference costs by 75% after moving its agents to Sapiom, according to the company.

"Agents are becoming employees with no manager and no budget, and increasingly, the CTO is the one acting as CFO, allocating real money with no visibility into where it goes. That's not a governance problem you solve with another dashboard. It's an infrastructure problem, and it needs to be solved at the point where agents act and money moves. Sapiom is the only team that we have seen execute this at scale."

Haseeb Qureshi, General Partner at Dragonfly

How Sapiom plans to use the funding

Three new products are launching alongside the round. Sapiom Router automatically matches each model call to a suitable, cost-effective model rather than defaulting to the most expensive one. Sapiom Agent Studio gives engineers a local environment to build, test, inspect, and deploy agents within their existing codebase. Sapiom Runtime provides managed production infrastructure with access controls, routing, automatic recovery, and step-level visibility.

Sapiom is hiring across engineering, machine learning, go-to-market, and developer relations, with a particular focus on routing policy, inference systems, and post-training. The company operates its own inference capacity and processes tens of millions of real agent executions.

Cost is the immediate priority, but not the full picture. The company says its longer-term goal is to remove every infrastructure barrier that prevents agents from reaching production at scale.

What Sapiom does and how its platform works

Ilan Zerbib founded Sapiom in San Francisco in 2025. The company provides a layer of infrastructure between an AI agent and the services that agent needs to function: language models, compute, search tools, email, databases, and more. Instead of hardcoding those connections, Sapiom manages them dynamically each time the agent runs.

In practice, this works through a single integration point. Every time an agent needs to take an action, Sapiom handles the routing, usage tracking, and permissions. Budgets and access rules are enforced before anything executes, and every action is recorded in a full audit trail. Teams can see exactly what their agents are doing and how much each step costs. Engineers can build with Sapiom from tools they already use, including Claude Code, Cursor, and Codex. In June 2026, the company also acquired Fewsats to bring machine-native payment capabilities into the agent execution layer.

"Teams can build remarkable agent demos in days, but getting those agents into production economically, reliably, and with control is still brutally hard. Every team is being forced to recreate the same infrastructure before its agents can perform real work. Sapiom exists to remove those barriers. We are starting with cost because it is where the economics break first, but our ambition is much larger: to remove whatever stands between builders and the next trillion agents."

Ilan Zerbib, Founder and CEO of Sapiom

The investors behind Sapiom's $50 million in funding

Sapiom has raised $50 million across two rounds in quick succession. The $15 million seed round, led by Accel, closed in February 2026, five months after the company was founded. The $35 million Series A, led by Dragonfly, followed six months later. Dragonfly is a crypto-focused investment fund with offices in New York and Singapore. Its Managing Partner, Haseeb Qureshi, will join Sapiom's board of directors as part of the deal.

The investor list is broad. Accel, a Silicon Valley venture firm, participated in both rounds. Anthropic, the AI safety company behind the Claude family of models, is also among the backers, along with Okta Ventures, Menlo Ventures, and Array Ventures. Newer participants in the Series A include Gradient, Coinbase Ventures, Operator Collective, Formus Capital, and VanEck Ventures.

The company raised $50 million in under a year from founding. That pace points to real investor appetite for the AI agent infrastructure layer, especially as more enterprises discover how difficult and expensive it is to move agents from prototype to production.

Funding details

Subscribe to JustAINews!

Get the industry's biggest AI news straight to your inbox.
Subscribe

Related posts

© Copyright 2025 - Just AI News - All Rights Reserved
linkedin facebook pinterest youtube rss twitter instagram facebook-blank rss-blank linkedin-blank pinterest youtube twitter instagram