Technology

Restate Raises $20M to Scale Durable Execution for AI Agents

Martin HollowayPublished 3d ago3 min readBased on 4 sources
Reading level
Restate Raises $20M to Scale Durable Execution for AI Agents
source:restate.dev

Restate has raised a $20 million Series A led by Singular, with participation from Redpoint Ventures and Capital One Ventures. The Berlin company announced the round on September 30, 2026, to fund durable workflow infrastructure for backend systems and AI agents. TechCrunch

Restate was co-founded in 2022 by Stephen Ewen. The company is based in Berlin and plans to use the fresh capital to hire a go-to-market team, hire more engineers, and expand its office in the San Francisco Bay Area.

The round follows commercial traction. Restate closed multiple six- and seven-figure customer contracts in the months before September 30, 2026. Its customers include vibe coding platform Replit.

The technical bet is architectural. Restate developed its own storage, replication, and redundancy layers instead of building its durable execution engine on top of an external database. The stated goal for the Series A is to make Durable Execution fast, efficient, and flexible enough to become a widespread building block for backends.

For AI workloads, Restate durably persists every LLM call, tool execution, and routing decision in AI agents. Restate Docs The system enables building AI agents that survive crashes and recover automatically.

That persistence model addresses versioning as well. Restate supports safely updating long-running AI agents in production with immutable deployments and pinned executions. Restate

Deployment flexibility is part of the pitch. Restate announced Restate BYOC on July 7, 2026. The company described Restate BYOC as fully managed Restate deployed into a customer's own cloud account and VPC.

The founding team comes from stream processing. Ewen was a co-creator of Apache Flink and later became CTO of Data Artisans. Alibaba acquired Data Artisans in 2019 and rebranded it as Ververica. Ewen served as CTO of Ververica for three years after the acquisition. Restate's other two co-founders are Igal Shilman and Till Rohrmann, former colleagues of Ewen from Data Artisans and Ververica.

One customer detail will draw attention from infrastructure engineers. Replit switched its AI coding agent from Temporal to Restate. Temporal, founded in 2019, announced a $550 million Series E at a $12.55 billion valuation in September 2026.

The broader context here is straightforward and, in this author's view, worth taking seriously. Workflows were already hard to operate reliably. Agents make the problem harder. They run longer, they call non-deterministic models and external tools, and they fail in the middle. Restarting from scratch is expensive and, for interactive coding or multi-step automation, it breaks the user experience.

Looking at what this means for practitioners, the interesting claim is not durability alone. It is durability without outsourcing durability to Postgres or Cassandra, and without forcing developers to re-architect around an external store for journals and timers. If Restate can deliver that with low operational overhead, it simplifies the stack. BYOC points in the same direction. Teams that cannot send execution state to a multitenant cloud still get a managed control plane inside their own VPC.

Worth flagging is the versioning problem. Long-lived executions outlive the code that started them. Pinned executions and immutable deployments are a pragmatic answer. They let an in-flight agent finish on the logic it started with while new runs pick up the fix. Anyone who has managed stateful workers through a hurried production patch will recognize the value.

In my view, the optimism around agents is justified only if the plumbing catches up. Durable execution does not make models smarter. It makes systems built on fallible models usable. That is less exciting than a demo, and more durable over the long arc.