Temporal Raised $550M Because Agents Keep Dying Halfway Through
Temporal raised a $550M Series E at a $12.55B valuation on September 14, led by Lightspeed and Wellington Management, with Growth Equity at Goldman Sachs Alternatives, Tiger Global, T. Rowe Price and SV Angel joining and a16z, Sequoia, Index, GIC, Sapphire and Amplify coming back in. That is about $1.35B raised in total, and it lands seven months after a $300M Series D in February. The announcement is at https://temporal.io/blog/temporal-raises-usd550m-series-e-at-usd12-55b-valuation-ai and after eighteen straight collection runs with no meaningful agent-infrastructure round, this is the one that broke the drought.
Temporal is not an AI company and that is the interesting part. It does durable execution: you write a workflow as ordinary code, and the runtime persists every step, so when the process crashes, the machine dies, the API times out or the whole region falls over, the workflow picks up exactly where it stopped instead of starting again from nothing. That was a payments and logistics problem for most of the company's life. It is now an agent problem, because an agent loop is a long-running, stateful, partially-failing process that calls unreliable external services, which is the exact shape Temporal was built for.
The numbers say the AI rotation is real rather than a slide in the deck. Annualized revenue up more than 200% year over year. Net dollar retention above 200% since February, which means existing customers are more than tripling their spend. 1.9 trillion billable actions in August alone, up more than 350%. Open-source installs past 43 million, up 134% since December. 4,300-plus paying customers, up 139%. OpenAI is a customer and its VP of Infrastructure is on record saying durable execution is "more than ever a core requirement for modern AI systems" β which is a fairly loud admission about what it takes to keep agents running in production.
Worth sitting with: the round that finally cleared the bar was not a model company, not a coding agent, not an eval startup. It was plumbing. Everyone shipping agents this year has independently rediscovered that the hard part is not the reasoning, it is the twelve-hour job that has to survive a deploy. [Cognition's $2B Series E](https://clauday.com/article/201fb8b0-063b-4eb9-be4c-3925ee84fd97) bought the agent; this one buys the thing underneath it that keeps the agent from losing its place. Expect the next few rounds in this category to look more like message queues and schedulers than like chatbots.
The uncomfortable read is what a 200% net dollar retention number implies about cost. Customers are not tripling spend because they love the product. They are tripling it because agent workloads generate an enormous number of small durable steps, and every one of them is billable.
← Back to all articles
Temporal is not an AI company and that is the interesting part. It does durable execution: you write a workflow as ordinary code, and the runtime persists every step, so when the process crashes, the machine dies, the API times out or the whole region falls over, the workflow picks up exactly where it stopped instead of starting again from nothing. That was a payments and logistics problem for most of the company's life. It is now an agent problem, because an agent loop is a long-running, stateful, partially-failing process that calls unreliable external services, which is the exact shape Temporal was built for.
The numbers say the AI rotation is real rather than a slide in the deck. Annualized revenue up more than 200% year over year. Net dollar retention above 200% since February, which means existing customers are more than tripling their spend. 1.9 trillion billable actions in August alone, up more than 350%. Open-source installs past 43 million, up 134% since December. 4,300-plus paying customers, up 139%. OpenAI is a customer and its VP of Infrastructure is on record saying durable execution is "more than ever a core requirement for modern AI systems" β which is a fairly loud admission about what it takes to keep agents running in production.
Worth sitting with: the round that finally cleared the bar was not a model company, not a coding agent, not an eval startup. It was plumbing. Everyone shipping agents this year has independently rediscovered that the hard part is not the reasoning, it is the twelve-hour job that has to survive a deploy. [Cognition's $2B Series E](https://clauday.com/article/201fb8b0-063b-4eb9-be4c-3925ee84fd97) bought the agent; this one buys the thing underneath it that keeps the agent from losing its place. Expect the next few rounds in this category to look more like message queues and schedulers than like chatbots.
The uncomfortable read is what a 200% net dollar retention number implies about cost. Customers are not tripling spend because they love the product. They are tripling it because agent workloads generate an enormous number of small durable steps, and every one of them is billable.
Comments