Temporal's second doubling in a year prices durable execution as agent plumbing

Temporal Technologies is in talks to raise about $500 million at a pre-money valuation of $12 billion or more, Bloomberg reported on Tuesday, citing people familiar with the talks ^[1][2]. If it closes at the reported floor, it would be the second time this year the company's valuation has doubled ^[6], which means the market is now pricing durable execution as a base layer for AI agents rather than a developer convenience.

The ladder is steep. Temporal announced a $146 million Series C at a $1.72 billion valuation in March 2025, led by Tiger Global ^[7], then a $300 million Series D at $5 billion in February, led by Andreessen Horowitz with Lightspeed Venture Partners and Sapphire Ventures participating ^[5]. At $12 billion the company would be worth roughly seven times its March 2025 mark ^[9] and about 2.4 times its February one ^[10]. Adding the new money would put the post-money figure near $12.5 billion ^[11], and a completed round would push total announced funding past $1 billion ^[8].

What is underneath that. Temporal said revenue grew more than 380 percent year on year and that annual recurring revenue passed $100 million as of February ^[12][13]. Those are the company's own numbers, and its usage figures - more than 2,500 cloud customers, monthly software installations past 20 million - mix commercial revenue with free open-source downloads ^[17][14]. Temporal has not publicly attached a current revenue figure to the proposed $12 billion ^[15]. Against the February ARR, the reported valuation is 120 times revenue ^[16]. The real multiple is lower, but nobody outside the round knows by how much.

The product argument is more durable than the valuation. Temporal sells an orchestration platform that sits underneath applications and keeps multi-step processes from failing partway through ^[21]. Developers write a workflow in an ordinary programming language, the infrastructure records each step, and after a crash the workflow resumes from its last saved point instead of starting over ^[23]. Temporal says that removes the need to hand-build timers, queues and retry logic ^[24]. The idea predates the current AI wave ^[31]: founders Samar Abbas and Maxim Fateev worked on Amazon's Simple Workflow Service, then built the open-source Cadence engine at Uber before starting Temporal in 2019 ^[27], and Abbas had led work on Microsoft's Durable Task Framework ^[28].

Agents are the demand shock. They run long chains of tasks and can fail silently, and Temporal says customers use it to keep agents alive over hours or days, trace where workflows break, and control compute costs ^[25]. The company does not build or train models ^[26]. Its biggest customer is OpenAI, according to the people Bloomberg spoke to ^[18], with Replit and Abridge also named ^[19]; Temporal has separately listed Nordstrom, Block, ADP and The Washington Post ^[20]. That is a good logo set and a concentration question at the same time.

For buyers, the lock-in shape is unusual. The core is released under an open-source licence and the commercial product is a managed service, Temporal Cloud ^[22]. The exit from the managed service is self-hosting the same engine, not switching vendors, because the workflow code is written against Temporal's model ^[23][22]. Meanwhile the primitive is being bundled by platforms teams already pay for: Cloudflare made Workflows generally available in April 2025 ^[29], and Vercel shipped its own durable workflows ^[30].

Watch three things: whether the round closes at the reported terms, which Bloomberg says could change ^[3]; whether Temporal publishes a current ARR figure to sit beside the $12 billion ^[15]; and whether new projects start defaulting to the bundled durable workflow products rather than a standalone engine ^[29][30].