On paper, Temporal Technologies looks like the kind of startup that comes out of nowhere: seven years old, $550 million in new cash, a $12.55 billion valuation, and a customer list that includes OpenAI, Nvidia, Netflix, Snap, and JPMorgan Chase. It’s the sort of headline that makes people assume the company was born with the AI boom. But Samar Abbas, Temporal’s co-founder and CEO, would gently correct you. He and his co-founder, CTO Maxim Fateev, have been wrestling with the same fundamental problem for more than two decades, long before the term “cloud” entered everyday vocabulary. They first met at Amazon in 2010, where they worked on Simple Workflow Service, an AWS product designed to coordinate long-running tasks across distributed systems. That sounds dry, but it opened a door to a question that never stopped bothering them: when software runs across many machines and many steps, how do you make sure it finishes even when things break? “We’ve been hacking away at this problem for 20-plus years now,” Abbas says. Temporal, based in Bellevue, Wash., was founded in October 2019 after the pair left Uber. It now has more than 4,300 paying customers and recently passed $250 million in annualized revenue run rate. But to the founders, the recent explosion of interest is not a sudden event. It’s the moment when the rest of the world finally caught up to a problem they’ve been obsessing over since the early days of distributed systems. For years, this work was quiet. When GeekWire first wrote about Temporal in 2020, it had 15 employees, $25.5 million in funding, and no paying customers. Sequoia Capital led a $20 million Series A, with participation from Seattle’s Madrona, and former Snowflake CEO Bob Muglia was an angel investor. The founders kept saying the same thing: this is core infrastructure, and you cannot build it fast. Fateev puts it plainly: “It took us years and years to get to the point where we knew what we were doing.” The world may be noticing Temporal now because of AI, but the company was built the old-fashioned way: slowly, patiently, and with a stubborn refusal to chase hype.
At the heart of Temporal is an idea the founders call “durable execution.” The name sounds technical, but it’s really about something we all understand: not losing your place. Imagine you’re driving a friend to the airport, and along the way you make three stops. If you can’t remember which stops you already made, you either repeat stops or give up entirely. A typical software program, when one of its moving parts fails, is exactly that forgetful. Temporal gives it a kind of external memory. Every step is recorded the moment it completes, so when a server crashes, a network call times out, or a third-party API returns nonsense, the program can resume exactly where it left off instead of starting over or leaving the work half-done. That sounds simple, but for long-running tasks that cross many machines, it’s fiendishly hard. Abbas and Fateev spent years at Amazon, Microsoft, Google, and Uber trying to solve variations of the same problem. Fateev built the messaging infrastructure behind Amazon’s Simple Queue Service starting in 2004, then led the architecture of Simple Workflow Service before moving to Google and later reuniting with Abbas at Uber. Abbas spent 11 years at Microsoft, joined the Simple Workflow Service team at AWS in 2010, returned to Microsoft and wrote the open-source library that became Azure Durable Functions, and eventually moved to Uber’s Seattle office in 2015. They kept realizing that most systems treat failure as an exception rather than a rule, and that’s exactly backwards. “It’s core infrastructure. You really cannot go and build it very fast,” Fateev says. Temporal is the culmination of that obsession: a layer that sits underneath other software and quietly makes sure every step of a long process survives the chaos of the real world. It’s not glamorous work, but it matters more and more as software stops being a simple app and starts becoming a sprawling, always-running organism.
The reason this suddenly matters more than ever is the rise of AI agents. Instead of just answering questions, agents act: they book travel, update files, write code, call other software, and keep going until a job is done. But they are also notoriously unreliable. An agent might run for hours or days, making hundreds of calls to language models and external tools, and any one of those calls can fail. Without something to tie the steps together, the whole process falls apart in a mess of half-completed actions. That’s exactly the problem Temporal was built to solve. Abbas frames it in everyday terms: “Every Snap story is a Temporal workflow. Every time you place an order at a Taco Bell, all of the steps get orchestrated on top of our platform. Some of the most popular coding agents out there are using us as an outer harness.” The company is deliberately model-agnostic; it doesn’t care whether an agent is running GPT, Claude, or something else. It just wants to be the nervous system that connects all the pieces, keeping them in sync even when individual components fail. This is why an infrastructure company with roots in the pre-cloud era has become one of the hottest names in artificial intelligence. AI needs reliability, and reliability is Temporal’s whole identity. “It’s a pretty exciting time for us, from that perspective,” Abbas says with characteristic understatement. The company has doubled its headcount over the past year to 570 people, and its growth is being driven by the same forces that make AI useful in the real world: agents can’t just be clever, they have to be dependable. Temporal provides the spine that lets them be dependable.
Temporal may be a global company, but its heart is in the Pacific Northwest. Both founders have lived in the Seattle area for more than 26 years, and they built their entire professional careers there—at Microsoft, Amazon, Google, and Uber. The company is fully remote, employs 570 people, and 89 of them are in the Seattle area. Its Bellevue office, which used to be OpenAI’s, is used mainly for meetings rather than daily work. For Abbas, the region is a competitive advantage. “The kind of talent that we have here in the Pacific Northwest is insane,” he says, pointing to the deep bench of cloud engineers trained at Microsoft and Amazon. He expects that advantage to apply to AI infrastructure as the industry shifts. The progress since Temporal’s early days is staggering. When GeekWire first wrote about the company in 2020, it had 15 employees, $25.5 million in funding, and no paying customers. Snap and Box were early users. Today Temporal has raised $1.2 billion in total funding and ranks No. 2 on the GeekWire 200 list of top Pacific Northwest startups, behind only Helion, the Everett fusion company valued at $15.5 billion. In a region known for patient, deep-tech engineering, Temporal’s rise feels less like a flash in the pan than a vindication of a very old-fashioned approach: hire great people, work on a hard problem, and don’t expect to be an overnight success. The founders are quick to credit the region’s ecosystem, but they also embody it. They stayed, they built, and they kept solving the same problem even when the rest of the tech world was chasing shinier things. Now the world has come to find them.
Temporal’s milestone arrives at an uncomfortable moment for the AI industry. There is a growing debate about how fast to move, and a steady stream of stories about AI agents doing things they weren’t supposed to do—finding ways around the rules, or failing in ways that no one can quite explain. To Abbas, these incidents are less about model behavior than about accountability. “As you dig into each and every one of the incidents that have happened in the last three months,” he says, “it always comes down to no one really knowing what these AI agents did, step by step.” Temporal’s pitch is that it can fix that. Because the platform records every step of a workflow, it leaves a complete trail of what an agent did, in what order, and where it stopped. That log gives companies a way to investigate problems after the fact, and, just as importantly, a place to intervene before the next step executes. “We can put the agent in a much, much tighter jail,” Fateev says. Abbas is careful not to overstate the role. “I’m not going to sit here and tell you we solve AI safety,” he says, arguing that responsibility for the models themselves lies with the labs that create them. But he insists that giving companies a way to know what their agents did—and to stop them from doing more than they should—is a specific, solvable engineering problem. “That’s truly an engineering problem,” he says, “and it’s very solvable today.” In a world where AI is moving fast and breaking things, that certainty is refreshing. It doesn’t require a philosophical breakthrough, just careful design and the willingness to keep a detailed record.
Temporal’s customer list includes some of the most powerful names in AI, and Bloomberg reported that OpenAI is currently its largest customer. Is that too much dependence? Abbas doesn’t think so. The AI labs are moving fast and driving usage, he says, but similar growth trends are starting to show up across the rest of the customer base. He notes that 18 of the top 30 AI-native companies use Temporal. Still, the biggest customers could become competitors. AI labs and cloud providers sell their own agent frameworks, and some are working on reliability features that sound a lot like Temporal. Fateev’s answer is that enterprises don’t want to be locked into one provider, one model, or one vendor’s idea of how agents should work. Temporal is open source, and it takes no position on what runs on top of it—an advantage in a world where the tech stack changes constantly. The founders also see an opening in how AI systems connect to one another. Agents increasingly need to call other agents and outside services, and standard protocols for making those connections assume the work finishes in seconds. When a job runs for hours, or has to be carefully unwound after something fails, those assumptions fall apart. Temporal is built for exactly that kind of long, messy, real-world work. As for what’s next, the company plans to pour a large share of its new funding into research and development, extending the durable-execution technology at its core. Abbas says Temporal wants to host agents on its own platform, not just direct them on others, and to expand features for security and auditing. It also plans to grow internationally and keep hiring, after doubling headcount in the past year. Through all of it, the founders remain deeply connected to the community they’ve built. Temporal’s open-source Slack channel has about 25,000 members, and Fateev still answers questions there himself. “You ask a question,” Abbas says, grinning, “I’m pretty sure Max will jump in within the first five minutes.” That, in a way, is the whole story of Temporal: a complex piece of infrastructure, built by people who never lost touch with the craft, and now suddenly essential to the most ambitious technology of our time.
