Derek Collison’s grandmother bought him a Commodore 64. He had been spending time with a friend’s TRS-80, trying to understand what the machine could do. His parents said they could not afford a computer. His grandmother decided he would have one anyway. She bought his sister one, too. Collison later recalled appropriating hers for parts. A family investment had acquired a small, unauthorized research department.
It is an appealing beginning for someone whose work usually lives out of sight. Today, Collison is the founder and CEO of Synadia and the creator of NATS, software that helps applications exchange messages. Before that came TIBCO, Google, VMware and Apcera. Much of his career concerns what happens between the things people actually notice.
A customer sees an application. An engineer sees the application, the services behind it, the data they need, and the increasingly awkward business of making all those pieces communicate. Collison has kept returning to that last problem. The machines have become more powerful. Their introductions remain surprisingly elaborate.
The plumbing acquires a passport
At TIBCO, Collison worked on enterprise messaging, including Rendezvous, and became senior vice president and chief architect. It was infrastructure for organizations that could put a price on an interruption. In one recollection, chief executive Vivek Ranadivé appeared with a suit because Collison needed to go immediately to a customer in New York. He took his computer and source code on a CD. The dress code was negotiable; the outage was not.
He spent 2003 to 2009 at Google. There he co-founded the AJAX APIs group, working on ways for developers to use Google’s services from their own applications. His work also included a content delivery network for popular JavaScript libraries. The setting had changed from enterprise messaging to the web, but the practical concern was familiar: make useful capabilities easier to reach.
At VMware, he designed and architected Cloud Foundry, an open platform for deploying applications. His 2012 presentation on its internals moved through controllers, provisioning agents, execution agents and messaging. A developer might experience deployment as one action. Underneath, several components had to agree about what that action meant. Somebody had to arrange the conversation.
A weekend with a long afterlife
NATS began inside that Cloud Foundry work in 2010. Collison needed messaging for the platform’s own coordination. He has described the first implementation as a weekend project: leave on Friday, plug it in on Monday. The ambition at the time was contained. Cloud Foundry needed a component that could handle its communication requirements.
Its guiding principle was equally compact: “Just do a couple of things very, very well.” That is an attractive instruction for software and a difficult one to enforce. Every additional capability arrives with an explanation for why it belongs. Before long, the small program has a reception desk and three committees.
The original NATS was written in Ruby. Collison did not initially set out to build a separate company around it. That came later, after another platform and another round of experience. The sequence matters. NATS had a job before it had a business plan, and users before Synadia existed to sell them services.
“Just do a couple of things very, very well.”
Derek Collison, on the principle behind NATS
A language chosen at ground level
In 2012, after leaving VMware, Collison founded Apcera. Its platform, eventually called Continuum, prompted another foundational choice: which language should carry the next system? He liked Ruby, but deployment dependencies and the demands of distributed infrastructure made him look elsewhere.
He had been experimenting with Go since 2011. His first program tested how its stacks and memory behaved. This was a fairly intimate first date with a programming language. Rather than starting with something photogenic, he wanted to know what happened to the furniture underneath.
Go’s compiled deployment model appealed to him, as did its build speed and performance. He moved the NATS server from Ruby to Go. In his 2014 account, keeping the architecture initially produced a jump from roughly 150,000 to 700,000 messages per second. Further optimization, he reported, reached six million on a MacBook Air.
The numbers are a record of his engineering journey, rather than a promise for somebody else’s workload. Their more revealing feature is the second improvement. Changing the language helped; learning how to use it changed the result again. A tool choice still left a great deal of work for the person holding it.
The question underneath the platform
Apcera also reflected a change in the question Collison wanted to answer. Running software efficiently was only part of an organization’s problem. How could people express what should be allowed, then establish that the system was actually enforcing it? Security teams, developers and compliance teams all had a stake in the answer.
He described this as a question of trust. An application platform had to accommodate organizational intent alongside the code. That made Apcera a continuation of his earlier work, with a wider set of concerns. Getting systems to cooperate involved people’s rules as well as machines’ protocols.
The move also showed a willingness to leave a project he had helped establish. Cloud Foundry was an achievement he could point to. Apcera gave him another attempt at the underlying design. For a founder with an architectural temperament, a successful diagram can still contain an irritating omission.
- TIBCOEnterprise messaging
- GoogleAJAX APIs
- VMwareCloud Foundry · NATS
- 2012 · ApceraApplication platforms and trust
- 2017 · SynadiaConnectivity across cloud and edge
The cloud gets a familiar nickname
Collison founded Synadia in 2017. His premise was that distributed systems would spread across regions, cloud providers and edge locations. The applications would still need services and data, but the assumptions about where everything lived would become harder to maintain.
He has a memorable phrase for the concentration of software inside large providers: “The cloud is the new mainframe.” The objection concerns dependence. Choosing a provider and a region can shape the rest of an application’s architecture, including the extra components needed to connect and secure it.
His preferred phrase for a different arrangement is “nomadic applications.” Software should be able to move without its developers rebuilding every connection around a new address. Synadia’s business combines that ambition with commercial products for operating, monitoring and supporting NATS-based systems.
In February 2024, Synadia announced a $25 million Series B led by Forgepoint Capital, taking its announced funding total to $51 million. The capital was intended to support NATS development and cloud and edge applications. The release named users including Rivian, Walmart and Replit. A program once made for one platform had found rather different places to work.
Synadia’s February 2024 Series B
Led by Forgepoint Capital · $51M total funding announced at that time
When the messages need to wait
Understanding the product helps explain Collison’s persistence. In core NATS, a publisher sends messages on a subject, and connected subscribers receive the messages they are interested in. The publisher does not need to know each recipient individually. That removes one kind of coordination burden.
JetStream adds persistence. It stores messages in streams and tracks consumers’ progress, allowing messages to be replayed and acknowledged. The participants can work at different times. A conversation gains a memory.
Core NATS delivers to subscribers connected when a message is published. It does not store messages for later replay.
That distinction matters more than an attractive speed figure. Different applications need different delivery behavior. The useful question is what happens when a recipient arrives late, restarts, or needs to read earlier events. Simplicity has to remain specific about what it promises.
The difficult business of keeping it open
By 2025, Collison also faced a different coordination problem: the relationship between Synadia and the open-source project. Following a public dispute, CNCF and Synadia announced an agreement on May 1. NATS would remain in CNCF, with development continuing under Apache 2.0 and Synadia’s continued participation.
Synadia agreed to assign its two NATS trademark registrations to the Linux Foundation. The project’s domain and repositories would remain held by CNCF. Those details put boundaries around a relationship that can look deceptively simple from outside: the creator, the company and the project each carry different responsibilities.
In October 2026, Collison’s public commentary was again focused on connectivity, data movement and manufacturing. The settings keep widening. His underlying interest is still recognizable in a factory’s systems, just as it was inside an application platform.
He has put his own attachment to the work plainly: “But I identify with builders.” The CEO title brings its meetings and negotiations. Beneath it sits someone who still wants to reason through how the parts fit together. There are more parts now. His grandmother’s investment has kept him occupied.