Spotlight Spotlight: Hedgehog is betting that open source networking will power the next generation of cloud AI

As AI workloads drive skyrocketing cloud bills, many companies are weighing whether to move computing from public clouds to their own data centers. But building and deploying an AI infrastructure is much more complex than just buying servers — communication and connectivity are already one of the biggest technical hurdles.

That’s the opportunity Seattle Hedgehog startups are chasing.

Founded in 2022 by CEO Marc Austin, a Cisco networking veteran, Hedgehog develops open source software designed to enable private AI data centers to function as hyperscale clouds. Raised $11 million in seed funding, with plans to raise Series A funding.

We caught up with Austin for the return of GeekWire’s Startup Spotlight to learn more about the 20-person company, the AI ​​networking boom and what surprised him most about building a startup in one of the fastest growing tech markets.

In 50 words or less, give us your elevator pitch?

Hedgehog is open source software that makes AI communication easy. AI clouds and enterprises are using it to run GPU networks the way hyperscalers do – deployed in hours instead of months, deployed by DevOps teams instead of armies of network engineers, on open hardware without vendor lock-in.

What problem do you like to solve?

GPU value time. A GPU cluster is the most expensive asset most companies will ever purchase, and every day it sits idle waiting on the network is a waste of money. That wait is rarely hardware – it’s fabric: weeks or months of rare network engineers manually designing, cabling, tuning, and verifying all the proprietary CLIs and vendor-locked gear.

Meanwhile the people who are told to “own the network” are often not network engineers at all — they are platform and DevOps teams. We’re worried about the collapse of that timeline: declare your network as a target in Kubernetes and go from GPUs attacked to thinking in hours instead of months – on open hardware, no lock-in, no room full of experts. Cloud-grade networking without hyperscaler headcount.

What surprised you after talking to customers?

It is not uncommon for a consumer to be a network engineer. Platform and DevOps teams, often in AI clouds that have just taken thousands of GPUs are told that “you own the network now.” They don’t want to learn BGP; they want a network that behaves like the rest of their cloud-native stack. Another surprise: they don’t just want to use the network, they want to sell by recording the volume of their customers, as the cloud provider does.

How has AI changed the way you build your company?

It is doubled.

Our product exists because AI has broken traditional networks. Training and inference networks are fuzzy networks designed for web applications.

And AI changed the way we build: we use it extensively across engineering, testing, and go-to-market, allowing a small team to continuously test all supported devices and configurations in our lab and fleet with hyperscaler-grade robustness. AI suggested what a startup-sized team could do.

What else do people misunderstand about your startup?

That “open source” means hobbyist. The opposite is true: openness is a feature of business. Our customers can test every line of code that runs their fabric, extend it, and become unstuck. Almost every competitor has marketed “open communication” while deploying a proprietary controller. Hedgehog is the only one that publishes the repo.

What was the hardest decision you made in the last year?

Bet on Ethernet entirely. We decided that open, standards-based Ethernet would win the AI ​​network and put everything behind it. Watching the biggest AI operators in the industry now standardize on that same path makes us feel good about the phone – but it was hard to say no.

What is one piece of advice you would give other entrepreneurs?

Choose a wave, not just a surfboard.

Product decisions are reversible; betting against structural industry change is not. Find a level, structure, or inevitable buyer behavior, align everything to it early, and be patient while the market hits your bet.

We’ll know our company has done it when…

Networking is boring too. When a platform engineer stands up the GPU cloud hires a lot of people and the network is just a few lines of intent that no one thinks twice about. When a “hyperscaler-like network” defines all AI clouds, not just the giants running on Hedgehog, we’ll have it!