Hedgehog bets open-source Ethernet can run AI clouds without vendor lock-in (interview)

Hedgehog makes open-source software that lets AI clouds and enterprises run GPU networks the way hyperscalers do.

$11M pre-seed/seed raised, per Austin in the interview; a Series A round is planned.

Marc Austin, a Cisco networking veteran, founded Hedgehog in Seattle in 2022. The product exists, he says, because AI broke traditional networking: training and inference traffic melts networks designed for web apps, and the wait is rarely the hardware — it's weeks or months of hand-designing, cabling, tuning and validating fabrics across proprietary CLIs and locked-in vendor gear.

The 20-person team ships software whose repo it actually publishes — Austin says nearly every competitor markets "open networking" while shipping a proprietary controller. Customers do more than run the fabric: they carve up GPU capacity and resell it to their own customers like a cloud provider. He calls betting entirely on Ethernet the toughest decision of the past year, and says the industry's largest AI operators are now standardizing on the same approach.

AI clusters made networking the bottleneck: GPU clusters idle for months while scarce engineers hand-build fabrics across proprietary vendor gear.

The buyer shifted from network engineers to platform and DevOps teams, who want networking declared like the rest of their Kubernetes-style stack.

Openness is positioned as the differentiator: publishing the repo against competitors that ship proprietary controllers under an 'open networking' label.

The all-in Ethernet bet rides a standardization wave Austin says the largest AI operators are now following.

Pick the wave, not the surfboard: product decisions are recoverable, betting against a structural shift isn't — find the standard that's inevitable, align early, wait out the market.

As of 14 July 2026, Hedgehog is a 20-person Seattle company with $11 million in pre-seed/seed funding and a Series A planned, using AI heavily across engineering, testing and go-to-market to test every supported device and configuration in its lab. Its success condition, per Austin: networking gets boring again, and "network like a hyperscaler" describes every AI cloud, not just the giants running on Hedgehog.