Hedgehog raises $11M to simplify AI data center networking

A Seattle startup is betting that the companies racing to build private AI infrastructure need a fundamentally different approach to networking — one that treats the data center fabric like cloud-native software rather than proprietary hardware.

Hedgehog has raised $11 million in seed funding to develop open-source software that automates the deployment and management of GPU cluster networks. The 20-person company, founded in 2022 by Cisco networking veteran Marc Austin, is now preparing to raise a Series A round.

Why it matters

As AI computing costs climb, enterprises face a strategic choice: continue paying hyperscale cloud bills or build their own GPU infrastructure. But standing up private AI data centers requires networking expertise most companies lack. Hedgehog's approach — treating network configuration as declarative code rather than manual CLI work — could lower the barrier for organizations that want to own their AI infrastructure without hiring armies of network engineers.

The GPU networking bottleneck

Austin describes the core problem as "time to GPU value." Companies spending millions on GPU clusters often wait weeks or months for network engineers to manually design, cable, and validate the fabric connecting those processors. That delay represents pure capital waste on idle hardware.

Traditional networking approaches buckle under AI workloads. Training and inference traffic patterns differ fundamentally from web application traffic, requiring networks purpose-built for the task. Yet most buyers aren't network specialists — they're platform and DevOps teams suddenly responsible for infrastructure they weren't trained to manage.

Hedgehog's software lets these teams declare network intent in Kubernetes-style configuration, then automatically provisions the fabric across commodity hardware. The company claims deployments that previously took months now complete in hours.

The open-source differentiation

Austin emphasizes that Hedgehog publishes its complete source code, distinguishing it from competitors who market "open networking" while shipping proprietary controllers. For customers, this means full code auditability, extensibility, and freedom from vendor lock-in — features Austin positions as enterprise requirements rather than hobbyist preferences.

The company made a significant technical bet last year by standardizing entirely on Ethernet rather than hedging across multiple fabric technologies. That decision now looks prescient as major AI operators converge on Ethernet-based architectures.

AI building AI infrastructure

Hedgehog uses AI extensively in its own engineering and testing workflows, allowing a small team to continuously validate every supported device and configuration. Austin notes this represents a meta-shift: AI workloads broke traditional networking, creating the market opportunity, while AI development tools enable his startup to compete with established vendors despite limited headcount.

The company's target customers increasingly include AI cloud providers themselves — startups that have taken delivery of thousands of GPUs and need to carve up that capacity for their own customers, mimicking hyperscaler multi-tenancy.

Austin's long-term vision: making networking "boring again" by reducing it to a few lines of declared intent that platform engineers never think twice about. GeekWire first reported these details in its Startup Spotlight series.