A networking startup called Eridu just pulled off one of the biggest Series A rounds this year, raising $200 million to build infrastructure for AI workloads. Co-founded by Drew Perkins, a networking pioneer who's been building internet tech since the early days, the company emerged from stealth with backing from legendary investor John Doerr at Kleiner Perkins. The hefty round signals serious belief that AI's explosive growth needs purpose-built networking hardware, not retrofitted data center gear.
Eridu just raised $200 million in Series A funding to tackle one of AI's most overlooked bottlenecks - the networking layer. The startup emerged from stealth mode with backing from Kleiner Perkins, where legendary investor John Doerr personally championed the deal. It's a massive vote of confidence for a company that's betting the AI boom needs fundamentally new networking architecture, not just faster versions of existing tech.
The round's size is striking. Most Series A deals hover around $15-30 million, making Eridu's $200 million haul more typical of late-stage growth rounds. That premium reflects both the capital intensity of hardware development and the pedigree of co-founder Drew Perkins, who's been inventing networking technology since the dawn of the commercial internet. Unlike the parade of twenty-something founders pitching AI wrappers, Perkins brings the kind of deep technical credibility that makes investors write bigger checks.
Perkins' track record matters here. He's not building a vibe-coded product - he's applying decades of experience to a real infrastructure problem. As AI models grow larger and training clusters expand to thousands of GPUs, the networks connecting them have become critical bottlenecks. Current data center switches and routers weren't designed for the massive all-to-all communication patterns that AI training demands. Meta, Google, and Microsoft have all built custom networking solutions internally, but the broader market still relies on retrofitted enterprise gear.
That's where Eridu sees its opening. The startup is developing networking hardware and software specifically optimized for AI workloads, though it's kept technical details under wraps during its stealth phase. The company's emergence comes as the AI infrastructure market heats up, with Nvidia dominating GPUs, startups like Cerebras and Groq building specialized chips, and cloud providers racing to offer the fastest training environments.
Doerr's involvement adds serious weight to the launch. The Kleiner Perkins chairman has backed some of tech's biggest winners, from Amazon and Google to more recent bets on AI infrastructure. His decision to lead Eridu's round suggests he sees networking as the next critical layer in the AI stack - one that could generate returns comparable to the chip and cloud sectors.
The timing looks right. Enterprise spending on AI infrastructure is exploding as companies rush to deploy large language models and train custom AI systems. According to recent industry estimates, the AI hardware market could hit $150 billion by 2027, with networking equipment capturing an increasing share as cluster sizes grow. Every percentage point improvement in network efficiency translates to millions saved in training costs and faster time-to-deployment for AI products.
Eridu faces stiff competition. Established networking giants like Cisco and Arista already serve major AI labs, while startups like Apstra and Volta Labs are also targeting the AI networking opportunity. But Perkins' experience building foundational internet protocols could give Eridu an edge in designing systems that scale to the massive clusters coming over the next few years. The company will need to move fast - its $200 million runway buys time to ship products, but hardware startups burn cash quickly on R&D and manufacturing.
The stealth emergence follows a pattern we've seen with other ambitious infrastructure startups. Rather than launching with a minimum viable product, Eridu spent its early months assembling a team and refining its technical approach before going public with major funding. That discipline, combined with Perkins' credibility and Kleiner's support, positions the company to make a serious run at redefining how AI systems communicate. Whether that vision translates to revenue depends on execution - but the capital and expertise are now in place to find out.
Eridu's $200 million Series A marks a major bet that AI infrastructure needs purpose-built networking, not recycled data center tech. With Drew Perkins' deep experience and John Doerr's backing, the startup has the resources and credibility to challenge both established networking vendors and fellow AI infrastructure startups. The real test comes next - can Eridu ship hardware that delivers meaningful performance gains for AI training clusters, and can it do so before the window closes on this infrastructure gold rush? For now, the company's stealth exit with one of 2026's largest Series A rounds signals that smart money sees networking as the next critical battleground in the race to power AI.
