A16z and Netris Get the SDN Band Back Together

Top venture capital firm Andreessen Horowitz is placing a key bet in AI network automation, investing $15 million in the Series A for Netris, a leading provider of network automation and multi-tenancy for AI infrastructure and a long-time member of our Futuriom 50 list (six years, to be precise).

Netris was founded by longtime network engineers CEO Alex Saroyan, Arsen Arakelyan, and Tigran Martirosyan, and it has had some twists and turns along the way. But through persistence, repositioning, and a tight relationship with NVIDIA, it may have solved one of the key bottlenecks in AI infrastructure—configuring a wide array of networking equipment on the fly, using programmatic techniques that combine software-defined networking (SDN) and intent-based networking.

This software is proving popular with some of largest neoclouds, sovereign AI providers, and AI factories globally, which have been struggling with networking configurations across a wide variety of networking hardware and vendors. Netris says it has already demonstrated success with 35+ live deployments in the AI cloud space.

The A16z All-Star SDN Team Steps In

At $15 million, it's true that the Netris deal is small. But this is a scrappy startup. And, as we've been telling you, it is on to something.

The A16z investment team is a "Who's Who" of experts from the early days of SDN. It's the original team that took startup Nicira and integrated it into VMware to build one of the industry's most successful virtualized networking platforms, NSX.

A16z General Partner Guido Appenzeller led the round and is joining the Netris board. The investment is also being supported by A16z general partners Martin Casado and Raghu Raghuram, who both helped build the datacenter networking franchise at VMware. Casado founded Nicira, the startup that pioneered software-defined networking and was acquired by VMware in 2012; Raghuram rose to become VMware's CEO. Backing Netris looks like a bet on reinventing SDN for AI cloud needs, just as the team led networking reinvention for the datacenter. Netris could become the NSX of AI clouds.

"Netris spent years building the platform AI cloud operators now standardize on," said CEO Saroyan. "That foundation is why we've reached 35+ live deployments at many of the largest AI clouds in the world and built the ecosystem around them. Andreessen Horowitz has a long track record backing category-defining infrastructure companies. Martin Casado, Raghu Raghuram, and Guido Appenzeller revolutionized networking for datacenters. Netris is doing the same for AI.”

The Need for NAAM

At last count, at least $800 billion is expected to be spent on AI infrastructure this year, with trillions of dollars to be deployed over the next few years, according to Futuriom research.

Deploying infrastructure at this scale is hard to do, and networking is the hardest part of it. An AI cloud runs across several network fabrics at once — Ethernet, InfiniBand, the NVL72 scale-up fabric, and virtual and edge networking. Each of these technologies has its own control plane. None of it sits still: Every time an operator adds, resizes, or removes a tenant, those fabrics have to be reconfigured in concert across every layer, and one misconfiguration can take a cluster down or leak one tenant's data into another. It demands a new approach to orchestration and automation.

That need is giving rise to a new category of orchestration software—one built to automate the network build-out of GPU clusters, abstract the fabrics beneath them, and enforce hard multi-tenancy in hardware. Netris labels its approach NAAM, for Network Automation, Abstraction, and Multi-Tenancy, and it has become the platform neoclouds standardize on as the category takes shape.

According to sources I have spoken with, Netris is especially good at solving the problems of abstracting networking and configuring multi-tenancy across a variety of multivendor hardware infrastructure. Multi-tenancy, which was a key driver of cloud services, enables infrastructure to be partitioned for separate customers. This is a particular challenge when selling GPU and AI compute as-a-service.

While many networking systems have their own orchestration and management tools—such as those from Arista, Cisco, HPE, and NVIDIA—they often fail to solve orchestration and multi-tenancy across complex, multivendor environments. This was an important topic at the recent Cisco Live event in Las Vegas and a driver behind Cisco’s own abstraction and orchestration product, Cloud Control. But it’s not yet clear how Cloud Control will interoperate with non-Cisco systems.

Futuriom research indicates that many AI clouds, including the hyperscalers, neoclouds, and altscalers, are looking for better software abstraction tools that allow them to manage many different types of infrastructure from a variety of vendors, especially in networking. So that’s why there is huge potential for Netris’s NAAM category.

NVIDIA Helps Fuel Netris Growth

Netris has been meeting strong demand, with an NVIDIA partnership behind its growth. In the last 12 months, Netris says it has experienced 800% annual recurring revenue growth, enabling network automation and multi-tenancy for 35+ AI clusters across the world’s largest neoclouds, sovereign AI operators, and AI factories.

NVIDIA customers use Netris orchestration and abstraction software for NVIDIA hardware deployments worldwide. In addition to NVIDIA, Netris has built an ecosystem of partners with other technology companies including Mirantis, Rafay, Red Hat, Spectro Cloud, vCluster, and HPE.

Netris customers include AI clouds such as Lightning AI, STN, Boost Run, and TensorWave; sovereign AI cloud providers, including TELUS, DCAI, and YOTTA; AI factories, including Foxconn-backed Visionbay.ai, operator of Taiwan's largest GPU cluster, and Firmus, operator of Australia's largest renewable-powered sovereign AI factory; and AI platform providers such as HPE, which delivers a full-stack AI solution for customers in research, education, and state and local government.

A16z Sees an AI Cloud Opportunity

The presence of A16z’s top partners in this deal shows the firm sees the network abstraction challenge as an intriguing opportunity.

"Every era of computing has needed a new networking foundation—first for the virtualized data center, then for the cloud, and now for AI," said Guido Appenzeller of Andreessen Horowitz. "GPU clusters run across many fabrics at once, and legacy automation was never built for that. Netris is the platform AI cloud operators standardize on to solve it, and we are excited to partner with them on this journey."

In a video shared with Futuriom, Casado reflected on some of the parallels between Netris and his successful SDN years with Nicira and VMware. “We’re in the early innings of the largest growth and compute we’ve ever seen in the AI wave. It will need SDN-type approaches. It will need what Netris is doing. So we think this is going to be the standard for every company going forward.”

Netris, which already has a strong list of companies along with the support of NVIDIA, says it will use the funding to expand its team and global presence and to grow its partner ecosystem as AI infrastructure scales worldwide.