Netris, a Santa Clara networking company, raised a $15M Series A led by Andreessen Horowitz to scale the least glamorous but most load-bearing layer of AI infrastructure: the network that ties GPU clusters together. Its NAAM platform gives AI cloud operators one control plane across every fabric they run and enforces hard multi-tenancy in hardware, so they can launch GPU clouds in weeks and provision isolated tenants instantly. a16z partner Guido Appenzeller led the round and joined the board, and the raise dwarfs the under-$6M Netris had taken before.

  • $15M Series A led by a16z (Guido Appenzeller), after raising under $6M total in the seven years since its 2018 founding.
  • Its NAAM platform (Network Automation, Abstraction, Multi-Tenancy) spans Ethernet, InfiniBand, the NVL72 scale-up fabric and virtual/edge networking under one control plane.
  • 800% ARR growth in 12 months across 35+ AI clusters and roughly 1 million GPUs, with about a 12% share of neoclouds by cluster count.
  • Customers include Lightning AI, Foxconn, TensorWave, Telus and HPE; Netris is the first ISV validated by NVIDIA for AI network automation.
Netris NAAM: one control plane across every AI fabric An AI cloud runs Ethernet, InfiniBand, NVL72 scale-up and edge fabrics at once. Netris sits above them as a single control plane, enforcing multi-tenant isolation across all four. Netris NAAM control plane single API · hard multi-tenancy in hardware EthernetInfiniBandNVL72Edge / virtual front-end fabricback-end fabricscale-up fabrictenant networks Without it, every tenant change means hand-reconfiguring four fabrics in concert. One slip leaks data or downs a cluster. genztech.blog
Fig 1 A modern AI cloud runs four separate network fabrics, each with its own control plane. Netris abstracts all of them behind one API and enforces tenant isolation in hardware, which is what turns a months-long buildout into a weeks-long one.

What problem is Netris actually solving?

An AI cloud is not one network, it is several running at once: Ethernet for the front end, InfiniBand for the back end, the NVL72 scale-up fabric inside a rack, and virtual and edge networks for tenants. Each has its own control plane, and none of it sits still. Every time an operator adds, resizes or removes a customer, all of those fabrics have to be reconfigured together, and a single misconfiguration can take down a cluster or leak one tenant's traffic into another's. Netris collapses that into one control plane with isolation enforced on the networking hardware itself, so operators provision tenants instantly, keep isolation automatic, and reallocate GPU capacity across customers without rewiring anything by hand.

RelatedAlpaca Raises $135M for Agent-First Brokerage Rails

Why is a16z writing this check now?

Because the neocloud boom made networking the bottleneck. Dozens of specialized GPU clouds, sovereign AI operators and AI factories are racing to stand up capacity, and the constraint is rarely the GPUs, it is getting the cluster networked, multi-tenant and live. Netris sits exactly there, and the traction shows it: 800% ARR growth in a year, presence across 35-plus clusters covering roughly a million GPUs, and an estimated 12% of neoclouds by cluster count, with customers like Lightning AI, Foxconn, TensorWave, Telus and HPE. Being the first ISV NVIDIA validated for AI network automation gives it a credibility moat in a market anchored on NVIDIA hardware.

Wait, there's no AI in the AI-infra startup?

Correct, and it is a feature. CEO Alex Saroyan is blunt that the platform relies on deterministic algorithms, not machine learning: reconfiguring thousands of switch configurations does not need creativity, it needs something persistent and repeatable that never makes a novel mistake. In infrastructure that must be correct every time, predictability beats cleverness, and a system that behaves identically on every run is exactly what an operator wants holding tenant isolation together.

What it means for the market

The signal for investors is that the AI-infrastructure trade is broadening past chips into the unglamorous plumbing that makes chips usable at scale. IDC projects the AI infrastructure market reaches $758B by 2029, and networking automation is a pick-and-shovel slice of that with real switching costs once an operator standardizes on it. For a16z, Netris is a bet on the neocloud layer maturing into durable businesses that need enterprise-grade operations. The clear risk, flagged even by supportive analysts, is platform absorption: if NVIDIA or the switch vendors fold comparable automation into their own stacks, Netris's independent value shrinks. This is analysis, not investment advice: the read is that networking is a genuine AI bottleneck, and whoever owns the abstraction layer captures durable value, if the incumbents do not swallow it first. For the wider money map, see our Funding Tracker and the ranked Biggest AI Funding Rounds.

RelatedNeko Health Raises $700M at a $7B Valuation

Why is the timing so favorable?

Netris is riding a specific window. The neocloud gold rush created a wave of operators who raised capital, secured GPU allocations, and then discovered that the last mile, turning racks of accelerators into a live, multi-tenant, revenue-generating cloud, is a networking problem few of them are staffed to solve. That mismatch between GPU supply and operational readiness is exactly where a specialist tool becomes indispensable, and it explains how a company that raised under $6M in seven years suddenly posted 800% ARR growth. The a16z check is a bet that this window stays open long enough for Netris to become the default networking layer across the neocloud tier before the market matures. Sovereign AI adds a second tailwind: governments standing up national AI capacity have the same networking problem plus stricter isolation requirements, which plays directly to Netris's hardware-enforced multi-tenancy. The Singapore office is a tell that the company sees that demand concentrating in Asia, where several sovereign-AI programs are scaling fastest.

What to watch
  • Neocloud durability. Netris rises with its customers. Watch whether the neocloud wave consolidates into a few durable operators or fragments.
  • NVIDIA's roadmap. The existential question is whether NVIDIA absorbs network automation into its own platform. Watch that relationship closely.
  • Singapore expansion. The funds target engineering, commercial hiring and an Asia office. Watch whether sovereign-AI demand there converts to logos.
Primary sources

Original analysis by GenZTech. Figures as reported; current as of July 2026.