a16z bets $15M on Netris to get AI neoclouds online faster
Andreessen Horowitz has led a $15M round into Netris, the quiet networking layer behind GPU clouds. For India's neocloud race, the timing could not be sharper.
The networking plumbing behind the AI boom rarely makes headlines, but it just attracted a fresh cheque from one of Silicon Valley's best-known firms. Netris, a software company that automates the network layer inside GPU data centres, has raised a $15 million Series A led by Andreessen Horowitz (a16z). Guido Appenzeller of a16z is joining the company's board as part of the deal.
The News
The eight-year-old firm sells hardware-accelerated network automation software that runs on data centre switches. Its pitch is narrow but valuable: it lets operators of large GPU clusters spin up multi-tenant, abstracted networks quickly, cutting the months-long slog of wiring and configuring a cluster down to something far shorter. Netris is vendor-agnostic, working across both Nvidia and AMD hardware, and it leans on deterministic algorithms rather than AI to manage configuration.
The company says its software is live across more than 35 GPU clusters that collectively manage roughly one million GPUs. Customers named include Lightning AI, Foxconn, Visionbay, Hewlett Packard Enterprise, Tensorwave and Telus. Two years ago, Nvidia was confident enough in the technology to recommend Netris to several of its own customers.
"For AI, software is not okay, because the amount of traffic is so high, everything must be hardware accelerated," said chief executive Alex Saroyan, explaining why traditional software-defined networking falls short for AI workloads.
Why It Matters
The funding is small by the standards of the current AI cycle, but the target is telling. Investors have spent two years pouring billions into chips, models and the "neoclouds" - independent operators renting out GPU capacity to compete with the hyperscalers. The bottleneck is quietly shifting. Buying GPUs is the easy part; getting tens of thousands of them to talk to each other efficiently, without idle hardware burning cash, is where build-outs stall.
That is the gap Netris is selling into. When CoreWeave and a wave of rivals proved that GPU-as-a-service could become a multi-billion-dollar business, they also exposed how unglamorous operational software became the difference between a profitable cluster and an expensive one. The last time networking startups drew this kind of attention was the software-defined networking wave of the mid-2010s; this round suggests AI has revived demand for the unglamorous layer beneath the models.
Indian Angle
For India, the timing is pointed. The IndiaAI Mission has committed to subsidising access to tens of thousands of GPUs, and a domestic crop of neoclouds - Yotta, E2E Networks, NeevCloud, Tata and Jio among them - is racing to stand up sovereign GPU capacity. Every one of them faces the same problem Netris addresses: idle, mis-wired clusters are pure cost, and time-to-live directly shapes the per-hour price they can offer Indian developers.
That matters because cost is the whole game for Indian AI builders. Startups such as Sarvam and Krutrim are training models on rented compute, and the rupee price of a GPU-hour is shaped as much by cluster utilisation as by the chips themselves. Faster, cleaner network provisioning is one of the few levers that can pull that price down without waiting for cheaper silicon.
There is a regulatory thread too. MeitY's push for sovereign compute assumes Indian operators can run world-class clusters at competitive economics. Software that compresses build-out timelines is precisely the kind of capability that decides whether India's subsidised GPU pool becomes genuinely usable infrastructure or an underused asset.
FAQ
What exactly does Netris sell?
Software that automates the network layer inside GPU data centres, running on switches to provide multi-tenancy and faster provisioning. It works across Nvidia and AMD hardware and uses deterministic algorithms rather than AI.
How big is the round?
A $15 million Series A led by Andreessen Horowitz, with a16z's Guido Appenzeller joining the board.
Why should Indian neoclouds care?
Build-out speed and cluster utilisation directly determine the rupee cost of GPU time, the key constraint for India's IndiaAI Mission and domestic AI startups.
Where can I read the original report?
TechCrunch published the original coverage, linked below.
This story was reported by TechCrunch. Read the full original coverage at TechCrunch.