Getting a GPU cluster online is one thing. Getting it profitable is another. The Netris Series A funding round, a $15 million raise led by Andreessen Horowitz (a16z), is a bet that the unglamorous plumbing between those two moments is where real money will be made.
Netris, headquartered in Santa Clara and co-founded by CEO Alex Saroyan, Arsen Arakelyan, and Tigran Martirosyan, all longtime network engineers according to Forbes, sells software that automates the setup, configuration, and ongoing operations of the networks inside GPU data centres. Its platform, formally named NAAM (Network Automation, Abstraction and Multi-Tenancy), is positioned by the company as the next generation of network automation, succeeding SDN and intent-based networking for the specific demands of AI infrastructure, according to the Netris Series A press release.
The Problem Every Neocloud Operator Faces
The neocloud boom has produced dozens of operators building GPU clusters to serve AI inference and training workloads. Unlike the traditional data centres run by the likes of Equinix, NTT, or hyperscalers such as Microsoft and AWS, these smaller outfits rarely have armies of network engineers available to configure everything by hand. Every day, operators must push configuration changes to every link in the cluster.
SDN, the software layer traditional data centres relied on, cannot keep pace. ‘For AI, software is not okay, because the amount of traffic is so high, everything must be hardware accelerated,’ Saroyan said. ‘So you need something like SDN, but completely hardware accelerated. This is what we do, and this is what we’ve been doing for eight years.’
There is also a structural reason older approaches fall short. According to the a16z investment announcement, AI GPU data centres now require not just a classic front-end network but a second, back-end network that directly connects GPUs with each other. Managing both layers simultaneously, across multiple customers sharing the same hardware, is precisely the challenge NAAM addresses. The platform provides network abstraction so hardware configurations can be changed as requirements shift, and isolates servers and resources at the hardware layer so that neoclouds can serve multiple tenants from a single cluster.
Netris Series A Funding: the a16z Rationale
The round was led by a16z, with partners Guido Appenzeller, Martin Casado, Raghu Raghuram, and Jason Cui involved in the investment, according to the Netris press release. Appenzeller is joining the company’s board. In their investment announcement, the a16z partners noted that GPU data centres built for close to a billion dollars are already entrusting their daily functioning to Netris software, citing this as a strong sign of market confidence.
The commercial numbers are hard to dismiss. The company says its platform is live at more than 35 GPU clusters worldwide, covering around a million GPUs operated by customers including Lightning AI, Foxconn, Hewlett Packard Enterprise, Tensorwave, and Telus. Annual recurring revenue grew by 800% in the last 12 months, according to the company’s press release. Cryptobriefing reports that Netris claims roughly a 12% share of the neocloud market by cluster count.
Nvidia’s endorsement arrived early in that trajectory. Two years ago, following a product demonstration, Nvidia recommended Netris to several of its own customers. Cryptobriefing also reports that Netris is the first independent software vendor validated by Nvidia for AI network automation, a distinction that carries weight when operators are choosing which software to trust with multi-million-dollar hardware estates.
No AI Inside the Machine
One point Saroyan is careful to make: NAAM does not use artificial intelligence. The company relies on deterministic algorithms built over years of network engineering work, predating the current AI wave entirely. ‘We started way before AI. We understood the challenge early on, and we started developing this algorithm early on,’ Saroyan said. ‘AI is not deterministic, right? Sometimes it likes to do things on its own. It’s good for creative work, but for changing many thousands of switch configurations, you don’t need to be creative. You need to be very persistent and repeatable.’
The vendor-agnostic design already covers networking equipment compatible with both Nvidia and AMD server environments. The $15 million will go towards hiring engineers and sales staff, adding support for more hardware vendors, and extending the platform’s algorithmic capabilities, according to the company. More detail on the broader ecosystem Netris has assembled around Nvidia is covered by Pulse 2.0.
The Netris Series A funding of $15 million arrives as the window for neocloud operators is at its most unforgiving. Every week a GPU cluster sits unconfigured is revenue that does not exist, and the cost of idle GPUs compounds quickly. Netris has built its business on that urgency. Whether this raise is enough to establish an unassailable position before a well-capitalised competitor or a hyperscaler decides the same problem deserves their attention is the next question on Saroyan’s list.
