NextFin News - Netris has raised $15 million in a Series A round from Andreessen Horowitz as AI infrastructure operators race to bring GPU clouds online faster. The funding underscores a new bottleneck in the AI buildout: not only getting compute, switches, and storage into a data center, but turning that hardware into a usable multi-tenant service quickly enough to capture demand. For neocloud operators, time-to-production is becoming as important as raw capacity.
The company’s software is built to automate network setup, configuration, and operations on switches while also providing network abstraction and hardware-layer isolation. In practical terms, that means Netris is trying to make it easier for operators to move from racks of hardware to a live cloud service without rebuilding the fabric from scratch each time a customer, workload, or hardware change arrives.
CEO Alex Saroyan framed the product around a simple operational problem. GPU cluster operators have to make repeated configuration changes across large numbers of links, and traditional software-defined networking is not enough when the traffic load is high and the environment has to remain repeatable. Netris says its platform is meant to make those changes persistent, deterministic, and hardware-accelerated.
The round also gives Andreessen Horowitz a board presence in the company. a16z partner Guido Appenzeller is joining Netris’ board, giving the investor a direct role in the startup as it tries to turn a technical thesis about AI networking into a broader commercial platform. The company plans to use the new capital to hire engineers and sales staff, add support for more hardware vendors, and expand the functionality of its algorithms.
Netris says it was co-founded in 2018 and has spent eight years building what it calls NAAM, short for Network Automation, Abstraction, and Multi-Tenancy. The company positions the category as infrastructure for AI operators that need secure multi-tenancy, frequent configuration changes, and a reliable path from installed hardware to a revenue-generating service.
The Real Constraint Is Operational, Not Just Computational
The headline number in the AI infrastructure boom is usually compute, but the harder problem is operational readiness. A cluster can have enough GPUs, yet still fail to become a business if the network fabric is too slow to configure, too brittle to change, or too difficult to segment for multiple customers. That is the gap Netris is trying to fill.
Saroyan said the challenge shows up in the day-to-day reality of a GPU cluster: network changes have to be made across every link, every day. In traditional data centers, operators have relied on software-defined networking, but he argued that approach falls short for AI-scale traffic because the system needs to be repeatable and hardware accelerated rather than software-only.
"As a GPU cluster operator, you need to make configuration changes to every link, every day. At traditional data centers, they were using something called SDN [software-defined networking] to do this, but SDN is falling short, because it’s a software technology," Saroyan said.
That is a meaningful distinction. AI infrastructure is not just a bigger version of normal cloud infrastructure; it is a faster, denser, and more operationally demanding version. Each new tenant, each maintenance event, and each hardware swap can create friction if the fabric cannot be reconfigured quickly and consistently. A networking layer that removes those steps can directly affect how soon a provider can bill customers.
The company’s broader pitch depends on that logic. If neocloud operators can go live faster, they can start generating revenue sooner and keep more expensive hardware in use. If they cannot, the economics of the whole stack get worse, because idle accelerators and delayed tenant onboarding are both costly in a market where margins depend on utilization.
That is why Netris is not selling networking as a back-office utility. It is selling it as part of the commercial engine of an AI cloud. The more quickly an operator can configure, isolate, and reconfigure the hardware, the faster it can convert capex into a customer-facing service.
Why a16z Is Backing the Layer Below the Layer
Andreessen Horowitz’s investment is also a signal about where infrastructure money is flowing. The firm is backing a company that sits below the application layer and below the model layer, in the part of the stack that helps AI clouds function at all. That is a bet that the market for AI infrastructure will not stop at chips or servers; it will extend into the systems that manage them.
The board seat for Guido Appenzeller strengthens that interpretation. His role suggests the investor sees Netris as a technical infrastructure business, not a broad software company that happens to touch networking. For a category still defining itself, that matters. A startup that can prove it is the control layer for AI fabrics may be able to turn a narrow technical advantage into a durable enterprise position.
Netris says its platform is built around network automation, abstraction, and multi-tenancy for AI infrastructure teams. Those functions matter because AI clouds need to serve different customers securely while changing the underlying hardware configuration often enough to keep operations efficient. In other words, the company is trying to solve the problem of how to make a GPU cloud both flexible and repeatable.
"For AI, software is not okay, because the amount of traffic is so high, everything must be hardware accelerated. 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," Saroyan said.
The argument is also about category creation. If Netris can establish itself as the software layer that makes AI networking operational, it may become harder for customers to rip and replace it later. Infrastructure tools often gain leverage that way: once they become embedded in configuration, isolation, and orchestration, they are less like optional software and more like part of the machine.
That creates a more interesting market than a simple point solution. The company is aiming to support more hardware vendors and broaden its algorithmic functionality, which implies a product strategy that has to work across different environments rather than only in a single standardized setup. That kind of flexibility can be valuable in a market where AI infrastructure is still being assembled.
There is also a timing element. The current wave of AI infrastructure spending has already lifted the importance of networking, but the next phase may be defined by who can operationalize those investments fastest. Netris is positioning itself squarely in that race.
What The Round Signals For AI Infrastructure
The deeper implication of the deal is that the AI infrastructure market is broadening. It is no longer just about acquiring GPUs or building data centers. It is also about making those assets usable, secure, and economically productive at speed.
That shift favors companies that can shorten the path from hardware delivery to live service. It also raises the value of tools that reduce manual intervention, because every manual step adds friction to onboarding, maintenance, and tenant changes. In a market where neoclouds compete on speed and specialization, operational software can become a differentiator in its own right.
For investors, Netris offers a specific thesis: the next layer of AI infrastructure value may sit in the networking fabric that operators rarely see but cannot do without. If that thesis proves right, the winners in the current buildout will not just be the companies that sell capacity, but the ones that make capacity easy to use.
The company’s next test is execution. It will need to convert the financing into broader hardware support, stronger engineering depth, and more customer adoption across AI operators. The market will then decide whether Netris is merely a useful tool for a niche of ambitious neoclouds or an infrastructure layer with staying power.
"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," Saroyan said.
That may be the cleanest summary of the company’s case. In AI infrastructure, the glamour sits at the top of the stack, but the revenue often depends on the layer that keeps the stack stable. Netris just raised $15 million to own more of that layer.
Explore more exclusive insights at nextfin.ai.
