Here's something that always gets me: the most important companies in any technology wave aren't usually the ones making the noise. They're the ones quietly solving the plumbing problem that everyone else ignores until it breaks.
Netris is one of those companies. They've been building switch-level networking automation for eight years — long before the current AI infrastructure mania made anyone care about data center topology. And today, they're announcing a $15 million Series A from Andreessen Horowitz to fund the next phase of growth.
I've spent enough time in production networks to know that switch configuration is the kind of work that makes engineers cry. Not because it's hard in an intellectual sense, but because it's tedious, error-prone, and every minute spent manually configuring a link is a minute your GPUs sit idle. That's real money burning.
What Netris Actually Builds
Let me cut through the jargon, because "network automation" means different things to different people.
Netris provides software that runs on network switches. Not a control plane sitting above them. Not an overlay that adds latency. The software lives in the switch hardware itself, and they offer a platform that connects to those switches to automate setup, configuration, and ongoing operations for data center operators.
The result is network abstraction. You can change hardware configurations without rewriting your entire network stack. Multi-tenancy at the hardware layer, so you can serve multiple customers from the same physical infrastructure without security headaches. All of this happens automatically, at line rate.
Think about that last part for a second. Multi-tenancy at the hardware layer isn't something you get from software alone. It requires deep integration with the networking equipment itself — which is exactly what Netris has spent eight years building.
Why SDN Doesn't Cut It for AI
Here's where things get genuinely interesting, and where most people in the space get it wrong.
Traditional data centers used something called SDN — software-defined networking — to handle configuration changes. It worked fine for email, web traffic, the usual enterprise stuff. You could spin up VLANs, adjust routing tables, manage bandwidth allocation. All through software.
But AI workloads? They generate traffic volumes that break software-based approaches. As Netris CEO Alex Saroyan put it: "For AI, software is not okay, because the amount of traffic is so high, everything must be hardware accelerated."
I think that distinction matters more than most people realize. When you're training a model that needs to move terabytes between GPUs every few seconds, you can't afford the overhead of software processing. You need hardware acceleration at the switch level. Period.
This is why Netris started eight years ago, before the current AI boom, before everyone was suddenly obsessed with GPU clusters. They understood the problem early and built a solution that's now proven at scale.
The Numbers That Actually Matter
Let's talk about what this looks like in practice, because the deployment numbers are genuinely impressive.
Netris is currently live at more than 35 GPU clusters around the world. That's roughly a million GPUs total, operating at production scale. Their customers include Lightning AI, Foxconn, Visionbay, Hewlett Packard Enterprise, TensorWave, Telus, and others.
That's not a beta program. That's production infrastructure handling real workloads for real companies. And the fact that they've been doing this for eight years means they've seen every failure mode you can imagine — and built their system to handle them.
Here's something that should give you pause: Nvidia recommended Netris to several customers two years ago, after seeing a demo of their technology. The chipmaker doesn't hand out recommendations lightly. If Nvidia thinks your networking solution is worth deploying at scale, that's a strong signal.
The Funding Round and What It Means
So what does the $15 million Series A from a16z actually mean for Netris and the broader market?
First, Guido Appenzeller, a partner at Andreessen Horowitz, is joining Netris' board. That's not just a check — it's strategic alignment. a16z has been heavily invested in the AI infrastructure space, and having someone like Appenzeller on board gives Netris access to a network of potential customers and partners that would take years to build organically.
The company plans to use the funding to hire more engineers and sales staff, add support for additional hardware vendors, and implement more functionality in their algorithm.
I should mention something here that might surprise you: Netris isn't using AI in their product. Saroyan was explicit about this. "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."
That's a refreshingly honest take. The industry is full of companies slapping "AI" on everything, but Netris has built something that works because it's deterministic, not because it's clever. There's a difference, and it matters.
The Bigger Picture for Neocloud Operators
The AI infrastructure market is growing at a pace that makes most other technology sectors look glacial. But here's the thing: building data centers isn't just about buying GPUs and hoping for the best. You need networking that can handle the load, configuration that can scale, and operations that don't require an army of engineers.
Netris is solving the networking piece. And with $15 million in funding and a proven track record across 35+ clusters, they're positioned to capture more of that market.
The question isn't whether this technology works. It's whether the neocloud operators who haven't adopted it yet will realize they've been wasting time and money on manual configuration for too long.
I suspect the answer is obvious. Time to market is everything in this space, and every day you spend configuring switches is a day your competitors are ahead of you.
What Comes Next
Netris has eight years of experience, a million GPUs running on their platform, and the backing of one of Silicon Valley's most prominent venture firms. They're vendor-agnostic, working with both Nvidia and AMD servers, which means they're not betting on a single hardware ecosystem.
The funding round suggests they plan to scale quickly. More engineers, more sales staff, broader hardware support. If they execute well, we could see Netris become the standard for AI data center networking within a few years.
That's a bold prediction, but the numbers support it. Thirty-five clusters isn't a niche player. That's infrastructure at scale.
The AI boom has encouraged everyone and their uncle to launch a data center business. But spinning up a data center isn't easy, even if you solve the problem of securing the GPUs, network switches, and storage. You still have to get everything configured, running, and able to cater to customers' various needs.
Netris claims it can make that problem disappear for neoclouds. After seeing what they've built, I'm inclined to believe them.
For more on the infrastructure challenges facing AI operators, see Scaling AI: Why Data Infrastructure is the Real Bottleneck, which explores how data delivery bottlenecks — not compute shortages — are what actually limit production AI performance.
The Vendor-Agnostic Advantage
One detail that doesn't get enough attention: Netris is vendor-agnostic. They work with both Nvidia and AMD servers, which means they're not betting on a single hardware ecosystem.
This matters more than you might think. The AI infrastructure market is still young enough that the dominant hardware standard hasn't been established yet. AMD's MI300X is gaining traction. Nvidia's next-gen Blackwell chips are coming. Who knows what comes after that?
If your networking layer is locked to a specific vendor's ecosystem, you're vulnerable. Netris' approach — building software that runs on the switches themselves rather than integrating with a specific GPU vendor's stack — gives operators flexibility. They can swap hardware as the market evolves without ripping out their networking layer.
That's not just convenient. It's a competitive advantage for operators who want to stay ahead of the curve.
The Observability Angle Nobody's Talking About
As someone who's spent years in technology observability, I find myself thinking about what Netris' platform means for monitoring and debugging.
When you're running a million GPUs across 35+ clusters, visibility into network performance becomes critical. Traditional monitoring tools often struggle with the scale and complexity of modern AI workloads. You need to see what's happening at the packet level, understand traffic patterns in real-time, and identify bottlenecks before they cascade.
Netris' switch-level approach means they have direct visibility into the network fabric. They're not guessing what's happening based on aggregate metrics — they can see exactly how traffic flows between switches, identify congestion points, and optimize routing in real-time.
This is the kind of observability that makes the difference between a data center that runs smoothly and one that experiences intermittent failures that drive engineers crazy at 3am.
The Timing Question
Eight years is a long time to build something before it becomes relevant. But I think that's actually Netris' greatest strength.
Companies that start building infrastructure solutions during a boom often rush to market. They cut corners. They skip the hard parts. And when the boom busts, they're left with products that don't work at scale.
Netris spent eight years refining their technology before the current AI wave hit. They've seen every failure mode you can imagine. They've built their system to handle edge cases that most startups haven't even considered.
That kind of experience can't be rushed. It has to be earned through years of real-world deployment and iteration.
The $15 million from a16z isn't just funding for growth — it's validation that the long game pays off. Sometimes the best strategy is to build something genuinely good before everyone else realizes it's needed.
The broader AI infrastructure spending wave that companies like Oracle are riding — see Oracle's Strategic Pivot: Balancing AI Infrastructure and Workforce Reductions — only makes Netris' switch-level automation more essential. As hyperscalers pour billions into GPU clusters, the networking layer underneath has to keep pace or become the new bottleneck.