Netris just raised $15 million in a Series A round led by Andreessen Horowitz, money aimed squarely at one of the AI boom’s quieter bottlenecks: getting new data centers online. TechCrunch AI reports the round exclusively, and it puts a16z partner Guido Appenzeller on the company’s board. The pitch is simple. Spinning up a GPU cluster is slow, and every day those chips sit idle costs real money.
What Netris actually does
The company builds network automation software for “neoclouds,” the wave of smaller cloud providers renting out GPU compute for AI training and inference. Securing the GPUs, switches, and storage is only half the battle. You still have to configure everything, wire up the network, and isolate customers from each other so multiple tenants can share the same hardware safely. That work can drag on for months.
Netris runs software directly on network switches and adds a platform that automates setup, configuration, and day-to-day operations. It also abstracts the network, so operators can change hardware configs on the fly, and it isolates servers at the hardware layer for multi-tenancy. As detailed in TechCrunch AI, the platform is vendor-agnostic and works with both Nvidia and AMD gear.
Why this matters
For years, data centers were the domain of giants like Equinix, NTT, Digital Realty, Oracle, Microsoft, AWS, and Google. Those companies solved network setup and multi-tenancy by throwing armies of engineers at the problem or building their own automation. Smaller neoclouds don’t have that luxury. Netris is selling them the shortcut.
What stands out here is the technical argument. CEO Alex Saroyan told TechCrunch that traditional data centers leaned on software-defined networking, or SDN, to manage constant configuration changes. For AI workloads, he says, that breaks down.
“As a GPU cluster operator, you need to make configuration changes to every link, every day,” Saroyan said. “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.”
The Nvidia signal
Two years ago, Nvidia was impressed enough by a Netris demo that it recommended the company to several customers. That endorsement carries weight in this market. Today Netris says it’s live across more than 35 GPU clusters worldwide, roughly a million GPUs in total, run by:
- Lightning AI
- Foxconn
- Visionbay
- Hewlett Packard Enterprise
- TensorWave
- Telus, and others
That’s real deployment, not a slide deck.
The twist: no AI inside
Here’s the detail I find most interesting. There’s no AI in Netris itself. The company runs on deterministic algorithms it developed years before the current boom.
“AI is not deterministic, right? Sometimes it likes to do things on its own,” Saroyan said. “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 sharp reminder. Not every problem the AI boom creates needs an AI solution. Sometimes the infrastructure layer just needs to be fast, predictable, and boring in the best way.
What to watch next
Netris plans to spend the new funding on engineers and sales staff, support for more hardware vendors, and deeper functionality in its algorithm. The broader signal is what counts. Capital is flowing into the unglamorous plumbing of AI, not just the models and chips that grab headlines. As neoclouds keep multiplying, the companies that shrink time-to-market for GPU clusters become quiet kingmakers.
If you’re building or buying GPU capacity, expect the competition over deployment speed to heat up. You can find the full details at the original TechCrunch AI report.