Netris AI network automation software turns physical data center networks into something GPU cloud operators can run more like a cloud service than a box-by-box wiring project.
The startup has raised a $15 million Series A from Andreessen Horowitz as AI infrastructure builders race to get GPU clusters live before expensive hardware sits idle for months. That timing matters because neocloud operators can buy the GPUs and still lose the market if setup, configuration, and tenant isolation take too long. Netris was founded in 2017 in Silicon Valley. Its leadership team includes CEO Alex Saroyan alongside co-founders Tigran Martirosyan and Arsen Arakelyan.
What does Netris AI network automation do?
Here’s the simple version: Netris sits on top of the switches and network fabric inside an AI data center and lets operators define the outcome they want instead of hand-configuring every switch, VLAN, route, and policy. The platform exposes cloud-style controls such as VPCs and V-Nets. It also includes elastic IPs, NAT, load balancers, ACLs, and peering through a web console, REST API, Kubernetes CRDs, and a Terraform provider. Under the hood, its software translates those inputs into hardware-level network configuration across the whole fabric.
For a real customer, the workflow looks pretty concrete. An operator loads inventory, topology, and IPAM data into the controller, then defines a server cluster template. From there, they create clusters for Ethernet, InfiniBand, or even NVLink-heavy deployments. Netris’ docs break those clusters down far enough to cover shared endpoints, subnet choices, and different fabric designs for GPU-as-a-service environments. It's much closer to a repeatable deployment recipe than the usual spreadsheet-and-CLI mess.
The part that matters most for neoclouds is multi-tenancy. Netris enforces hard isolation in networking hardware across bare metal, virtual machines, and container workloads. It can also extend VXLAN segments into hosts through its EVPN-on-Host model. That means an operator can sell isolated infrastructure to multiple customers without relying on soft guardrails alone. For AI clouds that also expose managed Kubernetes, that hardware-layer isolation is a pretty big deal.
There's also a before-you-break-production step that's easy to miss but useful. Netris ships an init workflow for rail-optimized GPU cluster topologies. It also has a plugin that can turn controller data into a digital twin inside NVIDIA Air, so teams can simulate the network before touching live hardware. That doesn't eliminate deployment risk. But it cuts out a lot of blind trial and error.
Who built Netris AI network automation, and why are customers using it?
Founding story
Netris didn't appear because someone suddenly noticed AI in 2024. Saroyan said the company has been building toward this problem for 8 years, which lines up with Netris' 2017 founding. The pitch is straightforward: hyperscalers solved network automation internally years ago, but smaller GPU cloud operators don't have the engineering armies to do the same thing from scratch.
Founder market fit
Saroyan looks like a classic domain founder. Before Netris, he worked on carrier and core networking at Orange and later ran core network work at Ucom, where he dealt with large-capacity, multi-vendor infrastructure and rollout problems that sound a lot like the operational headaches Netris now sells against. He also has deep hands-on experience across Juniper, Cisco, Brocade, Ericsson, F5, and other networking stacks. That helps explain why Netris leans so hard into vendor-agnostic positioning.
The broader founding team also skews operator-heavy rather than pitch-deck-heavy. Tigran Martirosyan leads software engineering as co-founder, while Arsen Arakelyan is listed as co-founder and VP of customer success. That matters because Netris isn't selling a toy admin layer. It's selling something that has to survive messy production networks and impatient customers.
Traction and early signals
The market has clearly given it a shot. Netris is already live at more than 35 GPU clusters worldwide, representing about 1 million GPUs, with customers that include Lightning AI, Foxconn, Visionbay, Hewlett Packard Enterprise, TensorWave, and Telus. In APAC, it has also been talking publicly about deployments with Visionbay in Taiwan and Firmus in Australia.
That's the strongest part of the story. Plenty of infrastructure startups can demo automation. Far fewer can point to production environments where operators trusted them with multi-tenant GPU networks.
Fundraising details
Andreessen Horowitz led the new $15 million Series A, and a16z partner Guido Appenzeller is joining the board. Netris plans to use the money to hire more engineers and sales staff. It also wants to broaden hardware vendor support and add more capability to the automation engine behind the product.
One detail here is easy to overlook: Nvidia had already recommended Netris to customers after seeing a demo 2 years ago. That doesn't make Netris a winner by itself. But it suggests the company had credibility with a key power center before this financing landed.
Competition and market positioning
Netris isn't alone. Juniper has been pitching an AI-focused networking stack for GPUaaS and AIaaS providers, with Apstra automation, EVPN-VXLAN multi-tenancy, and fabric-to-GPU visibility. Arrcus comes at it from a different angle but targets similar operator pain with multi-tenant EVPN-VXLAN Clos fabrics, open APIs, and support across open networking hardware.
So where does Netris fit? It sits between the old way and the full-stack vendor way. The old way means hiring skilled network engineers and living in manual change windows. The newer vendor approach often ties automation to a broader switch or NOS strategy. Netris' bet is that neoclouds want cloud-style abstraction and hardware-enforced tenant isolation. They also want automation that works across different hardware choices — including Nvidia and AMD server environments — without rebuilding the whole stack around a single incumbent.
Why did a16z back Netris AI network automation now?
This round matters because it's not just growth capital. It's a sign that investors think networking is becoming a choke point in AI infrastructure, not just a supporting function.
Netris will use the money to add engineers, sales capacity, more vendor support, and more functionality in its core algorithms. That suggests the next phase is less about proving the product exists and more about turning it into standard infrastructure for more GPU cloud operators. If that works, Netris becomes part of the control plane for AI compute delivery. Not just a nice add-on.
There's also a sharper thesis underneath Saroyan's comments. He argues that software-only SDN falls short for AI workloads because traffic levels are too high, and that operators need hardware-accelerated control instead. His dismissal of generative AI inside this workflow is blunt — AI is “not deterministic,” he said — and that's refreshing. For network change management, “persistent and repeatable” beats clever.
How big is the market for Netris AI network automation?
The backdrop is huge. Grand View Research estimates the global data center networking market was worth $38.49 billion in 2024 and could reach $154.83 billion by 2033, a 17.2% CAGR. North America held the largest share in 2024. Asia-Pacific is projected to be the fastest-growing region.
The AI data center buildout is even more aggressive. The same firm pegs the global AI data center market at $147.3 billion in 2025, rising to $810.6 billion by 2033, with 23.9% annual growth from 2026 through 2033. That's the structural reason startups like Netris exist now: compute is scarce, timelines are compressed, and nobody wants millions of dollars in GPUs waiting around for network teams to finish manual provisioning.
Should neocloud operators care about this Series A?
Yes — because Netris is trying to solve the ugly middle layer of AI infrastructure that buyers care about but rarely celebrate.
The company still has a real challenge ahead. Network automation is crowded, incumbents are strong, and every major AI infrastructure vendor now talks about multi-tenancy and fabric management. But this Netris AI network automation round gives the company more room to turn early deployment wins into a repeatable market position. The next thing to watch is simple: whether Netris can move from dozens of GPU clusters to becoming default plumbing for the broader neocloud market.
Read how SaffronStays raised $3.5M in a funding round led by Infinity Ventures to expand its managed holiday home network, strengthen its technology platform, and scale private villa stays across India's growing leisure travel market.
FAQ
- What funding did Netris raise?
Netris raised a $15 million Series A round from Andreessen Horowitz. The financing was disclosed on June 25, 2026, and includes a board seat for a16z partner Guido Appenzeller. - How does Netris’ product work for AI data centers?
Netris gives operators a software layer that turns physical networking into cloud-like primitives they can manage through a console, API, Kubernetes, or Terraform. It handles automation and abstraction. Hardware-level tenant isolation across GPU clusters is why it's aimed at neoclouds and AI factories rather than generic enterprise IT. - Who founded Netris?
Netris was founded in 2017, and its leadership team includes CEO Alex Saroyan with co-founders Tigran Martirosyan and Arsen Arakelyan. Saroyan's background in carrier-grade networking at Orange and Ucom gives the company real operator DNA, which fits the product it built. - Is Netris an AI company or a networking company?
It's really a networking infrastructure company built for the AI buildout. Saroyan has been explicit that Netris doesn't rely on generative AI for its core operations and instead uses deterministic algorithms for automation, because changing thousands of switch configs is the kind of job where predictable behavior matters more than creativity.
