Netris sells software that automates the network layer underneath smaller, GPU focused AI cloud operators (the so called "neocloud" category), where a paid for GPU rack can sit unconfigured for months before it earns revenue.
The race to stand up a new AI cloud has a quiet choke point that has nothing to do with chip supply. Once a cluster of Nvidia or AMD accelerators is paid for, the next step is a months-long engineering project: wiring the switches, configuring tenants, and stitching the rest of the data center around the rack. Netris, an 8-year-old network-automation startup, raised $15 million in Series A funding from Andreessen Horowitz (a16z) on Wednesday to sell software that automates exactly that work, according to TechCrunch and the company's press release on BusinessWire.
What makes the round interesting is the customer base Netris is selling into: a category the company and its peers have started calling "neoclouds." These are smaller, often venture-backed operators trying to sell GPU compute alongside the hyperscalers (AWS, Google, Microsoft, Oracle) and the older colocation giants (Equinix, Digital Realty, NTT). Neoclouds sit in a new middle of the AI infrastructure market, too small to build their own silicon supply chain and too specialized to just resell hyperscaler capacity. The pull is real, but so is the operational cost of being late. A GPU cluster sitting unconfigured is a GPU cluster not earning revenue, and on a high-end GPU that retails for $30,000 or more, that drag shows up on the monthly burn rate within weeks.
Netris pitches its software as the network layer those neoclouds should not have to build by hand. The product runs on commodity switches, exposes a programmable fabric, and gives operators a way to carve shared GPU infrastructure into isolated tenants at the hardware level, rather than as an overlay on top of it. The company frames this as a displacement of legacy software-defined networking, or SDN, for AI traffic, a positioning its CEO has pushed publicly and that the company repeats in its Series A announcement. Independent engineers and competing vendors have not yet weighed in on whether the claim holds at scale.
The customer list Netris and a16z are selling is unusually long for an early-stage infrastructure round. Stated neocloud users include Lightning AI, STN, Boost Run, and TensorWave. Stated sovereign-AI-cloud customers and partners include TELUS, DCAI, and YOTTA. Stated AI-factory operators include the Foxconn-backed Visionbay.ai, which runs Taiwan's largest GPU cluster, and Firmus, a renewable-powered sovereign AI factory in Australia. Nvidia has, by Netris's account, been routing customers its way. HPE is named as a platform partner (TechCrunch). Whether any of those customers would publicly endorse Netris as a category standard is a separate question; the names are on a vendor's list, not a buyer's case study.
The company also disclosed, in a March press release, that it has grown 622% in ten months and now claims 12% of the global neocloud market. Both numbers are self-reported and unaudited. The 12% figure is striking because it implies Netris has decided what counts as the neocloud market, a definition the rest of the industry has not agreed on. A reader should treat both numbers as the company's pitch, not market consensus.
The bet a16z is underwriting with this round sits on a layer of the AI stack the public conversation mostly ignores. Chip supply, model progress, and hyperscaler capex dominate headlines. The wiring underneath is, in most accounts, treated as plumbing that engineers will figure out. Netris's argument is that plumbing is now the gating constraint, and that a separate category of automation tooling is needed before the next wave of neocloud capacity can actually go live and start earning. If the bet is right, the bottleneck shifts from silicon procurement to network provisioning, and the companies that automate the second become as strategic as the ones that secured the first.
There is also a real reason to keep this in proportion. Netris has been around since 2018, and this is its first Series A. Eight years of existence before a Series A is not a red flag on its own, but it is a data point: the category it sells into did not exist at meaningful scale until very recently. The "vendor-agnostic" claim, common in this layer of the stack, has not been independently stress-tested in the public record. The framing of Netris as a replacement for legacy SDN is the company's positioning, not yet a market verdict. A reader who follows the money into this round should also follow whether the next generation of neoclouds cites Netris by name when they describe how they went live, and how long it actually took.