AI infrastructure booms bring a large number of new cloud service providers to the market, but data centres are often subject to long network configuration and preparation, from equipment to real service. For operators relying on GPU clusters for training and reasoning, the longer the deployment cycle, the greater the cost pressure on idle hardware.

TechCrunch reported that Netris, a network automation start-up company, had completed $15 million in A rounds of financing, with the investor Andreasen Horowitz (a16z). The company provides software running on the network switch, as well as a management platform to connect the switch to automate deployment, configuration and transportation processes to help AI neocloud provide cloud computing services more quickly.

Used for multiple GPU clusters

Netris ' core product is oriented towards the data centre network layer. According to the company, its platform could isolate servers and resources at the hardware level to support multi-tenant services while allowing operators to reduce the costs of adaptation when replacing or adjusting hardware configurations through network abstraction capabilities.

The company CEO Alex Saroyan states that while the traditional data centres have long relied on the software definition network (SDN) to process configuration changes, in the AI scenario, the network traffic is higher and the software alone is difficult to meet. The Netris programme emphasizes the acceleration of hardware and states that it has been developed in this direction for many years.

The Netris platform is reported to be compatible with a number of equipment manufacturers, adapt to common data centre network standards and also support the Nvidia and AMD server environments.

Nvidia recommended it to the client

Netris states that approximately two years ago, Nvidia, having seen its technical presentation, recommended the company to some of its clients. Today, the system is operational in more than 35 GPU clusters around the globe, covering a total of about $1 million GPU.

  • Deployment of more than 35 GPU clusters globally
  • Overwrite GPU totalling approximately $1 million
  • Clients include Lightning AI, Fuscon, etc.

It was also reported that Netris products were also being used by enterprises such as Visionbay, Hewlett Packard Enterprise, Tensorwave and Telus.

Funds will be used to expand the team.

After this round of financing, a16z partner Guido Appenzeller will join the Netris Board. The company plans to use new funds to recruit additional engineering and sales personnel, expand support to more hardware manufacturers and continue to refine existing algorithm capabilities.

At the same time, Netris emphasizes that its products do not directly use generated AI. According to Saroyan, the company uses automated and operational maintenance algorithms that were previously developed, as large-scale switch configurations require more stable and replicable implementation than open capacity to generate models.