DriveNets Scores $410 Million to Advance Its AI Fabric

DriveNets has scored $410 million in Series D funding that brings its total raised to $1 billion and its valuation to a reported $8.5 billion. The round includes investment from AMD, which validates DriveNets’ work integrating its AI Fabric technology with AMD’s GPUs and NICs—and those of other providers.

DriveNets, whose cloud-native software has been serving up telco-grade networking for a series of Tier 1 operators worldwide, such as AT&T and Comcast, for over a decade, has expanded its established service provider business into AI networking with its AI Fabric technology. This effort has focused on providing a full stack of networking software for linking AI clusters together in the datacenter and across sites.

With AMD, DriveNets has developed a reference architecture that pairs AMD Instinct MI355X GPUs and AMD Pensando Pollara NICs with DriveNets AI Fabric, which optimizes training and inference performance.

The solution incorporates DriveNets Fabric Scheduled Ethernet (FSE), which is based on Broadcom’s Jericho and Ramon ASICs and addresses congestion and load balancing at the fabric level using cell spraying; and DriveNets Endpoint Scheduled Ethernet (ESE), which works not only with AMD’s NICs but also with Broadcom’s and NVIDIA’s and prevents congestion before it enters the network by handling scheduling and traffic management at the NIC endpoint. DriveNets’ approach has been favorably received by the Ultra Ethernet Consortium, of which it is a member.

A Focus on Multivendor Architectures

One reason for DriveNets’ work with AMD has been a focus on open AI environments. Vamsi Boppana, senior vice president of AI at AMD, stated n the press release:

“AI infrastructure is entering a new era of open, integrated systems where compute, networking, and software scale together. Our support of DriveNets’ Series D reflects a shared commitment to scaling AI workloads efficiently with AMD Instinct accelerators and DriveNets’ high-performance fabric on open infrastructure, advancing open, standards-based AI data centers.”

Specifically, DriveNets has created an open-stack solution that starts at the network layer with a fabric that works with Broadcom and AMD NIC parameters. The approach also works with AMD’s ROCm Communication Collectives Library (RCCL), which is designed to be an open-source way to share and synchronize data across AMD GPUs; and with ROCm itself, the open-source software for programming AMD GPUs.

DriveNets emphasizes that its networking stack is designed to optimize performance across heterogeneous AI architectures that incorporate multivendor accelerators, NICs, and other elements. By working within the parameters of various chips, kernels, and ASICs, DriveNets can achieve optimal performance across the AI network.

To create its AMD- and Broadcom-compatible infrastructure, DriveNets has built on the networking expertise it developed over the years of supplying white-box-based networks to telcos worldwide, a business DriveNets continues to support.

A Look Ahead

DriveNets’ said its new funding will be used to scale inventory and to expand its technology to further support heterogeneous AI infrastructure solutions.

Things are looking good: DriveNets claims over $1 billion in secured business and notes that the company has been cash-flow positive since 2025. DriveNets also has a firm foothold in the telco market, which gives it extra credibility in the neocloud and hyperscaler markets.

DriveNets’ latest round was led by Bessemer Venture Partners and Atreides Management. New investors AMD and Red Dot Capital joined existing investors Pitango and D1 Capital Partners.

Futuriom Take: DriveNets’ latest funding with the participation of AMD testifies to its growing strength in heterogeneous networking for AI. With a solid position in the telecom networking market and blueprints for multivendor AI infrastructure, it is in a great position to challenge proprietary networking technologies as open architectures are increasingly favored.