DriveNets Lands $410M to Scale AI Networking

The big picture: DriveNets, an Israeli networking software company, has closed a $410 million Series D funding round. The company builds Ethernet fabric, the high-speed networking layer connecting thousands of processors inside large-scale AI data centers. Ido Susan and Hillel Kobrinsky founded DriveNets in December 2015.

Why it matters:

• Shifting Landscape: AI data centers are moving away from single-vendor hardware stacks, like NVIDIA’s, towards more flexible, multi-vendor environments, increasing demand for adaptable networking solutions.
• Ethernet Dominance: Ethernet switches in AI back-end networks more than tripled in 2025, accounting for over two-thirds of data center switch sales in AI clusters, according to Dell’Oro Group.
• Market Growth: The broader data center networking market is projected to grow from $55.6 billion in 2025 to $139 billion by 2031, driven by the escalating demands of AI training and inference workloads.

How it works:

• Disaggregated Networking: DriveNets applies its core approach of disaggregated networking, running its Network Cloud software on standard, commodity hardware, offering customers greater flexibility and cost control over their infrastructure choices.
• Optimized Fabric: The company optimizes across the full networking stack, including communication libraries, transport protocols, network interface cards, the fabric itself, and orchestration software, to achieve higher tokens per second and lower cost per token.
• Heterogeneous AI: DriveNets is expanding its Heterogeneous AI infrastructure offering by building deeper compatibility with a wider range of AI accelerators, collaborating with partners like AMD, Broadcom, Dell, and Supermicro to improve GPU utilization in mixed-vendor environments.

The catch: While DriveNets’ telecom history provides a strong foundation for high-performance networking, the AI infrastructure market is intensely competitive. Established players like NVIDIA continue to innovate with their own networking solutions (e.g., InfiniBand, Spectrum-X Ethernet), and other networking giants are also vying for market share. The complexity of integrating and maintaining performance across a rapidly evolving, multi-vendor AI hardware landscape presents a continuous challenge.