Broadcom (AVGO) backs DriveNets with Tomahawk 6 for open Ethernet AI data-center fabrics

Broadcom is lending support to DriveNets’ next push in AI data-center networking, with the chipmaker’s Tomahawk 6 used as a core component in an effort to build “open Ethernet” networking platforms for large-scale AI clusters. The move, highlighted in a report published by Yahoo Finance, frames Tomahawk 6 as enabling the kind of AI Fabric interconnect intended for hyperscale deployments.

DriveNets’ expansion centers on networking gear that is designed around open Ethernet principles rather than proprietary interconnects. In practical terms, the pitch is that hyperscale operators can deploy AI traffic networks using widely adopted Ethernet approaches, while still targeting the low-latency, high-bandwidth needs of modern AI training and inference at scale. The Yahoo Finance report characterizes DriveNets’ platforms as aimed at “large scale AI data centers,” with Tomahawk 6 positioned as the underlying accelerator for that goal.

Broadcom’s involvement underscores the broader pattern in the AI infrastructure stack, where merchant silicon and networking chips increasingly function as a “common engine” for multiple systems integrators. In Broadcom’s broader networking roadmap, the company has also emphasized distributed AI computing and data-center scale interoperability with its earlier switching and networking products, a theme reflected in its product release about shipping Jericho4, which Broadcom described as enabling distributed AI computing across data centers.

The Jericho4 release, dated Aug. 4, 2025, said Broadcom shipped the product and linked it to distributed AI computing across data centers. While the release does not address DriveNets or Tomahawk 6 specifically, it provides context for how Broadcom continues to position its networking hardware for scaling AI infrastructure and connecting computing environments at broader geographic and rack-to-rack levels.

For DriveNets, the technical direction implied by “open Ethernet AI data centers” is that its AI Fabric strategy depends on interoperable networking designs that can be adopted across different hardware ecosystems. Ethernet is already the dominant link layer in most data centers, and vendors have been working to tailor it for AI traffic patterns, including rapid collective communications and heavy east-west movement of data across racks and clusters.

Notably, the Yahoo Finance report does not provide additional publicly verifiable details in the material available here, such as performance benchmarks, list prices, customer names, or deployment timelines. It also does not specify whether DriveNets has announced a particular commercial launch date, service-level targets, or certification status with any hyperscale operator.

Broadcom’s Tomahawk 6 association with DriveNets suggests a continued strategy for expanding usage of its AI-oriented networking silicon among companies building fabric solutions. For the market, these alliances matter because AI clusters are increasingly tied to specialized networking, and the ability to scale fabric capacity can determine how quickly operators can add compute without suffering network bottlenecks.

Looking ahead, investors and customers will likely focus on whether DriveNets can translate the open Ethernet positioning into measurable improvements in throughput, latency, and fault tolerance at scale, and whether Broadcom’s role extends beyond silicon supply into joint validation efforts. Additional disclosures, such as customer deployments, reference designs, and performance data, would help clarify how broadly the approach will be adopted across hyperscale AI environments.