Intel unveils Ethernet E835 controllers and adapters aimed at cloud, edge and telco data centers

Intel has launched a new set of Ethernet building blocks - Ethernet E835 controllers and network adapters - designed for high-density cloud, enterprise, AI and telco workloads. Intel says the E835 series is the latest addition to its 800 Series portfolio and is intended to scale with rising data movement demands across modern data-center and edge deployments.

At the core of the announcement is throughput headroom: Intel says the Ethernet E835 solutions provide up to 200GbE bandwidth and can be configured for multiple port layouts. The company also says the platform scales from 10GbE to 200GbE, aiming to better match network capacity to changing application loads - an issue that matters as operators try to avoid overprovisioning while still keeping latency and CPU overhead under control.

Intel’s platform positioning is centered on “system throughput” and host connectivity. The company says the E835 uses high-speed host interfaces (PCIe 5.0/4.0, depending on configuration) to help drive total system bandwidth, and it includes RDMA support - Remote Direct Memory Access - via RoCEv2 and iWARP. In plain terms, RDMA is a method that allows network traffic to move data with reduced CPU involvement, which can help certain performance-sensitive applications (such as distributed training and storage systems) spend less time on packet handling.

Intel also highlights onboard network optimization and manageability. It points to Dynamic Device Personalization (DDP) to improve application performance and to DMTF specifications for manageability, including NC-SI 1.2 (a network controller interface standard used for communication between a system’s management components and network hardware). For deployments that rely on remote provisioning and fleet operations, Intel is effectively positioning E835 as a “data-plane” upgrade that is easier to operate as environments scale.

On security, Intel says E835 includes a hardware root of trust - cryptographic security anchored in hardware - plus standardized attestation using signed SPDM 1.2 (a protocol for device identity verification in security-conscious “zero-trust” architectures). The company also says its cryptographic modules meet FIPS 140-3 Level 1 requirements, a U.S. government-aligned validation commonly referenced in regulated environments.

The E835 launch is tied to Computex, where Intel says the Ethernet E835 controllers and network adapters begin shipping or launch there, with pricing and availability varying by configuration. Intel directs customers to an Intel Ethernet pricing page for recommended pricing, and it also notes that performance and security claims vary with use, configuration and deployment choices.

For buyers, the practical question is how the E835 family maps to real server designs. Intel says it plans to support a “versatile range of port configurations,” explicitly listing 2x25G, 4x25G, 2x100G and 1x200G. It also points to the Intel Ethernet Port Configuration Tool (EPCT), which it describes as a mechanism to validate a configuration and then adjust port counts and speeds later - an approach that could reduce the number of “fixed” SKU decisions customers must make up front when workloads evolve.

Intel is also leaning on an industry partner ecosystem to speed adoption. In its announcement, Intel names major OEM and platform partners including Cisco, Dell Technologies, HPE, Lenovo and Supermicro, framing the E835 lineup as intended to integrate across multiple vendor platforms. For the market, this matters because Ethernet adapters and controllers are rarely purchased in isolation; they are usually validated within specific server and switch combinations, and partner support can shorten the time from first lab testing to broader production rollouts.

Intel’s own materials include performance claims and test conditions, but they come with limits. Intel’s Ethernet product page describes a test platform and reports an example result of about 370Gbps of payload throughput per adapter, while also stating that performance varies by use and configuration and that results may not reflect later updates. The company did not provide a single universal “best” benchmark figure in the announcement itself, nor did it disclose exact pricing across the full E835 configuration range within the newsroom release.

Looking ahead, the main indicates to watch are integration timelines and how quickly the E835 family shows up in production systems for AI, cloud and telco mobile core deployments. Intel directs customers to its Ethernet Performance Index for workload-specific results and configurations, and the Computex launch suggests OEM validation efforts may accelerate over the coming weeks. In practical terms, buyers will likely focus on compatibility with their existing software stacks, RDMA behavior under their storage or training workloads, and whether attestation and manageability features are supported cleanly across their chosen server platforms.