400G Ethernet Overview Data center switch sales grew by double digits in 2023, and 200/400 Gbps port shipments nearly doubled that same year, according to Dell'Oro Group's data center switch market analysis. By 2027, Dell'Oro expects nearly half of all data center switch ports to run at 400 Gbps or higher.

The driver isn't a mystery. AI training clusters now involve trillions of parameters and thousands of accelerated nodes, all needing to talk to each other fast. That's straining networks built for 100G.

Here's the problem: upgrading isn't cheap. New 400G switches and optics carry steep OEM price tags, and many IT teams are stuck choosing between overspending or falling behind.

This guide covers what 400G Ethernet actually is, how it works, why it's worth the upgrade, and how to get there without blowing your budget — including through refurbished hardware.

Key Takeaways

  • 400G Ethernet delivers four times the throughput of 100G, using PAM4 modulation to pack more data per lane
  • IEEE ratified the 802.3bs standard for 200G/400G in December 2017; 800G is already shipping widely
  • Consolidating four 100G links into one 400G port improves density and reduces per-port overhead
  • Refurbished 400G switches and transceivers offer a lower-cost entry point for cost-conscious providers and enterprises

What Is 400G Ethernet?

400G Ethernet (400GbE) is the IEEE 802.3bs standard, ratified in December 2017, delivering 400 gigabits per second of throughput, four times the speed of 100G. It's part of the same standardization effort that introduced 200G.

Naming conventions vary, but they all mean the same thing:

  • 400 Gigabit Ethernet
  • 400GbE
  • 400 Gbps

It sometimes gets folded into "Terabit Ethernet" discussions, but that label is inaccurate: a terabit is 1,000 Gb/s.

What Is a 400G Switch?

A 400G switch is built around ASICs and ports capable of aggregating or breaking out 400 Gbps links. These sit at the spine and core layers of a data center network, where they handle the heaviest traffic loads.

Primary use cases include:

  • Hyperscale data centers
  • Cloud service providers
  • Telecom backbone networks
  • AI/ML training clusters, where node-to-node bandwidth is the bottleneck

The Evolution of Ethernet Speed Standards

Ethernet speeds have climbed steadily since the late 1990s. Here's the standardization timeline:

Speed IEEE Standard Year
1 Gb/s 802.3z / 802.3ab 1998/1999
10 Gb/s 802.3ae 2002
40/100 Gb/s 802.3ba 2010
200/400 Gb/s 802.3bs 2017
800 Gb/s 802.3df 2024

Ethernet speed standards evolution timeline from 1998 to 2024

What Is the Highest Speed of Ethernet?

800G is now the highest commercially deployed standard. Dell'Oro reported that 800 Gb/s ports surpassed 20 million shipped units within three years of launch, and 800G switches now dominate AI back-end network sales. A 1.6 Tb/s standard is under development through the IEEE P802.3dj task force, but it's not yet commercially deployed.

Is Gigabit Ethernet the Same as Ethernet?

No. Gigabit Ethernet (1G) is one speed tier within the broader Ethernet standard family, which now spans from 10 Mb/s up through 800G and beyond. Ethernet is the umbrella technology; Gigabit is just one rung on the ladder.

Is a 1G Ethernet Switch Good?

For legacy edge or access use, 1G still works fine for many endpoint connections. But it's nowhere near sufficient for data center core workloads, where 400G and 800G are becoming the baseline.

How 400G Ethernet Works

PAM4 Modulation

Older Ethernet standards relied on NRZ (non-return-to-zero) signaling, transmitting one bit per clock cycle. 400G shifts to PAM4 (4-level pulse amplitude modulation), which encodes two bits per symbol instead of one. This effectively doubles the data rate per lane without doubling the physical infrastructure.

At these speeds, forward error correction (FEC) keeps the link reliable. It catches and corrects transmission errors that PAM4's denser encoding makes more likely.

Form Factors and Lane Configurations

400G optics typically come in two form factors:

  • QSFP-DD — widely supported, backward-compatible with QSFP28 slots
  • OSFP — a separate footprint; Juniper, for instance, doesn't support OSFP in a QSFP-DD port, so the two aren't interchangeable

Common lane configurations include 8x50G and 4x100G, depending on the underlying optical technology.

400GBASE Variants Compared

Variant Max Reach Fiber/Connector
400GBASE-SR8 100 m OM4 multimode, MPO-16 APC
400GBASE-DR4 500 m Parallel single-mode, MPO-12
400GBASE-FR4 2 km Duplex single-mode, duplex LC
400GBASE-LR8 10 km Duplex single-mode, duplex LC

400GBASE variant comparison chart showing reach and fiber connector types

Breakout Capability

A single 400G port isn't locked into one configuration. It can split into 4x100G or 8x50G connections, giving network architects flexibility to match port speed to actual device needs without swapping hardware.

Benefits of Upgrading to 400G Ethernet

Moving to 400G cuts port and cable sprawl, frees rack space, and keeps interconnect capacity ahead of AI, 5G, and high-resolution video demand.

Cost and Density Efficiency

Consolidating four 100G links into a single 400G port reduces the physical ports, cables, and transceivers needed for the same aggregate throughput. Fewer components means:

  • Lower per-port overhead
  • Simplified cabling and rack space usage
  • Easier scale-out as networks grow

Bandwidth for Demanding Workloads

400G, and increasingly 800G, is becoming the default for:

  • AI training clusters, where node interconnects are the bottleneck
  • 5G backhaul networks
  • 4K/8K video streaming infrastructure

Strong Market Momentum

The Ethernet optics market overall is expected to grow 57% in 2024 and 62% in 2025, according to LightCounting's September 2024 forecast, spanning categories from 100G through 1.6T. That figure is not 400G-only, but it shows how fast the full high-speed optics category is scaling.

Ethernet optics market growth forecast for 2024 and 2025 percentages

Choosing 400G Switches and Hardware: New vs. Refurbished

Before picking hardware, place a 400G switch in the network stack:

  • Layer 1 (physical): Handles raw signal transmission: cabling, optics, electrical interfaces
  • Layer 2 (data link): MAC-layer switching within a local network segment
  • Layer 3 (network): Routing-capable switches that make forwarding decisions across subnets

Most 400G core and spine deployments use Layer 2/3 capable platforms.

The OEM Landscape

Major vendors now ship 400G-capable platforms:

  • Arista: 7800R3, 7500R3, 7388X5, and other 7000-series systems with 400G QSFP-DD/OSFP support
  • Juniper: QFX5130 series, offering 32 400GbE QSFP-DD ports in a 1U form factor
  • Cisco: Nexus 9300-GX2 series, with configurations up to 64 x 400G ports
  • Extreme: 7830 series, combining 100G and 400G ports in a single chassis

New units from these vendors deliver strong performance, but upfront costs climb quickly when you refresh an entire spine layer. Refurbished 400G gear from the same OEMs lowers that capital outlay when each unit is tested and warranted properly.

Where Refurbished Fits In

At Terabit Systems, we source and sell refurbished networking hardware from Arista, Juniper, Cisco, and other major brands, including 400G-capable systems like the Arista DCS-7060PX5 with OSFP configurations.

Every unit goes through our five-step quality control process:

  1. Physical inspection for connector and chassis damage
  2. Functional testing, including wire-speed traffic and optical transceiver checks
  3. Factory reset and firmware verification
  4. Cosmetic refurbishment
  5. Final QC and kitting before shipping

Five-step refurbished network hardware quality control inspection process

Each unit ships with a one-year replacement or refund warranty.

If you're retiring older 100G gear during the upgrade, our Sell Your Gear program buys back used equipment from Arista, Juniper, Extreme, and Cisco so you can recover value while you scale.

Frequently Asked Questions

What is a 400G switch?

A 400G switch is a network switch built with ports and ASICs capable of handling 400 Gbps connections. These are typically deployed in data center spine or core layers to handle heavy aggregate traffic loads.

What is the highest speed of Ethernet?

800G is the highest commercially deployed Ethernet standard today, with widespread adoption in AI back-end networks. A 1.6 Tbps standard is in active development but not yet commercially shipping.

What is the fastest Ethernet switch available?

Current top-tier platforms include Arista's 7800R4, supporting up to 576 800GbE ports, and Cisco's Nexus 9300-GX2 series.

Is a 1G Ethernet switch good?

1G switches remain fine for legacy edge and access-layer connections. They're inadequate, however, for modern data center core workloads that demand much higher throughput.

What are L1, L2, and L3 switches?

  • Layer 1: Physical signal transmission
  • Layer 2: Forwards traffic by MAC address within a local segment
  • Layer 3: Adds routing and forwards traffic across different subnets

Is Gigabit Ethernet the same as Ethernet?

No. Gigabit Ethernet is one specific speed tier (1G) within the broader Ethernet standard family, which spans everything from early 10 Mbps standards up through today's 800G deployments.