- Nvidias Quantum-X and Spectrum-X-Switches Reduces the Dependence of Traditional Optical Transceivers
- Silicon Photonics increases Nvidia’s network efficiency by 3.5x
- Quantum-X and Spectrum-X Switches supplies up to 400 Tbps flow to hyperscale AI factories
Nvidia promotes its networking technology by integrating co-packed optics (CPO) into its Quantum Infiniband and Spectrum Ethernet Switch, a step that is expected to reduce power consumption and costs in AI data centers.
At its GTC 2025 event, Nvidia detailed its plans for the implementation of silicon photonics, which will increase the effectiveness of reducing the need for traditional optical transceivers.
Instead of relying on traditional pluggable transceivers, Nvidia Fotonik directly embedded in Switch Asics, cuts energy consumption and minimizes signal loss. These progress benefits Hyperscale AI and could also improve small business routers with similar efficiency gains.
NVIDIA seeks to cut AI -Datacenter power by over 50%.
Nvidia’s Spectrum-X and Quantum-X switches use silicon photonics to supply higher bandwidth and lower energy consumption, supporting up to 1.6 terabits per day. Second (TBPS) per Port to effectively connect millions of GPUs.
The Quantum-X and Spectrum-X photonic switches offer configurations ranging from 128 gates at 800 Gbps to 512 ports at 800 Gbps, providing total flow of up to 400 Tbps.
While the Spectrum-X Ethernet platform improves multi-tenant hyperscale implementations, Quantum-X Infiniband-Switches delivers superior signal integrity and resilience, making them challengers to the best network switch.
“AI factories are a new class of extreme scale data centers, and network infrastructure needs to be reinvented to keep up with integrating silicon photonics directly into Switches, crushed Nvidia The old restrictions of hyperscale and corporate networks and open the gate to the million GPU AI factory,” Jensen Huang said, founder and CEO of Nvidia.
Nvidia’s new switches improve energy efficiency by a factor of 3.5 while reducing signal degradation. In a typical AI Data Center with 400,000 GPUs, conventional network setups require millions of optical transceivers that consume significant power.
Nvidia’s approach reduces the overall network power from 72 megawatts to 21.6 megawatts, which dramatically improves sustainability. These gains can also improve business mounting, enabling faster and more reliable connection.
While Nvidia is switching toward optical network, Copper remains relevant in specific configurations.
Systems such as GB200 NVL72 still use thousands of copper cables to connect GPUs and CPUs via NVLink 5, offering lower power consumption at stand level.
However, As Nvidia progresses to NVLink 6, Copper’s limitations will become more clear, strengthening the need for photonic resolutions in large AI tools.
Nvidia’s new switches are set to release at the end of 202 and 2026. The first model, Quantum 3450-LD Infiniband Switch, which is launched at the end of 2025, provides 144 ports of 800 GB/SEC connection and a total bandwidth of 115 TB/sec.
In 2026, Spectrum SN6810 debuts ethernet -Switch with 128 gates at 800 GB/SEK and a total bandwidth of 102.4 TB/sec. A larger spectrum SN6800 model also arrives in 2026 with 512 gates of 800 GB/sec and a total flow of 409.6 TB/sec.
Via NextPlatform
