Brother
Professional
- Messages
- 2,590
- Reaction score
- 500
- Points
- 83
Analysts predict a rapid increase in port throughput to 1.6 Tbit/s.
The market for switches for data centers (DPCs) will grow by 50% due to the growing demand for artificial intelligence. Dell'Oro analyst Sameh Bujelbene predicts that demand for AI will double the size of this market by 2027.
Currently, the share of AI in the total network switch market is less than 10%. At the same time, about 90% of AI deployments use InfiniBand technology, rather than Ethernet. Thanks to this, Nvidia took the second place in the switching market after Cisco.
In the AI segment, bandwidth and latency are key factors. InfiniBand is characterized by low latency and packet loss compared to Ethernet. And this is critical for expensive and lengthy AI training processes.
However, InfiniBand is inferior to Ethernet in terms of maximum bandwidth. Nvidia's latest Quantum InfiniBand switches offer 25.6 Tbps bandwidth and 400 Gbps ports. And Ethernet has already reached 51.2 Tbps and 800 Gbps, respectively.
At the same time, a regular server does not need such high-speed ports. But AI servers are equipped with a 400-gigabit network adapter for each GPU, and the GPU can be 4-8 pieces. Such AI nodes fill a huge bandwidth.
To compete with InfiniBand, Ethernet has even introduced packet loss and latency management technologies. This will allow Ethernet to take 20% of the switching market for AI by 2027, according to Dell'Oro forecasts. Hyperscale data centers and cloud providers will be the main drivers of growth.
By 2025, most switches deployed for AI will have 800 Gbps ports. And by 2027, their capacity may reach up to 1600 Gbit/s. Such high speeds will be possible thanks to the integration of PCIe switching mechanisms into network adapters.
So while InfiniBand is likely to maintain its market leadership in AI switching, the growing demand for AI technologies will only drive growth and innovation in other industry solutions. All this, one way or another, will lead to significant progress and rapid development in the field of AI and machine learning.
The market for switches for data centers (DPCs) will grow by 50% due to the growing demand for artificial intelligence. Dell'Oro analyst Sameh Bujelbene predicts that demand for AI will double the size of this market by 2027.
Currently, the share of AI in the total network switch market is less than 10%. At the same time, about 90% of AI deployments use InfiniBand technology, rather than Ethernet. Thanks to this, Nvidia took the second place in the switching market after Cisco.
In the AI segment, bandwidth and latency are key factors. InfiniBand is characterized by low latency and packet loss compared to Ethernet. And this is critical for expensive and lengthy AI training processes.
However, InfiniBand is inferior to Ethernet in terms of maximum bandwidth. Nvidia's latest Quantum InfiniBand switches offer 25.6 Tbps bandwidth and 400 Gbps ports. And Ethernet has already reached 51.2 Tbps and 800 Gbps, respectively.
At the same time, a regular server does not need such high-speed ports. But AI servers are equipped with a 400-gigabit network adapter for each GPU, and the GPU can be 4-8 pieces. Such AI nodes fill a huge bandwidth.
To compete with InfiniBand, Ethernet has even introduced packet loss and latency management technologies. This will allow Ethernet to take 20% of the switching market for AI by 2027, according to Dell'Oro forecasts. Hyperscale data centers and cloud providers will be the main drivers of growth.
By 2025, most switches deployed for AI will have 800 Gbps ports. And by 2027, their capacity may reach up to 1600 Gbit/s. Such high speeds will be possible thanks to the integration of PCIe switching mechanisms into network adapters.
So while InfiniBand is likely to maintain its market leadership in AI switching, the growing demand for AI technologies will only drive growth and innovation in other industry solutions. All this, one way or another, will lead to significant progress and rapid development in the field of AI and machine learning.
