AI Data Center Interconnects: AEC / AOC Cables and Retimers Market; Why Investors are Looking at Credo and Astera Labs

Credo, Astera Labs

Two studies from Google and Microsoft have shown GPUs spend 70% of their time sitting idle!

The faster we can feed data to the GPUs, and the more efficiently we can have them communicate with each other and with other system components, the higher the overall system performance and efficiency. Given GPU accounts for 65% of Server BOM, there is a significant stress on GPU utilization.

This is where interconnects come into play. Interconnects, including switches, cables, retimers, NVSwitch and interface technologies like PCIe and CXL, play a critical role in determining how quickly and efficiently data can move between GPUs and other components in the system.

In this article, we will review:

1. Existing Infrastructure Requires Upgrading to Deploy AI WorkLoads

2. Active Electric Cable (AEC) vs. Active Optical Cables vs. Direct Attach

   1. Compare Cost, Power and Range

   2. Operating Costs of AEC Advantage Over AOC

   3. Players - Credo / Microsoft alliance

   4. Hyperscaler’s adoption of AEC

3. Retimers

   1. What is a Retimer?

   2. Astera labs’s IPO and Products.

Please note: The insights presented in this article are derived from confidential consultations our team has conducted with clients across private equity, hedge funds, startups, and investment banks, facilitated through specialized expert networks. Due to our agreements with these networks, we cannot reveal specific names from these discussions. Therefore, we offer a summarized version of these insights, ensuring valuable content while upholding our confidentiality commitments.

Existing Infrastructure Requires Upgrading to Deploy AI WorkLoads

Cisco’s interviewed 8000+ senior business leaders at 500+ organizations and in its recent AI Readiness index report,

• 95% of respondents recognize that AI will increase infrastructure workloads.

• Over half of the respondents say their infrastructure is scalable only to a moderate or limited extent and requires upgrading for more complex AI technology.

what this means is,