Bay Street Wire
Tech & BusinessOpinion

The Invisible AI Race: Why 3.2T Interconnects Matter More Than Your Phone Plan

Portrait of Sam Whitfield
Sam Whitfieldtelecom & connectivitySep 8AI
Part of the storyline: Toronto's AI Buildout
The Invisible AI Race: Why 3.2T Interconnects Matter More Than Your Phone Plan

AI-generated image · Bay Street Wire

While the Big 3 push overpriced consumer upgrades, a $275 million bet on Celero Communications reveals where the real connectivity battle is being fought.

OPINION: Every time I look at my monthly telecom bill, I'm reminded of the gap between what the Big 3 sell us and where the actual innovation is happening. The major carriers love to brag about network upgrades to justify their premium pricing, but if you want to see where the real AI infrastructure race is being run, stop looking at your handset and start looking at the hardware layer.

As first reported by the Financial Post, Celero Communications has raised $275 million in Series C funding. This round, led by CapitalG, Valor Equity Partners, and Atreides Management—with participation from Maverick Silicon and Sutter Hill Ventures—brings total capital raised to $415 million and pushes the company's valuation above $3 billion. As part of the deal, Gavin Baker, the Chief Investment Officer and Managing Partner at Atreides Management, joins the board of directors.

Why does a hardware play in Irvine, California, matter? Because the bottleneck for the AI revolution is connectivity. The Financial Post reports that Celero has validated the industry's first 2nm coherent DSP silicon, establishing a scalable path from 1.6T to next-generation 3.2T AI optical interconnects.

Nariman Yousefi, the CEO and co-founder of Celero Communications, noted that traditional optical interconnect architectures are being rapidly outgrown by AI infrastructure. The 2nm process is designed to provide the power efficiency and bandwidth density necessary for hyperscale AI clusters, bypassing the practical limits of traditional PAM4 optical architectures.

The scale of this investment is staggering. The Financial Post cites McKinsey & Company, projecting that global data centers will require nearly $7 trillion in investment by 2030, with roughly $5.2 trillion supporting AI workloads.

While the Big 3 keep us locked into expensive plans with incremental speed bumps, the real heavy lifting is happening in the raw plumbing of the internet. The race to 3.2T interconnects is what makes the AI era possible, proving that the most critical advancements are happening far away from the retail stores.

Sources

More from Sam Whitfield