Salem Radio Network News Monday, November 17, 2025

Science

Celero raises $140 million from Alphabet’s CapitalG fund, others for AI networking chip 

Carbonatix Pre-Player Loader

Audio By Carbonatix

By Stephen Nellis

SAN FRANCISCO (Reuters) -Celero Communications, a U.S. startup founded by networking industry veterans, on Monday said it has raised $140 million to build a chip that will help link artificial intelligence data centers across vast distances.

The Irvine, California-based firm is designing a chip that will translate the bursts of light sent over fiber-optic cables between data centers into the electrical ones and zeros used in computing networks inside the data centers. Long-distance connections between data centers have become an area of renewed investment in the chip industry as firms such as Alphabet’s Google and Meta Platforms seek to build massive data center campuses near available power sources.

Celero founders Nariman Yousefi and Oscar Agazzi were formerly senior executives at Marvell Technology, one of the major suppliers of chips that connect data centers. Celero plans to use new algorithms that will be embedded in its chips to process light sent over existing fiber-optic cables in the ground faster and using less power.

“Every time the customer sends a pulse of light, you want to send more information,” Yousefi told Reuters in an interview. “That’s how you get the efficiency over the fiber optics.”

The funding announced on Monday includes a $100 million Series B round led by CapitalG, Alphabet’s independent growth fund, and a previously undisclosed earlier $40 million round led by Sutter Hill Ventures with participation from Valor Equity Partners, Atreides Management, Maverick Silicon and others.

James Luo, a general partner at CapitalG and now a member of the Celero board of directors, said the investment was attractive because existing chips for linking data centers on a campus have trouble stretching hundreds of miles, while telecommunications chips that can reach that far remain too expensive and power-hungry for wide use in AI, where networking chips compete for dollars and power with the rest of the chips in the systems.

“You’re kind of stuck now with a very big gap in the middle,” Luo told Reuters. “It’s clear there’s a new class of technology emerging.”

(Reporting by Stephen Nellis in San Francisco; Editing by Paul Simao)

Previous
Next
The Media Line News
Salem Media, our partners, and affiliates use cookies and similar technologies to enhance your browsing experience, analyze site traffic, personalize site content, and deliver relevant video recommendations. By using this website and continuing to navigate, you consent to our use of such technologies and the sharing of video viewing activity with third-party partners in accordance with the Video Privacy Protection Act and other privacy laws. Privacy Policy
OK
X CLOSE