AI data centers have a bandwidth problem. As models grow larger and computing clusters expand, the pipes moving data around simply can't keep up. Celero Communications thinks it has the solution, and Alphabet (GOOG) just bet $100 million that they're right.

The Irvine, California-based semiconductor startup announced on November 17 that it closed $140 million in total funding to tackle what might be artificial intelligence's most critical infrastructure challenge: how to move massive amounts of data without melting your data center or your budget. CapitalG, Alphabet's independent growth fund, led the $100 million Series B round. The company had previously raised $40 million across seed and Series A rounds led by Sutter Hill Ventures, with participation from Valor Equity Partners, Atreides Management, and Maverick Silicon.

The Bandwidth Bottleneck Problem

Here's the issue: AI systems need to transfer enormous amounts of data between accelerators and across data center networks. Think of it like trying to run a major city's worth of traffic through a two-lane highway. Current optical connectivity approaches are struggling with power consumption, cost efficiency, and bandwidth capacity as AI clusters grow beyond single-rack configurations. According to a symposium paper by Matthew Williams, Chief Technology Officer at composable infrastructure provider Cerio, the old ways of doing things just aren't cutting it anymore.

Celero's founders, Nariman Yousefi and Oscar Agazzi, know this territory well. They previously held senior roles at Marvell Technology (MRVL), Inphi (IPHI), Broadcom (AVGO), and ClariPhy, where they shipped networking solutions and digital signal processor products. Yousefi now serves as CEO of Celero Communications, which opened for business in 2024 with headquarters in California and design facilities in Canada and Argentina.

"As AI models and clusters grow exponentially, today's data centers face severe limits in bandwidth, cost, and energy efficiency," Yousefi said in the company's statement. "Our coherent DSP technology delivers the optical performance and scalability needed to remove these bottlenecks, establishing the foundation for large-scale, accelerated computing networks."

How Coherent DSP Technology Works

Celero's platform handles signal processing for optical connections between AI accelerators and cloud infrastructure. The technology aims to improve bandwidth efficiency while reducing power consumption compared to existing solutions. Translation: it helps data move faster through fiber optic cables while using less energy.

Data centers are transitioning from thousands to millions of AI accelerators, and they need optical connectivity that can maintain performance at that scale. Coherent DSP chips process light signals traveling through fiber optic cables, enabling faster data transmission with lower energy requirements. It's sophisticated technology, but the concept is straightforward: make the pipes bigger and more efficient without burning through electricity.

"Amid the explosive growth of artificial intelligence, Celero is uniquely positioned to meet the soaring demands of AI infrastructure," said James Luo, General Partner at CapitalG. "The experienced leadership and differentiated architecture at Celero position the company to lead in AI-driven network innovation."

Experience Matters in Semiconductor Scaling

The founding team's background includes multiple product cycles in coherent optical technology. Yousefi and Agazzi worked on DSP development and networking chip commercialization at their previous companies, which means they've actually shipped products before. In the semiconductor world, that experience matters tremendously. Building chips is hard. Building chips that work at scale is harder. Building chips that work at scale and can be manufactured economically is harder still.

"Celero's coherent DSP innovation directly addresses one of the most pressing challenges in AI infrastructure: efficient, scalable optical connectivity," said Stefan Dyckerhoff, Managing Director at Sutter Hill Ventures. "The team's track record in building and commercializing advanced coherent solutions is unmatched in the industry, and their focused approach positions Celero to lead the next wave of AI infrastructure."

The company plans to use the capital to support product development and market entry as cloud providers and enterprise operators continue expanding AI computing capacity. With AI workloads showing no signs of slowing down, the demand for better data center infrastructure is only going to intensify. Celero is betting that solving the bandwidth bottleneck problem puts them in the right place at the right time.