As AI workloads continue to balloon in size and complexity, the infrastructure that powers them is being stretched to its limits. Enter Celero Communications, a California-based startup that has just raised $140 million to tackle one of the most pressing challenges in AI scaling: how to seamlessly connect massive data centers across vast distances. With AI companies distributing their compute across multiple sites worldwide, the need for ultra-fast, energy-efficient long-range networking has never been greater — and Celero wants to be the company that stitches it all together.
Credits: Communications Today
Lighting the Path: How Celero’s Chip Works
At the core of Celero’s innovation is a coherent digital signal processor (DSP) — but not the kind that sits inside everyday telecom hardware. This chip takes the light pulses traveling through fiber-optic cables and converts them into the electrical signals that AI servers can understand. What makes Celero’s DSP remarkable is its intelligence. Instead of merely passing signals along, it uses advanced on-chip algorithms to interpret more information from every pulse, improving data density while lowering power usage.
This combination of high bandwidth, signal clarity, and energy efficiency could redefine how data centers talk to each other — especially when those centers are hundreds or even thousands of kilometers apart. For AI companies that rely on large, geographically distributed clusters, this is the kind of “glue” that could keep global compute tightly synchronized.
Founders With Deep Domain Expertise
Celero is led by two industry veterans: Nariman Yousefi and Oscar Agazzi, engineers with extensive experience at heavyweight chip and networking organizations like Marvell, Broadcom, Inphi, and ClariPhy. Their background gives Celero instant credibility. More importantly, it equips the team to tackle a notoriously difficult engineering problem: scaling optical interconnects for the era of AI-first data centers.
Their mission is simple but ambitious — fix the long-distance connectivity bottleneck that arises when AI computation outgrows traditional setups. Telecom chips weren’t built for AI, and AI interconnects weren’t built for long distances. Celero wants to bridge that gap with purpose-built silicon.
Backed by Major Investors
Celero’s funding journey is just as impressive as its technology. The startup raised $100 million in its Series B, led by CapitalG, Alphabet’s independent growth fund. This follows an earlier $40 million raised from Sutter Hill Ventures, Valor Equity Partners, Atreides Management, and Maverick Silicon.
CapitalG’s participation — and its new board seat — is a strong signal. It shows just how critical long-distance AI networking is becoming and how strategically important Celero’s technology could be for the next wave of AI scaling.
Why This Matters for the Future of AI
Modern AI infrastructure faces a real tension: as models grow, compute must expand across more servers, more clusters, and more physical locations. But efficient long-range communication remains one of the toughest challenges in the industry.
Traditional data-center interconnects offer speed but not distance. Telecom-grade optical systems offer distance but not the power or cost profile that AI deployments require. Celero hits a rare sweet spot — enabling long-range connectivity that is fast, efficient, and built specifically for AI workloads.
This unlocks new architectural possibilities:
-
More distributed data centers without sacrificing performance
-
Energy-efficient global compute fabrics
-
Faster scaling of next-generation AI models
If successful, Celero could become foundational to how AI is deployed worldwide.
Part of a Bigger Shift Toward AI-First Networking
Celero isn’t alone in rethinking the networking layer. The rise of companies like Enfabrica, which recently raised $115 million for its “SuperNIC” designed to connect massive GPU clusters, shows a clear trend: networking is becoming a first-class citizen in AI infrastructure.
Compute gets the headlines, but connectivity is what unlocks scale.
Credits: SiliconANGLE
What’s Next for Celero
With fresh capital in hand, Celero plans to accelerate R&D, expand its engineering roster, and push its coherent DSPs closer to production. As hyperscalers and AI companies race to build distributed mega-clusters, Celero’s technology is arriving at the perfect moment.
If the company executes its vision, it could supply the invisible backbone that allows the world’s AI systems to operate as one interconnected, high-speed organism.
Celero isn’t just building a faster chip — it’s building the future nervous system of global AI.
