Source: The Next Web
As GPU clusters scale beyond 100,000 units, traditional copper wiring becomes the physical constraint limiting AI training speeds and efficiency—a problem that silicon abundance has made visible. This funding round reflects a hard infrastructure shift: the bottleneck moved from compute to communication, making optical interconnects a suddenly urgent business, not a research curiosity. The investors backing this are hardware makers themselves, signaling that incumbents view optical switching as table-stakes infrastructure for the next generation of AI clusters, not a speculative bet.