The chips that wire GPUs together are the new prize. Xsight raised $300m to sell an open one.

The business move

Xsight Labs, an Israeli chipmaker, announced raising over $300 million in funding at a $2.8 billion valuation, led by Fidelity. The company designs the often-overlooked networking chips that link thousands of GPUs together, crucial for large-scale AI computations. This funding round values Xsight at more than five times its previous worth, marking significant investor confidence in open networking hardware as AI infrastructure demands grow.

Why it matters

GPUs have dominated the AI conversation, but their performance depends heavily on the chips that connect them. These networking chips handle the data traffic between GPUs to maintain fast, efficient communication. As AI model sizes and complexity skyrocket, the need for scalable, open, and interoperable networking hardware intensifies. Xsight’s approach of selling an open networking chip challenges big players like Nvidia, which tend to tightly integrate their own proprietary interconnect technologies. This push could pressure the AI hardware supply chain to prioritize flexibility, reduce costs, and accelerate AI service deployment.

Who gains and who gets squeezed

AI builders and cloud operators stand to benefit from more modular, open networking chips that promise lower lock-in and potentially better pricing. Investors backing Xsight are betting on networking chips becoming as strategically valuable as GPUs themselves. Meanwhile, Nvidia and other vertically integrated GPU vendors face pressure to either open their interconnect ecosystems or risk losing market share where open standards take hold. Suppliers of proprietary networking gear may see shrinking influence if open chips gain momentum.

What to watch next

Watch how Nvidia and other GPU makers respond to Xsight’s open networking solution. The pace of adoption by hyperscalers and cloud providers will also signal whether open networking chips become the new default for AI infrastructure. Funding activity in the chip interconnect space will indicate investor appetite for alternatives to incumbent vendors. Finally, monitor product releases and performance benchmarks to see if open networking chips can match or beat proprietary options on cost, speed, and scalability.

AI Quick Briefs Editorial Desk