Xsight Labs Raises $300M at $2.8B for AI Networking

An Israeli semiconductor company most people outside the networking industry have never heard of just closed one of the larger AI infrastructure funding rounds of the year, built entirely around a problem that rarely makes headlines: moving data fast enough between AI chips to keep them from sitting idle. Xsight Labs, a fabless semiconductor company based in Tel Aviv, closed a funding round exceeding $300 million, establishing a $2.8 billion post-money valuation, according to the company's announcement of the raise. Fidelity Management & Research Company led the round.

The scale of the valuation jump is genuinely striking. According to Calcalist's reporting on the raise, the new valuation represents more than five times the company's previous mark, driven by growing demand for advanced connectivity infrastructure needed to support AI and cloud computing at scale.

Founded in 2017, Xsight Labs builds high-performance connectivity chips designed to move massive amounts of data efficiently between servers, memory systems, and networks inside advanced AI data centers, a bottleneck that becomes increasingly critical as AI workloads scale. The company's E1 Data Processing Unit processes millions of connections per second at full 800 gigabits-per-second line rate with zero packet drops, and became the first 800G DPU to ship commercially, according to Pulse2's reporting on the technical milestones behind the raise.

The company's E1 chip and its companion X2 Ethernet switch have already been selected by multiple global network operators, including SpaceX's Starlink satellite network, and the technology is currently under evaluation at Tier-1 hyperscalers, according to Calcalist's reporting. Xsight Labs CEO Yossi Meyouhas framed the valuation as validation of that commercial traction: "This valuation is a testament to the team's relentless execution and reflects the market need for a high-performance alternative to closed, legacy architectures."

A genuinely notable technical detail is the X2 switch's software-defined design. Its data plane can be reprogrammed after deployment through Xsight's own instruction set architecture, letting network operators add new protocols and features through software updates rather than waiting for entirely new physical chip generations, while operating at under 200 watts with a 40% lower power profile than legacy switches of comparable capacity, according to PR Newswire's technical breakdown of the X2's specifications. That flexibility matters directly for AI workloads specifically, since model architectures continue evolving rapidly, and Xsight's approach avoids locking hyperscalers into hardware tuned for last year's dominant AI architecture, a dynamic worth understanding alongside our broader coverage of AMD's own Helios platform and the growing wave of specialized AI infrastructure challenging Nvidia's dominance from multiple angles simultaneously.

Xsight Labs' rapid valuation growth adds to a genuinely notable pattern of specialized AI infrastructure companies attracting massive capital by solving specific bottlenecks within the broader AI compute pipeline, rather than competing head-on with Nvidia's general-purpose GPU dominance. For businesses evaluating cloud and AI infrastructure vendors, the networking and data movement layer is becoming just as critical a differentiator as raw compute power, particularly as AI workloads scale toward increasingly large, distributed training and inference clusters.

Israeli semiconductor company Xsight Labs raised more than $300 million at a $2.8 billion valuation, more than five times its previous mark, to scale connectivity chips that move data between AI chips inside data centers. The company's E1 processor became the first commercial 800G data processing unit and has already been selected by network operators including SpaceX's Starlink. The funding reflects growing investor confidence in specialized AI networking infrastructure as a distinct, highly fundable category separate from general-purpose GPU competition.