Fidelity Management & Research Co. led the financing, with backing from a broad syndicate including Aliya Capital Partners, Atreides Management, Artisan Partners, Battery Ventures, Diagonal Capital, Intel Capital, Key1 Capital, Maverick Capital, Sienna, T. Rowe Price, Union Group, and Valor Equity Partners. The investment underscores a broader industry bet that AI data center networking will balloon into a $150 billion market by 2028. Xsight intends to use the funds to push its programmable Ethernet switches and 800-gigabit data processing units deeper into cloud and AI infrastructure.
Power Efficiency in Programmable Ethernet Switches
Founded in 2017 and headquartered in Tel Aviv, Xsight Labs builds two core components designed specifically for AI connectivity: the E1 data processing unit and the X2 Ethernet switch. The company’s primary differentiator is power efficiency. Its X2 switch provides 12.8 terabits-per-second of full-duplex bandwidth while consuming under 200 watts. Xsight claims this represents a 40% reduction in power draw compared to other 12.8 Tbps switches, with a first-bit-to-first-bit latency of less than 700 nanoseconds.
The X2 features 128 lanes of 100G PAM4 SerDes and a reprogrammable data plane that allows on-the-fly reconfiguration for different network applications. The chip is also one of the first to meet the 1.0 specification from the Ultra Ethernet Consortium. This standard defines a remote-memory-access transport with multipath routing and congestion control tailored specifically for AI training and inference traffic, rather than general-purpose computing workloads. As Ethernet closes the gap with proprietary interconnects used in AI chip clusters, Xsight is positioning itself at the forefront of this open-standards shift.
The 800-Gigabit DPU Advantage
Xsight’s E1 800-gigabit DPU pairs two 400G ports with 64 Arm-based Neoverse N2 cores built on a 5-nanometer process. The chip features 32 megabytes of shared cache, PCIe Gen5 connectivity, and encryption engines that run at line rate. Unlike traditional DPUs that hand off packet inspection to smaller exception-handling engines, Xsight places its 64 cores directly on the data pathway. This design enables routing, telemetry, and packet inspection on every data packet, delivering four times the performance per watt of previous-generation designs.
The E1 has also achieved SONiC-DASH Hero 800G validation, sustaining over 14 million connections per second without dropped packets. In AI data center racks, where accelerator chips consume the bulk of the power budget and cooling capacity, the watts spent on network fabric are considered overhead. Xsight’s performance-per-watt advantage directly addresses this constraint.
SpaceX Deployment and Hyperscaler Traction
The startup’s hardware is already proving itself in extreme environments. SpaceX selected the X2 switch as the high-speed networking core for its Starlink V3 satellites. The chips will handle more than 1 Tbps of fronthaul traffic and roughly 160 gigabytes-per-second of uplink capacity. Xsight noted that the switch was chosen not only for its performance but also for its ability to operate in radioactive environments and thermal extremes.
Beyond low Earth orbit, Xsight’s E1 and X2 chips have been deployed by multiple data center operators globally within edge and network infrastructure. The company is now being evaluated by several Tier-1 hyperscalers. “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,” said Yossi Meyouhas, co-founder and Chief Executive of Xsight Labs. To capture these hyperscaler accounts, the startup will have to leverage its lower power consumption and competitive cost per token processing against the deeper customer relationships and broader portfolios of established giants like Cisco, Broadcom, and Marvell.
What Happens Next
Xsight Labs faces a steep uphill battle against entrenched incumbents that currently dominate the data center networking market. However, as AI clusters scale, the overhead spent on network fabric power becomes a critical bottleneck. If Xsight can deliver on its promises of lower power consumption and competitive cost per token processing, its programmable architecture could disrupt the status quo. Expect the company to use its new $300 million war chest to aggressively expand its engineering footprint across Israel, the U.S., Europe, and Asia, while ramping up manufacturing to secure design wins with major cloud providers transitioning to Ultra Ethernet standards.
— David Kim, technology desk, AXO News