Xsight Labs Raises $300M at $2.8B Valuation to Challenge Nvidia with Programmable Networking Chips

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

The AI boom made GPUs famous. The chips that connect thousands of them are quietly becoming just as valuable, and Nvidia dominates that layer too. Israel's Xsight Labs has raised more than $300m at a $2.8bn valuation to sell an open, programmable alternative. It is already flying on SpaceX's Starlink V3 satellites. The AI boom made GPUs famous. The dull chips that wire thousands of those GPUs together are quietly becoming just as valuable. Xsight Labs, an Israeli chipmaker, has raised more than $300 million at a $2.8 billion valuation, it said on Thursday. Fidelity led the round. That is more than five times what the company was worth in 2021, a jump it made in a single year. Xsight builds the plumbing of the AI data centre. Its two chips, the E1 processor and the X2 switch, shuttle data between servers, memory and networks. As models grow, moving that data has become one of the industry's tightest bottlenecks, and the wiring between GPUs is now a prize of its own. Taking on Nvidia's other monopoly Nvidia does not just sell the GPUs. It also dominates the networking layer, which it bought its way into with Mellanox. Xsight's pitch is the opposite of that closed kit. Its chips are open and programmable. They run standard Linux and SONiC, and operators can rewrite the X2 switch in software after it ships, instead of waiting for new silicon. The market wanted "a high-performance alternative to closed, legacy architectures," said chief executive Yossi Meyouhas. Xsight joins a wave of startups chipping at Nvidia's grip on the AI stack. Already in orbit This is not a lab project. Several global network operators have picked Xsight's silicon, including SpaceX's Starlink. Last week its processors passed a test aboard SpaceX's new Starlink V3 satellites, Globes reported. Programmable chips suit space, where engineers fix a fault in software rather than swap out the hardware. The selling point is efficiency. The E1 is the first 800-gigabit data processor to ship, and runs under 75 watts. The X2 moves 12.8 terabits a second while drawing 40% less power than older switches. As AI's electricity bill balloons, every watt saved is a pitch. A serial exit machine behind it The pedigree helps. Veterans of EZchip, the Israeli networking firm Mellanox bought in 2016, founded Xsight in 2017, CTech reported. Its chairman is Avigdor Willenz, a serial chip entrepreneur who has sold past companies to Marvell, Amazon and Intel. His backers followed him in. The round drew Fidelity, Intel Capital, T. Rowe Price, Battery Ventures and Valor Equity Partners, among others. The cash will fund the next chip generations and a bigger sales push at Tier-1 hyperscalers. The money pouring into AI infrastructure is no longer chasing only GPUs. It is chasing everything around them, from power stations to the wires in the rack. Xsight is betting the winners in networking will be the open ones.

Programmable networking chip startup Xsight Labs raises $300M

Israeli network chip startup Xsight Labs Ltd. said today it has closed on a $300 million funding round that brings its valuation to $2.8 billion. It's planning to use the funds to help its programmable Ethernet switches and 800-gigabit data processing units push deeper into cloud and artificial intelligence networks. Today's round was led by Fidelity Management & Research Co., and saw participation from 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.. Investors are making a fairly straightforward bet on Xsight. They believe AI data center networking will balloon into a $150 billion market by 2028, and Xsight's chips, which prioritize power efficiency and programmability, are positioned to capture a meaningful slice of that growth Xsight, founded in 2017 and based in Tel Aviv, develops two key components that are used in data center networks - the E1 data processing unit or DPU and the X2 Ethernet switch, which are both designed specifically for AI connectivity. The company's calling card is power efficiency, with its chips consuming under 200 watts for a 12.8 terabit-per-second switch. It also emphasizes customer-driven software extensions, making its chips programmable rather than fixed-function. The startup is currently developing new generations of both products, which will introduce support for the latest standards backed by the Ultra Ethernet Consortium. To develop those products, it's planning to expand its engineering teams across its facilities in Israel, the U.S., Europe and Asia, and increase its manufacturing and supply chain capacity so it can take orders from "tier-1 customers." The startup's E1 chip pairs two 400G ports with 64 Arm-based Neoverse N2 cores built on a 5-nanometer process and 32 megabytes of shared cache. It also features PCIe Gen5 connectivity and encryption engines that run at line rate. Where it differs from other DPUs is that those 64 cores sit directly on the data pathway, so that routing., telemetry and packet inspection occur on every data packet instead of being handed off to a smaller exception-handling engine, enhancing overall network bandwidth with four-times the performance per watt of previous-generation designs. The chip has also achieved SONiC-DASH Hero 800G validation, sustaining more than 14 million connections per second with no dropped packets in that benchmark. As for the X2 switch, this provides 12.8 terabytes of full-duplex bandwidth at under 200-watts, plus first-bit-to-first-bit latency of less than 700 nanoseconds, with 128 lanes of 100G PAM4 SerDes. It features a reprogrammable data plane that allows it to be reconfigured for different network applications on the fly. According to Xsight, it draws 40% less power than other kinds of 12.8 Tbps switches. Xsight says the E1's performance-per-watt is its main advantage. 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 all considered overheads. Xsight's most notable customer so far is SpaceX Corp. The company said last year that the X2 has been chosen as the high-speed networking core of Elon Musk's Starlink V3 satellites, which will be capable of moving more than 1 Tbps of fronthaul traffic and around 160 gigabytes-per-second of uplink capacity. The switch was chosen as much for its performance as its ability to operate in radioactive environments and thermal extremes, the company said. Xsight's E1 and X2 have both also been deployed by multiple data center operators globally within edge and network infrastructure, and its products are now being evaluated by a number of 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 co-founder and Chief Executive Yossi Meyouhas. The startup is also hoping to increase its presence in AI data centers, as Ethernet is believed to be closing the gap with the proprietary interconnects that link most AI chip clusters today. Last year, the Ultra Ethernet Consortium published its 1.0 specification, which defined a remote-memory-access transport with multipath routing and congestion control that's designed for AI training and inference traffic instead of general-purpose computing workloads. Xsight says the X2 is one of the first switches to meet the new standard. Incumbents such as Cisco Systems Inc., Broadcom Inc and Marvell Technology Inc. currently dominate large portions of the data center networking market. The question is whether Xsight's promise of lower power consumption and competitive cost per token processing will hold up against those player's deeper customer relationships and broader product portfolios.