Nvidia has reportedly splashed some $900 million to snap up staff from AI networking hardware company Enfabrica, including CEO Rochan Sankar.

CNBC reports that the chip giant agreed to pay a combination of cash and stock, with the deal also encompassing a licensing agreement.

Founded in 2019, Enfabrica offers interconnect solutions it claims can connect hundreds of thousands of GPUs.

The firm also offers networking chips for AI, with its ACF-S line designed to resolve bottleneck issues by ensuring data is fed through networks at a sufficient speed to prevent idling – potentially significantly boosting the efficiency of Nvidia's hardware.

Enfabrica is well-known to Nvidia, having previously joined backers in its $125 million venture capital round in 2023. Nvidia was a strategic investor in the Series B round, with other notable names joining a later $115 million round, including Arm, Cisco, and Samsung.

Prior to founding Enfabrica, Sankar was an organizational leader for Broadcom's cloud and data center switch silicon business, where he helped to build the company's switching and routing product portfolio. He previously worked at Cypress Semiconductor and Cswitch.

Should he and his team join Nvidia, they’d follow in the footsteps of Mellanox staff who joined the chip giant following a $6.9 billion acquisition in 2019.

While the reported Enfabrica deal isn’t an acquisition, it comes as Nvidia is continuing to scale out its networking segment, which recently recorded its highest-ever revenue to date, reaching $7.3 billion.

Nvidia's networking business has been buoyed by demands for its Ethernet-based AI fabric, Spectrum-X, as well as more traditional InfiniBand and NVLink offerings. To further complement its offerings, the firm recently released Spectrum-XGS, a networking software platform that builds on Spectrum-X to connect multiple data centers.

Nvidia has not commented publicly on the reported Enfabrica investment.

Beyond the reported deal with Enfabrica, Nvidia announced this week a $5 billion investment in Intel, striking an agreement with the embattled chip firm to jointly develop custom data center and PC hardware.

In an agreement with far-reaching implications for the industry, the pair plan to connect their respective architectures via Nvidia’s NVLink technology to support custom x86 CPUs that will integrate into Nvidia AI infrastructure platforms.