Chip giant NVIDIA recently invested over $900 million to recruit Rochan Sankar, CEO of AI hardware startup Enfabrica, and his team, while also licensing the company's technology. Enfabrica, renowned for its technology that efficiently connects over 100,000 GPUs, is expected to strengthen NVIDIA's presence in large-scale AI computing clusters.
What is the background of Enfabrica’s founding and what are its strengths?
Enfabrica was founded in 2019 by Rochan Sankar , a former senior executive at Broadcom, and Shrijeet Mukherjee, who previously worked in technology development at Alphabet.
The two formed a team with 120 core engineers from Intel, Cisco, Meta and other companies to focus on solving problems in AI infrastructure expansion, such as improving data transmission and network computing bottlenecks.
The core of Enfabrica's technology lies in creating a dedicated network chip that connects 100,000 AI computing chips at high speed, allowing the massive chip cluster to operate collaboratively like a supercomputer. This design significantly reduces data transmission delays and chip idleness caused by insufficient network bandwidth, while also delivering higher computing efficiency.
Crucially, Enfabrica enables AI chips to directly connect to lower-cost DDR5 memory, replacing expensive high-bandwidth memory (HBM). This effectively reduces data center memory costs while maintaining system performance. As of 2025, the company has raised approximately US$260 million in funding.
The surge in technology demand has driven a major update of Nvidia's AI products.
As the generative AI craze continues to grow, Nvidia's products have evolved from single GPU cards to entire cabinets of high-density computing systems. Early A100 chips were mostly installed as single processors in servers, but the latest generation has been upgraded to tower-style racks capable of accommodating 72 GPUs per cabinet.
This architecture is the core configuration of Microsoft's recently announced $4 billion data center investment in Wisconsin. Enfabrica's hyperscale tandem technology complements Nvidia's massive system deployment, providing strong support for its continued global leadership in AI computing.
900 million magnesium recruits Enfabrica team, further expanding its global AI layout
According to the latest report from CNBC, Nvidia spent more than $900 million in cash and stock to recruit Rochan Sankar, CEO of AI hardware startup Enfabrica, and core members, and simultaneously obtained the company's technology license. The transaction was completed last week.
This is another major AI layout following the acquisition of Israeli software optimization company Run:ai for US$700 million in 2024, the investment of nearly US$700 million in British data center startup Nscale, and the announcement this week that it would spend US$5 billion to cooperate with Intel on AI processors.
Through the dual reinforcement of technology and talent, Nvidia continues to consolidate and expand its leading position in the global AI hyperscale computing field.
Nvidia invests $5 billion in Intel to jointly develop AI and PC chips
The article Nvidia spends over $900 million to recruit Enfabrica team and technology to strengthen its AI large-scale computing layout first appeared in ABMedia ABMedia .
