Nvidia just made its biggest talent grab yet, shelling out over $900 million to hire Enfabrica CEO Rochan Sankar and license the AI startup's networking technology. The deal, which closed last week according to sources familiar with the arrangement, positions the chip giant to connect massive GPU clusters - potentially over 100,000 units - as a single supercomputer.
Nvidia just rewrote the playbook for AI talent acquisitions. The chip giant's $900+ million deal to bring aboard Enfabrica CEO Rochan Sankar and license the startup's networking technology represents the company's most aggressive move yet to control the entire AI infrastructure stack.
The transaction, which closed last week according to two sources familiar with the arrangement, follows the acquihire playbook pioneered by Meta and Google. But Nvidia's price tag dwarfs most previous deals - except for Meta's jaw-dropping $14.3 billion grab for Scale AI founder Alexandr Wang in June.
"This isn't just about talent - it's about building the nervous system for AI," one industry insider told us. Enfabrica's core technology solves a critical bottleneck: how to network tens of thousands of GPUs so they function as a unified supercomputer rather than isolated processors.
Founded in 2019, Enfabrica claims its networking chips can connect more than 100,000 GPUs together seamlessly. That's exactly what Nvidia needs as its hardware evolves from single processors to massive integrated systems. The company's latest products already come in towering racks with 72 GPUs working in concert - the kind powering Microsoft's new $4 billion Wisconsin data center announced Thursday.
The timing isn't coincidental. Nvidia has dominated AI training with its H100 and upcoming Blackwell chips, but connecting thousands of these processors efficiently remains a major challenge for cloud providers building AI infrastructure. Amazon, Meta, and Google are all racing to build 100,000+ GPU clusters for their next-generation AI models.
Nvidia previously invested in Enfabrica as part of a $125 million Series B round in 2023 led by Atreides Management, giving the chip maker an inside track on the startup's progress. Late last year, Enfabrica raised another $115 million from investors including Spark Capital, Arm, Samsung, and Cisco, reaching a $600 million valuation according to PitchBook data.
The deal represents a strategic shift for Nvidia, which has historically focused on chip design rather than major acquisitions. The company's only previous billion-dollar purchase was Israeli networking company Mellanox for $6.9 billion in 2019 - technology that now powers much of Nvidia's current Blackwell lineup.
"Jensen [Huang] sees the future isn't just selling chips - it's about selling complete AI infrastructure solutions," explains one former Nvidia executive. The Enfabrica acquisition gives Nvidia the networking expertise to bundle its GPUs into turnkey supercomputing systems rather than leaving integration challenges to customers.
This follows the broader industry trend of tech giants bypassing traditional M&A to acquire AI talent through quasi-acquisitions. Beyond Meta's Scale AI deal, Google spent $2.4 billion to bring in Windsurf CEO Varun Mohan and his team in July. Microsoft made similar moves with Inflection, while Amazon grabbed Adept's leadership.
The regulatory advantage is clear - these deals avoid antitrust scrutiny while still capturing the talent and IP that drive AI breakthroughs. Nvidia's failed $40 billion attempt to acquire Arm in 2022 showed the limits of traditional acquisitions in today's regulatory environment.
But Nvidia isn't slowing down its investment spree. On Thursday, the company announced a $5 billion stake in Intel for AI processor collaboration, plus nearly $700 million invested in U.K. data center startup Nscale. The message is unmistakable: Nvidia is building an AI ecosystem, not just selling chips.
For Enfabrica, the deal represents validation of its networking-focused approach to AI infrastructure. CEO Rochan Sankar, who joined Nvidia as part of the transaction, brings deep expertise in high-performance computing networking - exactly what's needed to scale AI training to unprecedented levels.
The ripple effects are already being felt across Silicon Valley. Other networking startups focused on AI infrastructure are likely to see increased investor interest and acquisition approaches. The message from Nvidia is clear: in the race to build AI supercomputers, networking is just as critical as raw processing power.
Nvidia's $900+ million Enfabrica deal signals a fundamental shift in how the AI chip leader views its business. Rather than just selling processors, the company is building integrated infrastructure solutions that can scale to 100,000+ GPUs. As AI models demand ever-larger training clusters, Nvidia's bet on networking technology could prove as crucial as its chip dominance. The question now is whether competitors can match both Nvidia's silicon prowess and its growing ecosystem play.
