CoreWeave Secures $2 Billion from NVIDIA to Boost AI Computing Infrastructure Expansion

January 26, 2026 lands like a bass drop in a dark room. CoreWeave, born August 15, 2017 and headquartered in Livingston, New Jersey, just took a $2 billion strategic equity investment from NVIDIA. Not a press release moment. A signal. NVIDIA paid $87.20 per share for roughly 23 million Class A shares, pushing its ownership to 6.562 percent and implying a valuation north of $33 billion. This is not money chasing a story. This is silicon backing steel.

Michael Intrator did not start out trying to build the spine of the AI economy. Neither did Brian Venturo or Brannin McBee. They were commodities traders watching GPUs pay themselves off in days while Ethereum ran hot. When Hudson Ridge Asset Management underperformed the S&P 500, the hedge fund wound down and Atlantic Crypto spun up in a New Jersey garage. By 2019, crypto cooled, AI demand caught fire, and the company that would become CoreWeave found its lane. GPUs were no longer lottery tickets. They were infrastructure.

CoreWeave is not a cloud for everyone. It is a cloud for people who break clouds. Twenty eight data centers live today, thirty eight planned by the end of 2025, more than 850 megawatts of active power, and a roadmap to 5 gigawatts by 2030. The platform is purpose built for AI training and inference, running NVIDIA hardware at industrial scale with reported 96 percent node goodput. SUNK and Mission Control are not cute names. They are the reason trillion parameter models show up on time.

The numbers read like a dare. Q3 2025 revenue hit $1.364 billion, up 134 percent year over year. Full year 2024 cleared $2 billion after growing 737 percent from 2023. Backlog stands at $55.6 billion. OpenAI, Meta Platforms, Microsoft, and NVIDIA account for more than $52 billion in committed contracts. Capital expenditure runs $12 to $14 billion a year because power, land, and steel do not wait for permission.

This is why NVIDIA stepped in again after $100 million in 2023 and $250 million at the IPO. Chips are not the bottleneck anymore. Dirt and electrons are. Jensen Huang understands that AI factories need addresses, not just wafers. This investment accelerates access to power, real estate, and build speed, and it locks CoreWeave deeper into NVIDIA’s reference architecture future.

Michael Intrator, Brian Venturo, Brannin McBee, and Peter Salanki built this by treating compute like a commodity market and infrastructure like a discipline. The question now is not whether demand exists. It is who can deliver when the music speeds up again.