Nvidia Is Buying Loyalty: Why Jensen Huang Is Putting $2 Billion Into CoreWeave

Nvidia has put $2 billion into CoreWeave, and in return it gets something money can't measure: a guaranteed proving ground for the Rubin architecture, Blackwell's successor, before it reaches the mass market. According to TechCrunch, CoreWeave will also integrate Nvidia products across its entire platform, including the new Rubin chip architecture.

CoreWeave's path: from

Ethereum mining to the largest AI cloud

CoreWeave started as a modest Ethereum miner and grew into the world's largest specialized AI cloud. That aggressive expansion took enormous amounts of money, and skeptics began questioning the company's debt as early as last year. Nvidia's cash injection sends the market a clear signal: CoreWeave is too important to fail. The company has announced plans to deploy 5 gigawatts of capacity dedicated exclusively to neural network training and inference, a figure that makes both competitors and energy providers nervous. For comparison, that much power could supply electricity to several mid-sized European cities.

A closed loop: how the Nvidia–CoreWeave arrangement works

Nvidia's interest here is pragmatic. Nvidia provides the money, CoreWeave uses it to buy Nvidia chips, and those chips attract customers who pay for access to Nvidia technology. This vertical integration effectively turns CoreWeave into an external division of the company. It lets Jensen Huang control not just production but the actual consumption of compute, bypassing the bureaucracy and whims of hyperscalers like Microsoft or Google.

Rubin: a new chip and a faster release cycle

Rubin is Blackwell's successor, and the market is only beginning to learn the details. Nvidia is shortening its chip release cycle to one year, and it needs sites that can deploy tens of thousands of these devices immediately. CoreWeave is exactly that kind of site. The $2 billion investment guarantees that Rubin won't sit gathering dust in warehouses. Instead, it will start generating data and profit right away, proving Nvidia's edge over AMD and Intel.

What this means for the industry

The barrier to entry into the top tier of AI clouds is becoming unreachable: without a graphics giant with bottomless pockets behind you, competing on capacity is pointless. Nvidia is clearly preparing for an era when training models at the scale of GPT-5 and beyond will require resources that seem fantastical today, and 5 gigawatts is the infrastructure foundation for a world where AI is part of every process. Nvidia isn't just selling shovels during a gold rush; it's buying the mines and hiring the prospectors.

For AI startups, this is good news: there will be more compute. For Nvidia's competitors, it's a warning: they will have to do more than build better hardware. They will also need to find partners willing to build infrastructure at this scale under a heavy debt burden.

How much did Nvidia invest in CoreWeave?

Nvidia invested $2 billion in CoreWeave, the largest specialized AI cloud provider, which grew out of an Ethereum miner. The investment comes with the integration of Nvidia products across CoreWeave's entire platform, including the new Rubin chip architecture.

What is the

Rubin architecture, and why does CoreWeave need it?

Rubin is the successor to the Blackwell architecture and the next generation of Nvidia chips. The company is shortening its chip release cycle to one year, so it needs sites that can quickly deploy tens of thousands of devices. By integrating Rubin into its platform, CoreWeave becomes one of those sites, and the chips get tested under real-world conditions before they become widely available to the rest of the market.

How much capacity does CoreWeave plan to deploy?

CoreWeave has announced plans to deploy 5 gigawatts of capacity dedicated exclusively to neural network training and inference. For comparison, that much power could supply electricity to several mid-sized European cities.