Nvidia покупает лояльность: зачем Дженсен Хуанг вливает 2 миллиарда в CoreWeave

When a company with enormous debt receives a check for two billion dollars, it usually smells like either desperation or brilliant calculation. With CoreWeave and Nvidia, we're witnessing the latter. Jensen Huang decided not merely to support his favorite cloud provider, but essentially to reserve a prime spot for the upcoming Rubin architecture. This partnership looks like Nvidia's attempt to create an ideal testing ground for rolling out its most powerful hardware while bypassing the bureaucracy and caprices of hyperscalers like Microsoft or Google.

CoreWeave has traveled a remarkable journey from humble Ethereum miners to the world's largest specialized AI cloud. However, aggressive expansion requires colossal sums, and the company's debts began raising questions from skeptics as far back as last year. Nvidia's injection sends a clear market signal: CoreWeave is too important to fail. The plan to deploy 5 gigawatts of capacity is a figure that makes both competitors and energy companies nervous. For comparison, such power would be enough to supply electricity to several mid-sized European cities, but here all resources will be directed exclusively to training and inference of neural networks.

Nvidia's interest here is pragmatic to the point of cynicism. Deploying the Rubin architecture on CoreWeave's platform means that new chips will be tested in real conditions before they become widely available to the broader market. This creates a closed loop: Nvidia gives money, which CoreWeave uses to buy chips from Nvidia to attract customers who will pay for access to Nvidia technologies. This vertical integration de facto makes CoreWeave an external department of the corporation, allowing Jensen Huang to control not just production, but consumption of computational resources itself.

The Rubin architecture mentioned in the deal deserves separate attention. It is the successor to Blackwell, which we're only beginning to learn details about. Nvidia is accelerating its chip release cycle to one year, and it needs platforms capable of instantly deploying tens of thousands of such devices. CoreWeave becomes exactly such a platform. The 2-billion investment is a guarantee that Rubin won't collect dust in warehouses but will immediately start generating data and profits, proving the company's superiority over AMD and Intel.

What does this mean for the industry as a whole? First, the barrier to entry into the top tier of AI clouds becomes unreachable. Without a graphics giant with bottomless pockets behind you, it's pointless to compete on raw power. Second, Nvidia is clearly preparing for an era when training models at the scale of GPT-5 and above will require resources that seem fantastic today. Five gigawatts is the infrastructure foundation for a world where AI is present in every process.

We're witnessing the formation of a new type of monopoly. Nvidia doesn't just sell shovels during the gold rush—it's buying the mines themselves and hiring the prospectors. For AI startups, this is good news: there will be more capacity. For Nvidia's competitors, it's a warning bell. They'll need to not only make better hardware but also find partners willing to build infrastructure at this scale while bearing heavy debt burdens.

The bottom line: Nvidia is building its own cloud computing empire through CoreWeave's hands, insuring itself against competition from BigTech's cloud divisions. Will AMD be able to find an ally of the same scale?