Nvidia and co-investors inject $160 million into Applied Digital, a leading GPU cloud provider.

Nvidia's $30 billion in Q2 revenues from AI-powered GPUs underscore the critical importance of datacenters to the artificial intelligence boom. With GPU capacity in short supply, venture capital and chipmakers are investing billions to ensure the infrastructure can support continued growth. The latest example includes a $160 million investment by Nvidia and partners in Dallas, Texas-based Applied Digital, a bit-barn operator offering datacenter and cloud services built around Nvidia's GPUs. The company will use the injection to accelerate development of a datacenter complex in North Dakota and support additional debt financing for the costly accelerators. With bleeding-edge GPUs—particularly Nvidia's upcoming Blackwell chips—commanding $30,000 to $40,000 each, many datacenter operators have begun using them as collateral to secure massive loans.

Applied Digital is far from alone in seeking capital. In July, AI datacenter outfit CyrusOne secured $7.9 billion in loans to equip its facilities with the latest accelerators, on top of the $1.8 billion in capital it raised that spring. CoreWeave, arguably the biggest name in the rent-a-GPU market, closed a $1.1 billion Series-C funding round in May, only to raise another $7.5 billion in debt financing weeks later. The funding extends beyond the largest players: AI cloud upstart Foundry picked up $80 million in Series-A and seed funding ahead of its August launch, while Groq, whose inference cloud uses custom language processing units rather than off-the-shelf GPUs, secured $640 million last month to expand its offering.

Lambda exemplifies the broader strategy emerging across the industry. The GPU-cloud operator started the year with a $320 million funding round and secured another $500 million in loans that spring, with plans to add tens of thousands of Nvidia GPUs to its compute clusters. Other operators are pursuing similar paths: TensorWave is scaling compute clusters around AMD's MI300X accelerators, while Voltage Park follows Lambda's lead with Nvidia GPUs. Major venture firms including BlackRock, Magnetar Capital, and Coatue have backed these endeavors, alongside Nvidia, which previously financed CoreWeave.

Nvidia's motivation is straightforward: it can only sell as many GPUs as datacenter capacity allows. Once deployed, each accelerator can generate $1 per hour in subscription revenues from Nvidia's Enterprise AI suite, a figure that compounds dramatically in clusters with 20,000 or more GPUs. The economics work for datacenter operators and their financiers as well, provided revenues cover loan payments. According to The Next Platform, a $1.5 billion investment to build, deploy, and network a cluster of roughly 16,000 H100s would generate approximately $5.27 billion in revenues within four years—a compelling case for why it's currently an excellent time to be in the datacenter business.