AI technology has swept in like a storm, making GPUs the hottest commodity worldwide.
In the second quarter of this year, Nvidia's performance was explosive, with revenues exceeding $30 billion! But have you ever thought, without sufficient data centers to house these powerful chips, wouldn't it all be in vain? To meet the sustained demand of this AI boom, Nvidia and its partners have started to increase their investments.
Recently, Nvidia and its partners injected $160 million into Applied Digital, a company based in Texas. This company is not ordinary; they focus on providing various data center and cloud services around Nvidia's GPUs. This funding will help Applied Digital accelerate the construction of data centers in North Dakota and support their further financing plans to cover the high costs of accelerators.
You might wonder, why is the construction of data centers so important? That's because advanced GPUs are as expensive as cars, with Nvidia's upcoming Blackwell chips costing between $30,000 and $40,000 each. Many data center operators have started using GPUs as collateral to borrow huge loans from banks to raise this capital.
Moreover, Applied Digital is not the only successful company in terms of financing. In July this year, AI data center company CyrusOne secured a loan of up to $7.9 billion to support their new projects, even before securing $1.8 billion in capital. Such amounts are staggering! CoreWeave, a giant in the GPU rental market, successfully raised $1.1 billion in May and soon after, collected $7.5 billion in debt from investors.
Of course, besides these giants, there are many emerging startups also vying for this market. AI cloud startup Foundry secured $80 million in Series A funding before officially launching in August. Unique chip manufacturer Groq, with its custom Language Processing Units (LPU), also successfully raised $640 million last month.
Honestly, this is a very hot market, and almost every data center operator wants to hop on the AI bandwagon. New companies like TensorWave and Voltage Park are striving to expand computing clusters based on AMD and Nvidia GPUs, hoping to find their niche in this competitive market.
It's worth noting that Nvidia's willingness to invest in these projects is also very clear. After all, GPU sales are limited by data center capacity. Each time these accelerators are deployed, they can bring Nvidia an hourly subscription income of $1. If you have a cluster of 20,000 GPUs, the revenue is substantial.