New business model: Nvidia's deals with data centre operators temporarily suspended

Nvidia has temporarily put its new business model plans on hold, according to reports from German media outlet heise online. The company had aimed to support smaller AI infrastructure operators financially in exchange for a share of their revenue. However, this arrangement would have imposed additional obligations on the partners, leading some Nvidia employees to find it too intrusive. The company, known for manufacturing AI accelerators, sought to leverage its substantial financial resources to equip growing global AI infrastructures with its GPUs. While demand for these chips remains high, the challenge for potential buyers, especially smaller ones, is often the cost. Equipping data centers with such technology can require billions in investment. To address this, Nvidia launched a special program offering long-term partnerships. Under this plan, customers were supposed to purchase Nvidia GPUs initially, supported by a loan from the company. In return, they would build data centers, rent out the capacity to their own clients, and then allocate a portion of their revenue to Nvidia over a pre-set period. Additionally, Nvidia aimed to mitigate the risk of unmet customer demand by agreeing to lease the data centers itself if no other buyers emerged. Nvidia had planned to allocate $36 billion for these initiatives, with each partnership typically lasting six years. According to a statement by CFO Colette Kress during a shareholder meeting, as reported by the Wall Street Journal, the program was designed to cover essential costs including chip amortization, fixed data center expenses, and personnel costs. Partners were expected to charge around $1 per hour per GPU to their customers, which would help offset these costs. However, the minimum rental price had to be negotiated beforehand with Nvidia, raising concerns among some employees who found the approach overly controlling. These concerns led to the suspension of preparations for the program, though the exact reasons remain unclear. This move is not the only expansion in Nvidia’s business strategy. Recently, the company acquired the AI platform Hugging Face for $1.29 billion. Hugging Face functions similarly to GitHub, serving as a hub for developers and companies to publish, discover, and refine AI models and datasets, playing a crucial role in the AI industry. Furthermore, it was recently revealed that Nvidia holds stakes in both SpaceX and Intel. The company invested approximately $21 billion in SpaceX and around $30 billion in Intel, making them direct customers of Nvidia's AI accelerators. In terms of stock holdings, Nvidia owns about 123 million of the roughly 640 million shares of SpaceX currently available, and approximately 214 million of the roughly 5 billion shares of Intel in circulation. These figures come from a recently published document by the U.S. Securities and Exchange Commission.