Nvidia's GTC conference in Berlin ended with a product announcement and a stock market setback for AI cloud providers. The company revealed a new design that packs seven Vera Rubin AI racks into a standard shipping container, aiming to speed up data center deployment. The reveal came as shares for neoclouds, companies that lease AI compute resources, fell sharply.
Chief Executive Officer Jensen Huang spent the week meeting with political leaders, including U.K. Prime Minister Andy Burnham, French President Emmanuel Macron, and German Chancellor Friedrich Merz. The trip concluded with a media question-and-answer session where Huang discussed his work habits, stating that he tries to sleep less to bring data centers online faster. Following the event, Huang traveled to Kazakhstan to promote the country's potential as an artificial intelligence hub.
The final day of the conference coincided with a decline in public stocks for neoclouds. CoreWeave shares fell by 10% and Nebius dropped by 7% by market close. This move continued a period of turbulence for the sector, driven by concerns over heavy borrowing. One executive at an AI infrastructure company noted that the scale of businesses without credit ratings was alarming the market.
Despite the stock prices, Huang remained optimistic about the sector's fundamentals. In an interview with CNBC, he said demand for neoclouds is extremely high. He argued that moving compute into the supply chain will increase the speed at which systems can be turned on. Huang warned that delays in getting data centers running can result in billions of dollars in lost revenue.
The high cost of AI infrastructure is a central challenge. Huang told the Berlin audience that a 1 gigawatt data center costs approximately $50 billion. Goldman Sachs estimates that AI-related capital spending could reach about $7.6 trillion between 2026 and 2031, covering compute, data centers, and power. Huang stated that working capital is currently missing from the middle of this growth, creating a scarcity of funds despite high demand. He added that this capital gap will be solved.
An executive at an AI infrastructure financing company said capital will soon become the bottleneck. Meanwhile, major consumers of compute are looking for smaller deployment options. Reports indicate that OpenAI and Anthropic have been seeking smaller AI data center deals to meet their needs.