Nvidia Conference Outlines Three Major Problems in the AI Market
In short — neoclouds, data centers, and a shortage of funds
Nvidia CEO Jensen Huang concluded his one-and-a-half-week tour of Europe. He met with world leaders, including UK Prime Minister Andy Burnham, French President Emmanuel Macron, and German Chancellor Friedrich Merz, and even extended the trip with a visit to Kazakhstan to discuss the country’s prospects as a new AI hub.
Following meetings with investors, top executives of "neocloud" providers, and data center builders on the sidelines of GTC, observers identified key trends and challenges in the AI infrastructure market.
1. The "neocloud" sector under pressure. The end of the conference coincided with a sell-off of neocloud company stocks: CoreWeave shares fell by 10%, while Nebius dropped by 7%. The market is concerned about the high debt burden on players leasing AI computing power.
2. Data center in a container. To eliminate barriers in AI capacity deployment, Nvidia announced the Vera Rubin racks, packaged directly into standard shipping containers for transport on regular roads.
According to Huang, delays in launching data centers cost the industry billions of dollars in lost revenue. Mobile modules will drastically accelerate the deployment of computing systems. The project has already attracted interest from giants like OpenAI and Anthropic, which have been seeking solutions for more compact data center placement.
3. Capital deficit. Scaling AI infrastructure requires enormous financial investments. According to Goldman Sachs estimates, total capital expenditures on AI from 2026 to 2031 will reach $7.6 trillion.
A single 1 GW data center costs approximately $50 billion.
Infrastructure financing experts confirm that in the near future, the lack of available funds, rather than technology, will become the primary obstacle to industry development.
