Jensen Huang, CEO of NVIDIA, criticized the annual planning cycles of hyperscalers as fundamentally inadequate in the AI era. He stated that companies like Amazon, Microsoft, Google, and Meta plan their infrastructure once a year, while the AI market moves so fast that these plans are rarely accurate. Hyperscaler capital expenditures are projected to reach approximately $800 billion in 2026, then $1.3 trillion in 2027, representing 63% growth in a single year. Huang proposes « neoclouds » as an alternative: smaller, regionally focused, AI-native cloud providers built for speed rather than scale. He suggests the market might eventually need between 50 and 1,000 of these neoclouds to create a distributed network of AI capacity.
Source: Read the original article
