Nvidia GPUs are everywhere. Here are the ways companies are accessing them
Nvidia's GPUs are in unprecedented demand, pushing the chipmaker's market cap toward $6 trillion and projected revenues for the October quarter to $108 billion, a staggering 89% increase year-over-year.
Companies are exploring diverse access methods for Nvidia's GPUs, ranging from major cloud providers like Amazon, Microsoft, and Google, to emerging 'neoclouds' such as CoreWeave. Key industry clients including OpenAI and Anthropic are committed to spending over $500 billion on these GPUs. However, tight capacity has led to challenges in availability and pricing, with companies increasingly looking to smaller neoclouds and even borrowing GPUs for short-term projects. Future competition and partnerships, like SpaceX's GPU rental agreements, reflect the critical need for scalable AI infrastructure.
Understanding the dynamics of GPU access is crucial, as it impacts competitive positioning and cost structures in AI-related services. Startups and tech companies need to navigate these options strategically to secure necessary computing power without prohibitive costs.