Sam Altman announced that OpenAI aims to surpass 1 million GPUs by late 2025, positioning itself ahead of competitors. However, Altman hinted at a more ambitious goal of deploying 100 million GPUs, a feat projected to cost around $3 trillion and present significant challenges in manufacturing, power consumption, and infrastructure. For context, Elon Musk’s xAI operates on approximately 200,000 GPUs, making OpenAI’s goal five times larger than that. OpenAI’s partnerships with Oracle and exploration of Google’s TPU reflect its need to break free from existing cloud limitations while Microsoft’s Azure remains its primary platform. The increasing GPU demand benefits companies like SK Hynix, which expects record growth. Despite this, OpenAI’s rapid expansion raises financial sustainability concerns, particularly with SoftBank reconsidering its investment. Achieving the 100 million GPU target would require breakthroughs in efficiency, manufacturing, and energy infrastructure, making it more of an aspirational vision than a practical roadmap.
Source link
Share
Read more