According to Sawyer Merritt, xAI’s new Colossus 3 building, combined with Colossus 1 and 2, brings the company’s total datacenter footprint to approximately 2.5 million square feet. These three sites are set to form an unprecedented nearly 2GW supercomputer, powered by over 1 million GPUs and representing an investment likely exceeding $35 billion (source: @SawyerMerritt). This massive scale positions xAI at the forefront of AI infrastructure, enabling accelerated model training, advanced inference capabilities, and expanding opportunities for enterprise AI services. The deployment highlights a growing trend in hyperscale AI datacenter development, reflecting escalating demand for compute power in generative AI, autonomous systems, and large language models.

According to Anthropic (@AnthropicAI), the company has announced plans to expand its use of Google TPUs, securing approximately one million TPUs and more than a gigawatt of capacity for 2026. This large-scale investment aims to significantly boost Anthropic's AI model training and deployment capabilities, positioning the company to scale up its advanced AI systems and support enterprise demand. This move highlights the accelerating trend of hyperscale AI infrastructure investment and demonstrates the growing importance of robust, energy-efficient hardware for training next-generation foundation models and powering AI-driven business applications (Source: AnthropicAI on Twitter, Oct 23, 2025).