Summary
In May 2024, reports said xAI planned a 100,000-GPU “Gigafactory of Compute.” Reuters reported a $6 billion funding round at a company value of $24 billion. xAI later said its Colossus system had 200,000 GPUs. Most startups can rent AI computing power instead of building data centers.
Introduction
In 2024, xAI called its proposed computing project the “Gigafactory of Compute.” xAI later called its system Colossus. This article explains what the reports mean for startups.
xAI’s 2024 supercomputer plan
In May 2024, The Information reported xAI planned a supercomputer. The plan called for 100,000 Nvidia H100 graphics processors, or GPUs. The company aimed to have it running by fall 2025. Reuters reported a $6 billion Series B round at a $24 billion company valuation. Those figures came from 2024 reports. They are not current company figures.
xAI’s later Colossus page calls the system its gigafactory of compute. xAI says it reached 200,000 H100 GPUs in February 2025, after doubling its first system in 92 days. The same page’s specifications list 180,000 GPUs. The figures conflict. The 200,000 figure comes from xAI and is not an independent audit.
What large groups of AI processors do
A graphics processing unit (GPU) can help train AI models. A company can connect many GPUs to train or run larger AI systems. It also needs power, cooling, network connections, and a large budget. A data center is one part of this setup.
The project is one example of the race to build computers for AI. Large companies compete to get enough computer power to train and run AI models. This can affect which models are available and what AI companies charge. It does not mean every new model will help a small business.
What small businesses can learn
Most startups do not need a supercomputer. They can test an online AI service or an API on one clear task. An API lets software use another company’s AI. Before moving more work to it, compare quality, response time, data use, and total cost.
Keep a second AI service or a manual way to do the work in case one service stops working. If a tool handles customer information, check how the company stores and uses it. Review your choice when the price or amount of work changes.
Computer availability and prices can affect what AI services a startup can afford. Compare providers before running a larger test. The AI tools for entrepreneurs guide lists options to evaluate.
Sources
- Musk Plans xAI Supercomputer, Dubbed ‘Gigafactory of Compute’ (The Information, May 2024)
- Elon Musk’s xAI Valued at $24 Billion After Fresh Funding (Reuters, May 27, 2024)
- Colossus: xAI’s compute cluster (xAI)
