This week, we explore why Anthropic turned to SpaceX for compute infrastructure, how “compute” became one of the defining currencies of the AI race, and why the future of AI increasingly depends on the cloud ecosystems, power systems, and infrastructure operating underneath the software itself.

AI in Action

The AI Race is Expanding Beyond Software [AI Infrastructure/Compute]

What does one of the world’s leading AI companies have to do with a rocket company? Anthropic recently signed a partnership to access SpaceXAI’s Colossus 1 compute infrastructure, gaining access to more than 220,000 Nvidia GPUs as demand for Claude and AI coding workloads continues to rise. The companies also discussed the possibility of future orbital AI compute infrastructure. The scale behind frontier AI increasingly resembles industrial infrastructure projects more than traditional software development. Training and operating these systems now requires enormous amounts of compute, power, cooling, networking, and operational coordination.

That shift is spreading across the industry. OpenAI runs closely within Microsoft’s cloud ecosystem. Anthropic is deeply integrated into Amazon’s AWS infrastructure. Google combines frontier models with its own chips, cloud infrastructure, and global network. Nvidia increasingly sits at the center of the compute layer itself. Earlier software eras scaled primarily through code deployment and internet distribution. Frontier AI scaling increasingly depends on GPU clusters, hyperscale data centers, energy supply, networking capacity, and the infrastructure surrounding them. Even leading AI labs increasingly rely on outside infrastructure providers because building frontier-scale compute environments has become extraordinarily capital intensive.

The effects increasingly reach beyond AI labs themselves. Enterprises adopting AI are gradually building workflows inside ecosystems controlling compute access, developer tooling, cloud infrastructure, identity systems, and AI services. Similar dynamics shaped the cloud computing era, where infrastructure concentration created deep platform dependence over time. AI deployment increasingly follows the same pattern. Infrastructure access increasingly shapes how quickly models can scale across products, regions, and enterprise environments. The model still matters, though the competitive advantage increasingly sits underneath it, inside the ecosystems powering deployment and distribution.

The next bottlenecks in AI increasingly involve physical systems. Power grids, cooling systems, optical networking, energy supply, and compute capacity are becoming part of the race to scale intelligence. Governments are also starting to treat AI infrastructure as a strategic asset tied to economic competitiveness and national capability. AI may still appear as software through the interface people interact with every day, while the economics underneath increasingly resemble industrial infrastructure operating at global scale.

Bits of Brilliance

What is “compute”? [AI Infrastructure]

“Compute” has become one of the defining currencies of the AI race, usually mentioned alongside companies like Nvidia, Microsoft, Amazon, Google, OpenAI, and now SpaceXAI. The term refers to the processing power required to train and run AI systems. Every chatbot response, coding suggestion, AI-generated image, search result, or enterprise AI workflow depends on enormous numbers of calculations happening behind the scenes. Frontier AI models can require tens of thousands of GPUs running simultaneously across massive data centers consuming enough electricity to rival small industrial facilities. As models become larger and more capable, access to compute increasingly shapes how quickly companies can train models, launch products, and compete at the frontier of AI.

Cloud and compute are closely connected, though they serve different functions. Compute is the actual processing power performing the calculations. Cloud providers like Microsoft Azure, AWS, and Google Cloud build and manage the infrastructure giving companies access to that compute at global scale. Most firms do not build frontier AI infrastructure themselves for the same reason most companies do not operate their own electrical grid. They connect into larger systems already running enormous networks of servers, power systems, cooling equipment, storage, and networking infrastructure. That dynamic increasingly concentrates frontier AI development around companies with access to massive cloud ecosystems, GPU clusters, and capital-intensive infrastructure.

The demand does not stop once a model finishes training. Running AI systems across millions of daily interactions also consumes substantial compute. Every prompt, search query, coding request, AI-generated image, or enterprise agent workflow requires processing power in real time. An enterprise AI agent drafting reports, searching internal documents, querying databases, updating systems, and coordinating workflows may trigger dozens of compute-heavy operations behind a single request. As AI adoption expands, ongoing inference demand may become just as infrastructure-intensive as training frontier models themselves.

This is one reason compute increasingly sits at the center of the AI race. The companies controlling access to large-scale compute increasingly influence how AI systems are built, deployed, scaled, and distributed. Intelligence may appear through a chat window or software interface, while the infrastructure underneath increasingly determines who can realistically compete at frontier scale.

_{Byte-Sized Intelligence is a personal newsletter created for educational and informational purposes only. The content reflects the personal views of the author and does not represent the opinions of any employer or affiliated organization. This publication does not offer financial, investment, legal, or professional advice. Any references to tools, technologies, or companies are for illustrative purposes only and do not constitute endorsements. Readers should independently verify any information before acting on it. All AI-generated content or tool usage should be approached critically. Always apply human judgment and discretion when using or interpreting AI outputs.}