To The Moon

Forget rockets for a moment. Elon Musk’s AI infrastructure business may be becoming one of the most important parts of the $SpaceX(SPCX)$  story.

Two deals tell us why.

💰 Anthropic: $1.25 billion per month

Anthropic agreed to use computing capacity at SpaceX’s Colossus data centre, giving its Claude AI business access to a massive Nvidia GPU cluster.

💰 Google: $920 million per month

Google signed a multiyear agreement for access to approximately 110,000 Nvidia GPUs, along with CPUs, memory and related infrastructure. Payments are scheduled to reach the full monthly rate from October 2026. 

Combined, the contracts represent approximately $2.17 billion per month, or $26 billion annualised.

For context, SpaceX and xAI reported combined 2025 revenue of approximately $18.67 billion. These two compute contracts alone could therefore represent a revenue stream larger than that historical figure. 

But there is a bigger story here.

🤔 Why would Google rent Nvidia GPUs when it builds its own TPUs?

Google has its own AI chips and one of the world’s largest cloud infrastructures. So why rent computing capacity from Musk’s company?

Because designing chips and building cloud infrastructure does not guarantee unlimited capacity whenever demand surges.

Google has described the agreement as a way to secure additional capacity amid stronger-than-expected demand for its Gemini Enterprise platform. Renting Nvidia GPUs can provide another source of computing power without waiting for every new internal infrastructure project to come online. 

The lesson is that even the biggest technology companies can face physical infrastructure constraints.

⚡ The real AI bottleneck may be execution

AI infrastructure requires more than powerful chips.

It needs:

• Reliable electricity and sufficient power capacity.

• Data centres that can be built and brought online.

• Cooling, networking and supporting infrastructure.

• Enough GPUs to meet customers’ computing needs.

SpaceX’s advantage is not that it owns every part of the AI supply chain. It is that it has developed large-scale computing infrastructure and can potentially monetise that capacity by renting it to customers.

Nvidia supplies the GPUs. SpaceX supplies access to the computing infrastructure. Anthropic and Google pay to use it.

Different businesses, different roles — but all essential to the AI ecosystem.

⚠️ One important caveat

US$26 billion in annualised contract value does not mean US$26 billion in profit, nor does it guarantee that the full amount will be recognised as revenue in a single year.

Infrastructure costs, electricity, depreciation, capital expenditure and contract terms all matter. SpaceX and xAI also reported substantial losses in 2025, despite their revenue growth. 

My takeaway: The AI opportunity is expanding beyond model developers and chipmakers. Companies that can deliver computing capacity at scale may capture a significant share of the value — provided they can operate that infrastructure profitably.

The gold-rush analogy still applies: selling the gold is one business, supplying the tools is another, and providing the road and power needed to reach the mine can be just as important.

💬 Which part of the AI ecosystem do you think has the strongest long-term advantage: Nvidia’s chips, Google’s AI infrastructure, or SpaceX’s ability to deploy and monetise computing capacity?

Sharing this perspective for discussion purposes only.