SpaceX’s External Compute Deals Add a Potential Rival for CoreWeave
A confirmed Anthropic arrangement and reported outside-customer contracts put more large-scale AI capacity into the market. The unanswered question is whether it can become a repeatable cloud service.
Listen to this story
The audio brief
Story brief
3 key pointsAn Anthropic agreement to use xAI’s Colossus 1 gives the clearest evidence yet that SpaceX-linked, Nvidia-based infrastructure can serve customers beyond xAI. The Motley Fool reports $14.1 billion in external compute contracts in SpaceX’s latest quarter, but the commercial model is still unproven. CoreWeave’s $2.6 billion quarterly revenue and $104 billion backlog show strong demand, while its operational edge lies...
- 01
Anthropic plans to use Colossus 1 for training, fine-tuning, inference, and HPC supporting Claude Pro and Claude Max.
- 02
CoreWeave’s latest-quarter revenue rose 112% to $2.6 billion; backlog reached $104 billion, up 246%.
- 03
Physical scale may ease power, land, cooling, and networking constraints, but does not prove SpaceX can match CoreWeave’s cloud operations.
CoreWeave may face a new kind of potential AI-cloud rival: SpaceX-linked computing capacity that is beginning to reach external customers, not only xAI’s own workloads.
The competitive issue is not rockets. It is whether large Nvidia-powered capacity associated with xAI’s Colossus facilities can be sold more broadly to AI companies. The Motley Fool reports that SpaceX has signed agreements to supply outside customers, including major AI companies, and contracted $14.1 billion in external computing power in its latest quarter.
The commercial proof point
One external arrangement is confirmed. In May, xAI said it had agreed to give Anthropic access to Colossus 1 for training, fine-tuning, inference and high-performance-computing workloads. xAI said Anthropic planned to use the added capacity to improve service for Claude Pro and Claude Max subscribers.
Owning hardware is not the whole service
CoreWeave provides GPUs, data centers, networking, storage and software for rented AI computing capacity. Its specialized operating work includes scheduling workloads, maintaining utilization and coordinating large GPU clusters through software. Those capabilities are distinct from assembling physical infrastructure at scale.
SpaceX’s cited advantage begins with the physical limits of AI buildout: electricity, suitable land, data centers, cooling and networking. That may help it add capacity, but it does not by itself demonstrate superior AI-cloud operations.
The measurement ahead
CoreWeave reported $2.6 billion in latest-quarter revenue, up 112%, and a $104 billion revenue backlog, up 246%, according to The Motley Fool. Those figures show substantial current demand; they do not settle how pricing or GPU utilization would hold if more large-scale capacity is marketed externally.
Signals that would show a competitive effect
- Pricing as AI-computing supply grows.
- GPU utilization, which indicates whether costly accelerators are being used.
- Capital efficiency and customer diversification as CoreWeave expands.
The Anthropic deal confirms that Colossus capacity can serve an outside customer. Whether SpaceX-linked infrastructure becomes a broad, repeatable AI-computing service remains unresolved—and so does the durability of CoreWeave’s operating edge.
