A recent compute deal between SpaceX and Anthropic may have given Meta Platforms, Inc‘s (NASDAQ:META) bulls a fresh way to think about the company’s massive AI spending.
The reason is simple: AI compute is increasingly looking like a valuable asset rather than just a cost center.
The SpaceX-Anthropic Signal
Reports that Anthropic secured compute capacity through SpaceX highlighted just how tight the AI infrastructure market has become.
As leading AI developers race to train larger models and serve growing numbers of users, access to compute has emerged as one of the industry’s most important resources. The deal suggests companies are willing to look beyond traditional cloud providers to secure the capacity they need.
That has sparked a broader debate among investors.
If compute capacity is becoming a scarce and monetizable resource, what could that mean for companies building some of the world’s largest AI clusters?
Meta’s Quiet Compute Lead
Meta is pouring hundreds of billions of dollars into AI clusters. According to projections from SemiAnalysis, the Prometheus cluster will be one of the largest AI training clusters among the industry’s three most ambitious projects by the end of 2026.
The research estimates Prometheus will reach 1,020 megawatts of IT power, ahead of OpenAI’s Stargate project at 880 megawatts and Anthropic’s Rainier cluster at 780 megawatts.
But the more striking difference shows up in the compute output.
SemiAnalysis projects Meta’s cluster will deliver roughly 3.17 billion TFLOPS (Tera Floating Point Operations Per Second), compared with 2.47 billion TFLOPS for OpenAI and 1.04 billion TFLOPS for Anthropic.
The chip counts make the comparison even more interesting. Meta is projected to achieve that compute output using around 500,000 Nvidia Corp (NASDAQ:NVDA) GB200/300 chips, while Anthropic’s Rainier project is expected to deploy roughly 800,000 Trainium 2 accelerators.
In other words, Meta’s cluster is projected to generate roughly three times the compute of Anthropic’s while using significantly fewer chips.
A New Bull Case Emerges
That is where the SpaceX-Anthropic deal becomes relevant.
Meta has never indicated that it plans to sell compute capacity to outside customers. But the emergence of a market for AI compute raises a question that investors weren’t asking a year ago.
If companies are willing to pay a premium for access to compute, then one of the world’s largest AI clusters may have value beyond supporting Meta’s own models and products.
For Meta bulls, that creates a different lens through which to view the company’s AI spending. Instead of focusing solely on the cost, investors may increasingly focus on the asset being built—and whether owning one of the industry’s largest compute footprints becomes a competitive advantage in its own right.
Image via Shutterstock
Login to comment