Is xAI a neocloud now?

The neocloud framing matters because it repositions xAI not just as a model competitor but as a compute vendor, meaning its revenue thesis now depends on selling or leasing infrastructure capacity to third parties, not solely on Grok adoption. That is a fundamentally different business with different margin profiles and customer relationships.

This move lands directly inside the infrastructure arms race our coverage has been tracking. The Decoder's piece on big tech committing $725 billion to AI infrastructure in 2026 established that compute depth is now the primary competitive lever, and xAI appears to be drawing the same conclusion independently. What makes this more complicated is the earlier disclosure, covered here from the Musk-Altman trial, that xAI has relied on OpenAI model outputs for its own training. A lab with upstream model dependencies trying to sell downstream compute capacity is an unusual position, and it raises real questions about whether xAI's infrastructure play is a genuine moat or a way to generate revenue while the model stack matures.

Watch whether xAI announces external customers for its datacenter capacity within the next two quarters. If it does, the neocloud thesis has legs. If the infrastructure stays captive to internal Grok training, this is capital expenditure dressed up as a business model.