Turns out infrastructure beats ideology when you own the pipes and the competition needs somewhere to run their models.
The Summary
- SpaceX's Stargate venture signed a $4 billion deal to supply data center capacity to Anthropic, the company Musk called "evil" just three months ago for allegedly censoring political speech.
- The deal gives Anthropic infrastructure independence from Google, its primary investor and cloud provider, while positioning Musk as the kingmaker in the AI race.
- Musk now profits from both sides of the frontier AI competition: xAI (his own lab) and Anthropic (his former nemesis), while controlling the data center layer both depend on.
- The move demonstrates that "he who controls the data center, really does control the application of artificial intelligence right now", regardless of philosophical differences.
The Signal
Three months ago, Elon Musk went on a public tirade against Anthropic, calling the AI safety company "evil" for what he characterized as left-wing political censorship in its Claude models. He accused them of prioritizing ideology over truth, a familiar refrain from someone who's made "anti-woke AI" a selling point for his own xAI venture. Now SpaceX's Stargate subsidiary just inked a $4 billion contract to become Anthropic's infrastructure provider. Principles, it turns out, have a price point.
The deal is nakedly strategic on both sides. Anthropic gets independence from Google, which has been both its largest investor and its primary cloud infrastructure provider through Google Cloud. That's a precarious position when you're trying to compete with Gemini, Google's own frontier model. Diversifying compute suppliers isn't just smart, it's existential. Musk gets revenue from a competitor and cements his position as the infrastructure king of the AI era.
"He who controls the data center, really does control the application of artificial intelligence right now."
The Algorithmic Bridge frames this as Musk becoming "kingmaker" in AI, and that's not hyperbole. Consider the position:
- xAI competes directly with Anthropic, OpenAI, and Google in frontier models
- SpaceX Stargate now supplies infrastructure to Anthropic
- Musk has visibility into Anthropic's scaling plans, capacity needs, and architecture constraints
- He profits whether xAI wins or Anthropic does
This isn't a conflict of interest. It's a conflict-of-interest arbitrage strategy. The incentive structure is almost beautiful in its cynicism. If Claude beats Grok in the market, Musk still wins because Anthropic's training runs and inference loads flow through his data centers. If Grok wins, even better. If both succeed, best case scenario.
The $4 billion number matters. That's not a small pilot program or a tactical hedge. That's multi-year, multi-data-center commitment money. It suggests Anthropic is planning significant capacity expansion beyond what Google can or will provide, likely to support both larger training runs and growing enterprise inference demand. It also means Musk is betting big that AI compute scarcity remains the binding constraint for the next several years.
The timing is telling too. Google just posted disappointing cloud growth numbers, missing analyst estimates for the third straight quarter. Amazon and Microsoft are both tightening capital allocation to AI infrastructure, signaling that the hyperscalers see margin pressure ahead. Anthropic probably saw the writing on the wall: relying on a single cloud provider who's also a direct competitor is untenable as models scale. Better to diversify now, even if it means writing a check to someone who called you evil on Twitter.
The Implication
Watch for other frontier AI labs to follow Anthropic's lead. If you're OpenAI and you've bet everything on Microsoft's infrastructure, this deal just made your risk management team very nervous. Vertical integration into compute was already a theme for 2025. This accelerates it. Expect more AI companies to either build their own data centers or strike deals with non-hyperscaler infrastructure providers who don't have competing model businesses.
For Musk, this is the playbook: own the layer everyone depends on, stay ideologically flexible enough to do business with anyone, and profit from the arms race regardless of who wins. The "evil" rhetoric was theater. The $4 billion contract is the real signal about how power accumulates in the agent economy. Infrastructure always beats applications in the long run.