The AI startup that promised independence is now a distribution channel for the cloud oligopoly that bankrolled it.
The Summary
- Anthropic is increasingly dependent on Google, Amazon, and Microsoft to sell Claude, the same companies that compete with each other, invest in Anthropic, and provide its compute infrastructure
- The company built to be the "safe AI" alternative is discovering that reaching customers means going through Big Tech's tollbooths
- This reveals the real moat in AI: not models, but distribution and cloud infrastructure
The Signal
Anthropic's growing reliance on its three cloud partners exposes the fundamental tension in the AI economy. You can build the best model in the world, but if you need billions in compute and lack enterprise sales infrastructure, you end up renting both from companies that also build competing models. It's a strange arrangement: Google, Amazon, and Microsoft are simultaneously Anthropic's investors, suppliers, distributors, and rivals.
This isn't unique to Anthropic, but it's particularly revealing for a company founded by ex-OpenAI researchers specifically to chart a different path. That path still runs through the same three cloud platforms that control enterprise AI adoption. When a CIO wants to deploy Claude, they're likely buying it through their existing AWS, Google Cloud, or Azure contract, not going direct to Anthropic.
"The AI startup that promised independence is now a distribution channel for the cloud oligopoly that bankrolled it."
The implications ripple outward:
- Model builders increasingly look like content creators on someone else's platform
- The real margin in AI accrues to whoever owns the customer relationship and the infrastructure
- "Building the best AI" matters less than "getting the AI into production environments"
The irony is thick. Anthropic positioned itself as the principled alternative, the company that would do AI differently. And it has, in terms of safety research and model architecture. But when it comes to go-to-market, it's running the same playbook as everyone else: get your models into the clouds, let the cloud vendors handle the enterprise sales, hope the revenue split works out.
The Implication
Watch how many "independent" AI companies follow this path. The pattern is clear: raise billions from cloud providers, build on their infrastructure, sell through their channels, and slowly become a SKU in someone else's catalog. The companies that figure out direct enterprise distribution, or build on infrastructure they control, will capture asymmetric value. Everyone else is building software for landlords.
For anyone planning an AI company: distribution and infrastructure aren't "boring problems to solve later." They're the entire game. The model is table stakes.