AIThe case for doing AI’s hard infrastructure work
One post argues that AWS got companies used to outsourcing hard technical problems, but AI infrastructure isn’t mature enough for the same approach.
TLDR
A post argues that abundant money makes easy AI infrastructure work profitable, leaving little incentive to tackle harder problems. It points to the proliferation of model routers and claims most inference companies do minimal work on compute someone else set up. The author sees a huge gap between OpenAI, Anthropic, DeepSeek and Moonshot and everyone else—and argues that closing it requires work companies can’t simply assume someone else will handle.
The case for doing AI’s hard infrastructure work
One post argues that AWS got companies used to outsourcing hard technical problems, but AI infrastructure isn’t mature enough for the same approach.
TLDR
A post argues that abundant money makes easy AI infrastructure work profitable, leaving little incentive to tackle harder problems. It points to the proliferation of model routers and claims most inference companies do minimal work on compute someone else set up. The author sees a huge gap between OpenAI, Anthropic, DeepSeek and Moonshot and everyone else—and argues that closing it requires work companies can’t simply assume someone else will handle.