The Stack That Feeds Itself
Google is pouring up to $40 billion into Anthropic, following Amazon’s recent $5 billion injection. Both deals value the Claude maker at a staggering $350 billion. But this isn’t charity. It’s a closed loop: Google and Amazon invest cash, Anthropic uses that cash to buy compute and chips from the very same investors. Amazon gets Anthropic hooked on AWS. Google locks Claude into TPU pods. The AI startup becomes a captive customer, not a true partner. This incestuous arrangement masks a simple truth: the supposed independence of leading AI labs is a fiction.
Performance Targets vs. Real World Performance
Anthropic’s growth is real, fueled by controversies at OpenAI and its own agentic tools like Claude Code and Claude Cowork. Demand is so high it’s causing outages. But the investment is contingent on hitting ‘performance targets.’ That’s a setup for perverse incentives. When a $40 billion check depends on hitting metrics, the pressure to cut corners, fudge benchmarks, or push unstable code into production becomes immense. Meanwhile, Google and Amazon sit on both sides of the table: they supply the chips, host the models, and will profit whether Claude succeeds or fails. This isn’t a bet on AI. It’s a hedge on the entire market.
The Illusion of Choice
Let’s call this what it is: a cartel dressed up as competition. Google owns stakes in Anthropic while pushing Gemini. Amazon backs Anthropic while developing Olympus. Microsoft orchestrated the OpenAI rescue. The big three cloud providers now have a finger in every pie. No CVE disclosures or safety audits can fix a fundamental structural rot where the companies that control the infrastructure also control the competition. This arrangement ensures no lab can truly challenge the platform owners. The AI boom isn’t creating a vibrant ecosystem. It’s building a walled garden with three entrances, all owned by the same landlords.
Source: Arstechnica
