Amazon’s recent $50 billion bet on OpenAI might seem like a conflict of interest given its existing $8 billion partnership with Anthropic. To AWS CEO Matt Garman, however, it is simply business as usual. Speaking at the HumanX conference in San Francisco, Garman explained that navigating high-stakes rivalries is a “muscle” Amazon has been building since the cloud’s inception.
The Strategy of “Coopetition”
Garman, an Amazon veteran since 2005, noted that AWS was built on the once-radical idea that a company can compete with its partners while still providing them a platform. This philosophy of “coopetition” is now a tech industry standard. For example, even Oracle, one of AWS’s fiercest rivals, sells its database services on the AWS cloud.
For Amazon, the goal is to maintain a neutral ground. While the company may develop first-party products that compete with its partners, it maintains a commitment to avoid giving itself an unfair competitive advantage. In the AI arms race, this means providing customers with the best models available, even if those models come from companies that are “fierce and petty” competitors.
Survival in a Multi-Model World
Amazon isn’t the only one playing both sides. Microsoft also holds stakes in multiple AI ventures. For AWS, securing access to OpenAI’s models was a strategic necessity. With both OpenAI and Anthropic already available on Microsoft Azure, AWS had to ensure its customers weren’t forced to migrate to their biggest rival to access top-tier tech.
The Rise of Model Routing
Garman envisions a future defined by model routing, where businesses don’t rely on a single AI. Instead, they will use a variety of models tailored to specific tasks:
- Complex Logic: High-end models for reasoning and long-term planning.
- Efficiency: Cheaper, lightweight models for routine tasks like code completion.
By hosting a suite of competing models, AWS positions itself as the essential infrastructure for this multi-model future. It also creates a pathway for Amazon to eventually slide its own homegrown AI models into the rotation—proving that in the world of cloud computing, all is fair in love and AI.
