Amazon’s cloud chief Matt Garman isn’t losing sleep over backing two fierce rivals. At the HumanX conference in San Francisco, the AWS CEO defended Amazon’s $50 billion investment in OpenAI alongside its existing $8 billion stake in Anthropic, according to TechCrunch AI.
Garman’s argument boils down to muscle memory. AWS has spent two decades competing with its own partners, and he says the company knows how to keep things fair. “We’ve promised them we won’t give ourselves unfair competitive advantage,” he told the audience.
It’s a tidy narrative. But the real story is about something bigger: the cloud wars have entered a phase where neutrality is no longer optional.
Why AWS Had No Choice
Before the OpenAI deal, both Claude and GPT models were available on Microsoft Azure, AWS’s biggest competitor. That’s a dangerous position for any cloud provider. If enterprise customers can get every top-tier model on Azure but only Anthropic on AWS, the math is simple.
The $50 billion OpenAI investment wasn’t about picking favorites. It was defensive infrastructure spending. AWS needed model parity to stay competitive, and it paid a massive premium to get it.
The “Conflict” That Isn’t Really a Conflict
Garman framed this as business-as-usual, pointing to Oracle selling its database on AWS despite being a direct cloud rival. That comparison holds up. Cloud platforms evolved into marketplaces years ago, and customers expect access to best-in-class tools regardless of competitive dynamics.
What’s more telling: when Anthropic raised its $30 billion round in February, at least a dozen investors were also backing OpenAI. Microsoft itself was among them. TechCrunch AI notes that investor loyalty has essentially dissolved in the AI gold rush. Everyone is hedging.
This isn’t a bug. It’s the new normal.
The Model-Routing Play
The most strategic part of Garman’s remarks wasn’t about investments. It was about model routing. AWS, like Microsoft, is building services that automatically select different AI models for different tasks: one for planning, another for reasoning, a cheaper option for code completion.
“I think that is where the world will go,” Garman said.
This is where the real competitive advantage lives. Whoever controls the routing layer controls which models get used, how often, and for what. It’s also the quiet doorway through which Amazon and Microsoft will push their own homegrown models into production workloads.
Think of it like shelf placement in a grocery store. You stock every brand, but your store brand sits at eye level.
What This Means for Practitioners
A few practical takeaways:
- Multi-model is the default now. If AWS is building routing infrastructure to swap models per task, your architecture should assume the same flexibility. Don’t lock into a single provider.
- Cloud choice increasingly means model access. Enterprise decisions about AWS vs. Azure vs. GCP will hinge on which models are available, how they’re priced, and how well the routing works.
- Homegrown models will get aggressive distribution. Amazon’s Titan and Nova models, plus Microsoft’s Phi family, will show up as default options in routing recommendations. Test them against alternatives before accepting the defaults.
The Bigger Picture
The AI industry is settling into a pattern that looks a lot like the early smartphone era. Back then, carriers subsidized phones from multiple manufacturers while pushing their own services. The cloud giants are doing the same thing with AI models: subsidize access to the best ones, build the distribution layer, and quietly insert your own offerings wherever possible.
Garman’s calm confidence about managing conflicts is genuine. AWS really has done this before. But the stakes are orders of magnitude higher now, and the model makers know it. Anthropic and OpenAI are both taking billions from companies that openly plan to compete with them.
For more details on Garman’s remarks and the evolving cloud-AI dynamics, check out the full piece from TechCrunch AI.