Amazon Web Services CEO Matt Garman is defending the cloud giant's unusual strategy of pouring billions into both Anthropic and OpenAI - two fierce rivals in the AI race. In an interview with TechCrunch, Garman argues that AWS has a deeply ingrained culture of what he calls "coopetition" - competing with the same partners it invests in. It's a delicate balancing act that's raising eyebrows across Silicon Valley, but Garman insists it's just business as usual for the world's largest cloud provider.
Amazon Web Services is playing both sides of the AI war, and CEO Matt Garman thinks that's perfectly fine.
The cloud computing behemoth has invested billions in Anthropic - including a $4 billion commitment that made AWS its primary cloud provider - while simultaneously backing OpenAI through a separate multi-billion dollar infrastructure deal. It's the kind of move that would trigger conflict-of-interest alarms at most companies, but Garman insists AWS has been navigating these choppy waters for years.
"We have an ingrained culture of handling coopetition," Garman told TechCrunch in a new interview. The comment reveals how AWS views its complex web of partnerships in the AI era - not as contradictions, but as strategic imperatives.
The logic goes like this: AWS competes with nearly everyone it does business with. Netflix runs on AWS infrastructure while building its own content delivery systems. Salesforce uses AWS compute while offering competing cloud services. Even Apple relies on AWS for iCloud storage while positioning itself as a privacy-first alternative to Amazon's data-hungry ecosystem.
For Garman, backing both Anthropic and OpenAI is just an extension of that philosophy. AWS gets to position itself as the infrastructure layer for whoever wins the foundation model race, while also developing its own AI services through Amazon Bedrock and other initiatives. It's hedge-fund thinking applied to cloud computing - diversify your bets, capture value regardless of the outcome.
But the AI landscape is different from traditional cloud partnerships. Anthropic and OpenAI aren't just customers - they're competitors racing toward artificial general intelligence, each claiming their approach is safer or more capable. When AWS backs both horses, it gains access to cutting-edge research and model development, potentially creating information asymmetries that smaller AI labs can't match.
The arrangement also raises questions about how AWS allocates resources. Does Anthropic get priority access to AWS's latest Trainium chips? Does OpenAI receive better infrastructure pricing? And what happens when AWS's own foundation models - built using insights gleaned from hosting both companies - start competing directly with its partners?
Garman's coopetition argument works when you're providing commodity infrastructure. It gets murkier when you're deeply embedded in your partners' R&D processes, which is exactly where AWS sits with both AI labs. Microsoft faced similar scrutiny for its $13 billion investment in OpenAI while offering Azure AI services to competitors, but at least Microsoft picked a side.
Google and Meta have avoided this complexity altogether by building their AI capabilities in-house. Google's Gemini models run on Google Cloud infrastructure, period. Meta's Llama development happens entirely within its own data centers. They don't have to explain why they're funding rivals.
The real test will come when one of these partnerships starts to falter. If Anthropic's Claude models fall behind OpenAI's GPT series, does AWS shift resources? Or if AWS's own foundation models become competitive enough to threaten both partners, how does Garman navigate those conversations?
For now, AWS is collecting checks from all sides. Enterprise customers pay for Bedrock access, Anthropic and OpenAI both pay massive infrastructure bills, and AWS maintains pole position as the AI industry's landlord. Garman's coopetition defense is really just another way of saying AWS wants to own the pipes no matter which way the water flows.
The strategy reflects a broader truth about big tech in 2026: the platform always wins. Whether it's Apple's App Store, Google's search dominance, or AWS's cloud infrastructure, controlling the underlying layer means you don't have to pick winners. You just collect rent from everyone.
But coopetition only works when your partners believe you're playing fair. The moment Anthropic or OpenAI suspects AWS is using their infrastructure needs to gain competitive advantage, the whole arrangement collapses. Garman's betting that won't happen - that AWS's reputation as a neutral platform provider is strong enough to withstand the inherent conflicts in backing rival AI labs.
Garman's coopetition defense is ultimately a bet that AWS can maintain Switzerland-level neutrality while profiting from the AI arms race. It's a high-wire act that works brilliantly until it doesn't - and the stakes have never been higher. As foundation models become the new operating systems of computing, AWS's dual-investment strategy either looks like genius-level hedging or a conflict of interest waiting to explode. The answer probably depends on which AI lab you're backing.