The AWS chief explains why investing billions in Anthropic and OpenAI is an acceptable struggle

Amazon’s recent $50 billion investment in OpenAI, following its long partnership and including an $8 billion investment in Anthropic, is the kind of conflict of interest the cloud giant is accustomed to dealing with, AWS CEO Matt Jarman said.

Jarman has worked at Amazon since he was a business school intern in 2005, before launching AWS in 2006, he told an audience at the HumanX conference this week in San Francisco.

When asked about the conflict inherent in working closely with two prototypical AI companies that are fierce (and perhaps sometimes petty) competitors, he said it’s not a problem. He explained that because AWS itself often competes with its partners, it has significant first-hand experience with such competition.

In the early years of AWS, it realized it couldn’t build every cloud it offered itself, so Unity partnered with others.

“We also knew that we would have to compete with our partners, because the technology is interconnected,” Jarman said. “So, for a very long time, we’ve built this muscle around how we go to market with our partners,” he continued. “But we may also have first-party products that compete with them, and that’s OK, and we promised them that we wouldn’t give ourselves an unfair competitive advantage.”

Today, the world is accustomed to Amazon competing with those who sell on its cloud. Even Oracle, one of AWS’s biggest competitors, sells its database and other services on AWS. But it was a radical idea in 2006, when technology partners did their best to never compete with the partners who helped them succeed.

However, Amazon is not a pioneer in eliminating investor loyalty and conflict-of-interest obligations in the wild world of money-grabbing AI. When Anthropic announced its latest $30 billion round in February, it included at least a dozen investors who were also backing OpenAI. This included OpenAI’s main cloud partner, Microsoft.

For AWS, making a huge investment in OpenAI to get its model to its customers (and as a technology development partner) was almost a matter of life or death. Both models were already available on Microsoft Cloud, AWS’s biggest competitor.

Cloud giants are also keeping themselves front and center by offering AI model routing services. These services allow their customers to automatically use different models for different tasks as a way to maximize performance and reduce costs. As Jarman explained, one model might be ideal for planning, another for reasoning, and a cheaper model for easier tasks, like code completion. “I think this is where the world is going,” Jarman said.

This is also how Amazon and Microsoft will put their on-premises models to use — the old mode of competing with partners, again.

All’s fair in love and artificial intelligence these days.