Let's get straight to the point, because right now everyone is talking about OpenAI, Anthropic, and Nvidia. But they're forgetting someone. Amazon.
Yes, Amazon. The company most people know as the world's largest online marketplace is quietly fighting its own battle in artificial intelligence. And here's what makes it interesting. It has invested billions in both OpenAI and Anthropic, yet at the same time it wants to build its own complete AI ecosystem. In other words, it's trying to win from both sides.
Let's break it all down.
THE NEW $1 BILLION TEAM
AWS, Amazon's cloud computing division, has announced a $1 billion investment in a new organization called Forward Deployed Engineering.
So what does this team actually do?
In simple terms, Amazon sends its own engineers directly into customer companies. They work side by side with employees to build AI systems on site. These engineers are known as FDEs. The concept itself isn't new. Palantir introduced the idea more than a decade ago. But now it's making a comeback because every company wants to move faster.
How does this work in practice?
Amazon says it plans to deploy thousands of these engineers. They typically work in teams of five or six people alongside AI agents that can automate many tasks. The goal is simple. Within a matter of weeks, they leave behind an internal team that is capable of operating independently.
Francesca Vasquez, who leads the organization, said customers are asking for one thing above all else: speed.
It's also worth noting that AWS is the first major cloud provider to formally launch this type of initiative. OpenAI and Anthropic created their own Forward Deployed Engineering organizations earlier this year, but AWS is now making its own move. The team is already working with organizations including the Allen Institute, the NBA, the NFL, and Ricoh.
THE MODEL STRATEGY
Now let's move to the second piece of the puzzle.
People are one thing. AI models are another.
Here, Amazon made a surprisingly honest admission.
Peter DeSantis, one of the company's top AI executives, openly acknowledged that Amazon's models have not been at the very top for the most demanding workloads. In other words, they've been trailing behind OpenAI and Anthropic.
But they have ambitious plans.
He says the goal is to catch up within the next year.
"We took our time to build the right foundation," he explained, referring to data, architecture, and infrastructure.
Amazon's strategy is built on two pillars.
The first is Bedrock, which you can think of as a marketplace for AI models. Customers can choose models from multiple providers depending on their needs.
The second is Amazon's own family of AI models, Nova, launched in December.
How is it performing?
According to DeSantis, Nova now has around 50,000 customers. He admits it isn't yet among the industry's top performing models, but that's exactly the objective. Amazon wants to build a model that competes with the very best.
THE HIDDEN WEAPONS
This is where things become even more interesting.
Amazon isn't just building engineers and models.
It's also building chips.
Yes, its custom silicon strategy has been developing for years through products such as Trainium and Graviton.
DeSantis even compares Amazon's position to Nvidia's.
Why?
Because Amazon is one of the few companies designing its own chips, controlling the hardware architecture, and bringing those products into production.
So what could this mean for the market?
Amazon may eventually begin selling these chips directly to other companies.
CEO Andy Jassy said back in April that Amazon is considering selling complete racks powered by Trainium chips to third parties. There is still no official timeline, but the idea is clearly on the table.
For now, customers access that computing power through AWS.
And one of AWS's biggest AI customers?
Anthropic.
The irony speaks for itself.
Now put all the pieces together.
Engineers embedded inside customer organizations.
Proprietary AI models.
Custom-built chips.
This isn't just another AI product.
It's an attempt to build a complete AI ecosystem from end to end.