Amazon.com Inc. (AMZN) is making a big bet that the future of computing won't just be about AI models—it'll be about AI agents that actually do things. At its annual AWS re:Invent conference in Las Vegas, the company laid out a vision that attempts to balance its massive cloud computing legacy with an aggressive push into the AI era.

Goldman Sachs analyst Eric Sheridan maintained a Buy rating on Amazon with a price forecast of $290, arguing that the company's recent announcements should give investors more confidence that AWS can reaccelerate growth in an AI-driven world.

Here's what Amazon is banking on: AWS executives hammered home three core themes at the conference that reveal how they're thinking about the next phase of cloud computing.

The Custom Chip Strategy

First, AWS plans to keep developing its own custom chips to drive down computing costs while maintaining its partnership with Nvidia Corp (NVDA) for premium AI capacity. It's a both-and strategy rather than either-or. Amazon introduced the Graviton 5 custom chip and Trainium3 instances, along with new AI Factories that allow customers to deploy dedicated AI infrastructure directly in their own data centers.

The logic makes sense: if you can build cheaper chips that handle certain workloads efficiently, you can offer better pricing to customers while protecting margins. But when customers need the absolute cutting-edge performance, Nvidia's chips are still in the mix.

Betting Big on AI Agents

Second, Amazon believes AI agents—not just models—will fundamentally transform how businesses operate. These aren't passive tools that answer questions. They're systems designed to automate tasks, learn from interactions, and improve efficiency over time. Amazon rolled out new autonomous agents and the Nova Act service specifically for automating complex workflows.

Think of it as moving from "AI that tells you things" to "AI that does things for you." That's a meaningful shift, and it's where Amazon sees the real enterprise value emerging.

Bedrock's Rapid Growth

Third, Bedrock—AWS's platform providing access to various AI models—continues scaling fast. The platform now serves more than 100,000 customers, roughly double from a year ago. Amazon expanded it with new controls and fine-tuning options, launched its Nova 2 family of multimodal AI models, and added more open-weight models to give customers broader choices.

Can Amazon Reaccelerate Growth?

The central question investors keep asking is whether Amazon can accelerate growth again, especially compared to cloud competitors in an AI-driven market. Sheridan thinks the answer is yes, pointing to both recent earnings results and the re:Invent announcements as evidence.

His argument: AWS is well-positioned to reaccelerate revenue growth as AI evolves from a handful of foundational models into a much broader ecosystem of applications, agents, and workloads that require lower-cost, more efficient infrastructure. According to Sheridan, AWS is preparing for this wave of AI adoption the same way it prepared for earlier shifts in computing, storage, and databases.

As the world's largest cloud provider, AWS can support enterprises, developers, and model builders at scale. Sheridan noted the company's approach mirrors what made Amazon successful in e-commerce: broad supply, easy access, strong matching of supply and demand, and a relentless focus on customer needs.

Long-Term Growth Projections

Sheridan's recent research strengthened his conviction that AWS can deliver more than 20% compounded revenue growth over the next several years. He also expects AWS to reach GAAP EBIT margins in the low-to-mid 30% range, even as Amazon continues investing heavily in AI infrastructure and absorbs higher depreciation costs.

Beyond AWS, Sheridan added that Amazon's e-commerce and digital advertising businesses also show constructive trends heading into the next quarter.

At re:Invent, Amazon reinforced what it calls its three-layer AI strategy: infrastructure, models, and applications. The infrastructure layer includes those custom chips and new data center options. The models layer encompasses Bedrock's expanded capabilities and the Nova 2 family. The applications layer focuses on those autonomous agents and workflow automation tools.

Market Share and Competition

Sheridan acknowledged that AWS has lost some cloud market share, but he attributes it to having a larger revenue base, Azure's enterprise strength, and intense competition for AI workloads. Still, AWS backlog growth remains strong, and AI demand is driving significant new commitments from customers.

His analysis estimated AWS AI services could generate approximately $10.7 billion in revenue in 2025 and $23.1 billion in 2026, contributing several hundred basis points to AWS's overall growth rate.

The bottom line: Amazon is trying to position itself not just as a provider of cloud infrastructure, but as the platform where enterprises build and deploy AI systems that actually run their businesses. Whether that strategy pays off depends on execution, but the pieces are starting to come together.

AMZN Price Action: Amazon.com shares were up 0.17% at $229.47 at the time of publication on Friday.