Just now, has Amazon's "AI turning point" emerged?

With the official operation of its core data center, Amazon’s AI infrastructure has reached a key milestone.

Just a few days ago, Amazon CEO Andy Jassy announced on social platform X:

A cornfield near South Bend, Indiana, USA, has now become one of the world’s largest AI compute clusters — the core data center of Project Rainier. This system, jointly built by AWS and AI unicorn Anthropic, has deployed nearly 500,000 self-developed Trainium2 chips, making it 70% larger than any previous AWS AI platform. It is now fully operational.

According to Jassy, Anthropic, Amazon’s partner, is using the system to train and run its large model Claude, providing more than five times the computing power previously used for its AI model training. By the end of the year, the system is expected to double its Trainium2 chip deployment to one million.

This means Amazon's AI infrastructure expansion is shifting from strategic planning to production capacity realization, marking a major turning point in its AI business development.

Morgan Stanley predicts that AWS's revenue growth over the next two years will reach 23% and 25% respectively. According to Bank of America, Anthropic alone could bring AWS up to $6 billion in incremental revenue by 2026.

Supercomputing Cluster: Redefining the Scale of AI Infrastructure

The official launch of Project Rainier marks the beginning of AWS’s large-scale capacity expansion for AI.

The system is distributed across multiple U.S. data centers, connecting tens of thousands of super servers via NeuronLink technology to minimize communication latency and enhance overall computing efficiency.

With nearly 500,000 Trainium2 chips, it is one of the largest AI training computers in the world. Amazon plans to further expand 1GW of capacity by year-end and increase the Trainium2 chip count by another 500,000. Even more ambitious, the company aims to double AWS’s GW capacity by 2027.

AWS CEO Matt Garman previously emphasized that these self-developed chips outperform general alternatives. On an earnings call, Jassy said, “Adoption rates for Trainium2 continue to rise, and all current capacity is fully booked. This business is expanding rapidly.”

Self-Developed Chip Strategy Shows Initial Results

The core of Amazon’s AI strategy is not the model, but the ‘computing base’ — its self-developed chip system: The Trainium series (for AI training) and Inferentia series (for inference) make up AWS’s “dual engines” in AI computing.

Now, this strategy is showing results.

The Trainium chip series has become a multi-billion dollar core business, with quarterly growth of 150%. This approach not only helps reduce the costs of model training and inference, but ultimately improves AWS’s business profit margins.

Meanwhile, Amazon is preparing to launch the Trainium3 chip, expected as soon as this year's re:Invent conference, with larger deployments planned in 2026. The new generation boasts not just improved performance but will also expand to a broader customer base, meaning AWS’s AI services will go beyond “top customers” to reach the wider enterprise market.

Bank of America analyst Justin Post noted that the cost optimization effect of self-developed chips is already apparent: Trainium adoption has significantly lowered model training and inference costs, improving AWS's margins and becoming a new multi-billion-dollar growth engine.

Jassy previously revealed that the company's AI platform, Bedrock, aims to become "the world's largest inference engine," with long-term potential rivaling AWS’s core compute service EC2. Currently, the vast majority of token usage on Bedrock runs on Trainium chips.

Morgan Stanley Upgrades Amazon Rating: AWS Enters an “AI Growth Acceleration Cycle”

Morgan Stanley’s latest research listed Amazon as a “Top Pick” and raised the target price from $300 to $315, about 25% above the current share price.

The reason is simple — AWS is entering an “AI growth acceleration cycle.” Analysts summarized four key growth drivers:

1. Rapid capacity expansion: 1GW more computing power by year-end, doubling Trainium2 chip count;
2. Structural expansion cycle: AWS plans to add 10GW data center capacity in the next 24 months;
3. AI order surge: New orders in October surpassed those in the entire third quarter, with about $18 billion of new business in just one month;
4. Accelerated innovation: Trainium3 likely to debut this year, with software platform Bedrock continuing expansion.

The report says AWS is still in a "capacity constrained" state — supply falls short of demand, which becomes a key driver of growth.

According to Morgan Stanley analysis, AWS’s new business signed in October exceeded the whole third quarter. Their backlog analysis suggests Amazon may have signed around $18 billion in new business that month.

“If not for compute bottlenecks, AWS’s growth would be even faster,” wrote analysts. “These capacity expansion plans are paving the way for another acceleration in 2026–2027.”

Morgan Stanley expects AWS revenue growth of 23% and 25% over the next two years, with the AI boom projected to bring as much as $6 billion in incremental revenue in 2026, boosting overall growth rate by about 4 percentage points.

Morgan Stanley also raised Amazon’s capital expenditure estimates for 2026/2027 by 13%/19% respectively, now forecasting total capex at $169 billion/$202 billion. This means the company will invest $140 billion/$170 billion in technology and infrastructure, surpassing other giants like Microsoft, Meta, and Google.

Analysts believe that while Amazon is making massive investments to expand computing capacity, “capacity is absorbed as soon as it becomes available,” meaning it’s still at a very early stage, bringing unprecedented opportunities to AWS clients.

Risk warning and disclaimerThe market involves risks; investments should be made with caution. This article does not constitute personal investment advice, nor does it take into account individual users' specific investment objectives, financial situation, or needs. Users should consider whether any opinions, views, or conclusions in this article are suitable for their own situation. Investments made accordingly are at your own risk.