Amazon (AMZN) has achieved a critical milestone in its major AI initiatives, with its supercomputer named “Rainier” now officially operational. This system is one of the world’s largest AI training computers, powered by nearly 500,000 Trainium2 chips. These chips are distributed across multiple US data centers and efficiently connect tens of thousands of ultra-servers through NeuronLink technology, aiming to minimize communication latency and enhance overall computational efficiency.
The official operation of the “Rainier” system marks a significant turning point for Amazon Web Services (AWS). It not only better meets the market’s growing demand for AI model training and inference but also confirms that Amazon’s large-scale capacity expansion for AI workloads has officially begun. Although this launch was slightly delayed compared to expectations, its successful deployment lays a solid computational foundation for AWS to face future competition. With the aid of this system, Amazon’s AI partner Anthropic will gain more than five times the computing capacity it had before. By the end of 2025, Anthropic’s Claude model is expected to run on over 1 million Trainium2 chips.
Amazon’s continuous investment in self-developed technology is a core pillar of its AI strategy. Both the Trainium series chips and the Nova model are crucial strategic moves designed to give AWS a differentiated advantage in the highly competitive AI market. AWS CEO Matt Garman has emphasized that the performance of these in-house chips can surpass that of general-purpose alternatives. Justin Post, an analyst at Bank of America Securities, also pointed out that this in-house development approach helps reduce model training and inference costs and ultimately improves AWS’s business margins.
This strategy is already showing results. During an earnings call, Amazon revealed that demand for its self-developed AI chip, Trainium, is extremely strong, having grown into a core business worth billions of dollars with 150% quarter-over-quarter growth. Amazon CEO Andy Jassy stated, “The adoption rate of Trainium2 continues to rise, and current capacity is fully booked. This business is scaling rapidly.” He also mentioned that while Trainium processors are currently primarily used by a few large customers, the next-generation Trainium3 is expected to attract a broader customer base upon its release.
Looking ahead, the AI platform Bedrock, built by Amazon, is highly anticipated. Jassy noted that the platform aims to become “the world’s largest inference engine,” and its long-term development potential is comparable to AWS’s core compute service, EC2. He added, “Currently, the vast majority of token usage on Bedrock already runs on Trainium chips.” This fully demonstrates the tight integration between self-developed chips and upper-layer application services. According to the plan, the more powerful Trainium3 chips will be available for customer preview by the end of 2025, with larger-scale deployment expected in 2026.
This series of forward-looking deployments in AI infrastructure is widely viewed by the market as likely to significantly enhance AWS’s revenue-generating capabilities. Analyst Justin Post maintains a “Buy” rating on Amazon and predicts its stock price will reach $272. He forecasts that by 2026, AWS’s year-over-year revenue growth rate will reach 19%, primarily driven by the explosion in AI demand from partners like Anthropic. He estimates that this alone could bring up to $6 billion in incremental revenue to AWS in 2026, boosting the overall growth rate by 4 percentage points. This positive outlook is also directly reflected in the capital markets; previously, thanks to strong performance, Amazon’s US stock price rose over 12% during the trading session.
While vigorously advancing its self-developed technology, Amazon also emphasizes its partnerships with other industry giants. Jassy clearly stated that Amazon continues to maintain close collaborations with NVIDIA (NVDA), Advanced Micro Devices Inc. (AMD), and Intel (INTC) and plans to further expand cooperation with these chip suppliers in the future. He concluded, “We will continue to aggressively invest in expanding computing capacity because we see an explosion in demand. Although we are rapidly increasing capacity, it is absorbed as soon as it becomes available. This is still a very early stage, presenting unprecedented opportunities for AWS customers.”
