Challenge NVIDIA's dominance! Amazon plans to sell self-developed AI chips to external data centers, and the stock price responds with a rise of over 2%

According to Zhitong Finance APP, Amazon (AMZN.US) is considering selling its self-developed artificial intelligence (AI) chips for use in data centers of other companies, marking an important step for the company in challenging NVIDIA's (NVDA.US) dominance in the AI chip market.

Peter DeSantis, an executive responsible for Amazon's AI business, stated in an interview that the company has begun discussions with potential customers but did not disclose specific partners. "We believe AI infrastructure is evolving rapidly," DeSantis said, "and we are always looking for ways to serve more customers."

Boosted by the news, Amazon's stock price rose over 2% during trading on Thursday, reaching $242.86.

Amazon launched its self-developed AI acceleration chip, Trainium, in 2020, aiming to provide a more cost-effective computing solution for AI model training. Currently, companies such as OpenAI, Anthropic, and Uber (UBER.US) are using Trainium chips through Amazon Web Services (AWS). Amazon previously disclosed that as of April this year, Trainium-related projects had accumulated over $225 billion in revenue commitments.

In fact, Amazon has long signaled its intention to expand the commercialization of Trainium. In April this year, CEO Andy Jassy stated in a letter to shareholders that it is "entirely possible" to sell complete racks equipped with Trainium chips directly to third-party customers. The market generally views this as an important move for Amazon to strengthen its AI strategy and narrow the gap with competitors.

With ChatGPT igniting a global AI boom, cloud computing giants such as Amazon, Google (GOOGL.US), and Microsoft (MSFT.US) are accelerating the development of their own AI chips to reduce dependence on NVIDIA GPUs.

At the same time, the rapid development of the AI industry has also given rise to numerous cloud service providers focused on artificial intelligence, driving increased demand for "sovereign AI" infrastructure in regions such as Europe. "Sovereign AI" refers to data storage and computing resources deployed within a country's borders to meet local legal and data security requirements.

In April this year, Google CEO Sundar Pichai also announced that the company would open its self-developed Tensor Processing Units (TPUs) to certain customers, allowing them to deploy related hardware in their own data centers. Analysts believe that Amazon's plan to sell Trainium chips to external customers indicates that competition among cloud computing giants in the AI chip market is further escalating.

DeSantis noted that part of the reason Amazon is pushing Trainium into markets outside of AWS is the growing demand for locally controllable computing resources, especially in overseas markets like Europe.

However, regarding calls from some European countries to reduce dependence on American tech companies, DeSantis stated that AWS's business has not yet been substantively affected.

He also revealed that the third-generation Trainium chips, which began shipping this year, are currently "basically sold out," and the fourth-generation Trainium, expected to launch next year, has already garnered significant customer interest Regarding external concerns that selling Trainium outside of AWS may erode cloud business revenue, DeSantis disagrees. "There is still a huge computing power gap in the AI field," he said, "I am not worried about this issue."

In addition to Trainium, Amazon has also been actively promoting its general-purpose processor Graviton in recent years. DeSantis revealed that Meta (META.US) has begun adopting Graviton chips, and in the past three years, the number of Graviton chips deployed by Amazon has exceeded that of any other type of processor in the company