Amazon: The Invisible AI Giant?

When it comes to AI giants, investors often first think of hardware companies like NVIDIA, AMD, and Broadcom, or software companies like OpenAI. However, Amazon is one of the few companies that fully exploits every aspect of AI. Its AI product layout includes AI chips, AI cloud computing, and highly competitive large models.

● AI Application:

Monetizing AI directly to consumers (C) is more challenging than to businesses (B). AI products aimed at consumers are still predominantly free, as seen with over 95% of ChatGPT users utilizing the free version. However, businesses are willing to pay directly for B2B AI products because many companies are concerned about falling behind in the AI wave and have found that using AI can genuinely enhance efficiency and reduce costs.

NVIDIA especially values Amazon's leading position in cloud technology. Jensen Huang even claimed at Amazon's re:Invent conference in December that their Blackwell chip would first be supplied to Amazon. This is due to Amazon's advantage in cloud applications on the B2B side.

● AI Model:

Amazon's large model Nova outperforms GPT-4 and ranks among the top three in the industry, only behind Anthropic's Claude model and Google's large model Gemini. However, Google's market share in cloud computing is much lower than Amazon's, and its cloud revenue as a percentage of total revenue is less than Amazon's, making AI's positive impact on Amazon's earnings even more apparent.

Additionally, Amazon's Nova model is more cost-effective, which is priced up to 75% cheaper than other leading models in similar categories.

In 2024, Amazon also launched the AI application development platform Bedrock, which not only helps businesses reduce costs but also enhances business automation efficiency, leading to widespread adoption. Amazon Bedrock acts like a "supermarket," offering both its own models and third-party models, including over 100 professional or industry models from companies like IBM and NVIDIA.

Amazon explained that Bedrock gives developers the freedom to choose. Having multiple options when using models is crucial, as "there will never be one tool that dominates a field completely."

● AI Hardware:

Launched in December 2024, Amazon boasts that the Trainium 2 costs about 30% less than its competitors. Compared to its predecessor, its performance has quadrupled, and its memory capacity has tripled.

The Trainium 2 chip has begun deployment in Amazon's data centers and is expected to be extensively promoted in several core data centers, including those in Ohio, soon. The company aims to cluster up to 100,000 Trainium 2 chips by 2025, with the Trainium 3 chip set to further expand its performance cost advantage.

Amazon's chip business is led by James Hamilton, a pioneer in cloud computing. Hamilton's team proposed the idea of developing chips independently as early as 2013. Amazon's first AI chip, Inferentia, launched in 2019, focused on inference tasks, while the Trainium series is primarily aimed at training machine learning models.

Besides internal use, Amazon's AI chips are also being sold externally. Prominent companies like data analytics firm Databricks and Claude's parent company, Anthropic, have begun using Amazon's Trainium chips and achieved initial success in some projects. Tom Brown, Chief Computing Officer at Anthropic, commented, "We are impressed by the cost-effectiveness of the Amazon Trainium chips. We have been steadily expanding their application across increasingly broad workloads."

Benoit Dupin, Director of Machine Learning and Artificial Intelligence at Apple, revealed at the AWS re:Invent conference that Apple has adopted Amazon Web Services' custom Trainium and Graviton AI chips in its search services. Dupin emphasized that Amazon's AI chips are "reliable, efficient, and capable of providing high-quality services to global Apple customers," highlighting a "solid partnership" between Apple, AWS, and Amazon.

Over the years, Apple has extensively utilized AWS to support services like Siri, Apple Maps, and Apple Music. Furthermore, due to direct competition between Apple's iOS and Google's Android as well as Microsoft's Windows operating systems, Apple is unlikely to support Google and Microsoft's cloud computing, making Amazon a preferred choice in AI hardware and software for Apple.

Notably, Amazon has also announced a rental service for the Trainium 2 chip, which will further expand Amazon's influence in the AI chip field.

● Amazon e-commerce's underrated demand for AI

Beyond cloud computing, Amazon's AI can also be applied in its traditional e-commerce domain, providing value in inventory management, logistics optimization, seller customer service, and robotic warehouse delivery. With e-commerce still having significant room for growth in the U.S., as indicated by higher growth rates in online retail compared to offline during Black Friday sales, the further development of e-commerce will continuously provide Amazon with a steady cash flow.

Recently, several major banks have raised their target prices for Amazon. On December 18, J.P. Morgan raised Amazon's price target from $250 to $280, maintaining an "Overweight" rating; Bernstein raised its price target from $235 to $265. On January 3, Wolfe Research increased its price target for Amazon from $250 to $270, maintaining a "Buy" rating.

Overall, considering earnings growth and PE levels, Amazon's valuation is relatively reasonable compared to other major U.S. stocks, and its business segment is more easily monetized through subscription models, making Amazon an undeniable giant in the AI era.

Still, Morgan Stanley previously noted that Amazon needs to achieve at least a 15% revenue growth and over 20% profit growth in 2025 to sustain its current valuation level.

Looking ahead to 2025, whether Amazon's stock price can continue its growth momentum and set new records, whether the application of AI in retail and logistics can bring efficiency improvements and cost reductions, and whether investments in AI and cloud technology can translate into significant returns in the future are all points of interest for investors.