Amazon’s custom chips power Meta, giving the cloud giant another way to win in AI
Amazon stock is finally getting the breakout it’s been waiting for — and another intelligence chip made Friday may help explain why. Shares of the e-commerce and cloud giant rose nearly 3% to more than $263 each, en route to a second record high close this week. The latest catalyst: Friday morning’s announcement that Meta Platforms has agreed to use Amazon’s Graviton chips to help run the social networking and AI company’s massive computing needs. The agreement, which will last at least three years, makes Meta the top five customer of Graviton, which are central processing units (CPUs) based on the design of Arm Holdings, our new Club stock. The partnership is important because it illuminates the growing position of Amazon Web Services in one of the most important races in technology: providing the infrastructure behind AI. As hyperscalers seek enough computing capacity to meet their AI ambitions, Amazon is increasingly showing that it can compete as a cloud and chip provider. That’s a key part of Amazon’s bullish case for investors. “People are realizing that everything they said in the (investor’s) letter is actually coming true,” Jim Cramer said Friday during the Investing Club Morning Meeting, referring to Amazon CEO Andy Jassy’s latest shareholder letter published earlier this month. “Almost all AI so far has been done on Nvidia chips, but a new shift has begun,” Jassy said in the letter, noting how companies are looking for alternatives to Nvidia’s graphics processing units (GPUs), such as AWS Graviton and Trainium. The CEO said Amazon’s chips could help lower costs for its AWS cloud customers. While Nvidia GPUs remain popular chips for AI training due to their sophisticated ability to process large amounts of data, the rise of CPUs is driven by cost-effectiveness and shrinking real-world AI applications. Amazon has dedicated years to building a chip portfolio that includes both Graviton CPUs and Trainium accelerators, which act as GPUs. Amazon describes CPUs like the Graviton as being best suited for “always on thinking workloads” that require constant decision making. Trainium and GPUs, on the other hand, excel in training AI models. In his letter, Jassy said, “The annual revenue level in our chips business (including Graviton, Trainium, and Nitro – our EC2 NIC) is now over $20 billion, and growing at three-digit percent YoY.” For Meta, the appeal is clear. Training AI and running day-to-day tasks is expensive, and moving parts of those workloads to Graviton can reduce computational costs. That’s important for a company that deploys AI to billions of users through Facebook, Instagram, and its own ad platform where content and recommendations must continue to evolve at scale. Jim is also always cheap on the Meta. He called the stock “a screaming buy right now because the stock has come to him as a good performer,” referring to Meta CEO Mark Zuckerberg. “He seems to understand when to use agents and when to use people,” Jim added, following Meta’s announcement Thursday of layoffs. In the end, the Meta-Amazon Graviton partnership favors both tech giants in some ways, but it’s a definite win for AWS as it scrambles to capture more of the AI boom by selling the rare chips that power it. We have a buy rate of 1 on Amazon with a price of $250. We also have a rating of 1 on Meta and a price target of $825. Both Amazon and Meta report earnings after the closing bell next Wednesday. Microsoft and Alphabet also deliver their quarterly results that evening. Microsoft owns Azure, the second largest cloud after AWS, and develops custom chips. Alphabet-owned Google Cloud is No. 3 and is building its own custom chip business, with Broadcom as a co-designer, while also using its own AI model, Gemini, to compete with the likes of OpenAI’s ChatGPT. (Jim Cramer’s Charitable Trust is long AMZN, META, NVDA, GOOGL, MSFT, AVGO. See here for a full list of stocks.) As a subscriber to the CNBC Investing Club with Jim Cramer, you will receive a trade alert before Jim makes a trade. Jim waits 45 minutes after sending a trade alert before buying or selling stock in his charity portfolio. When Jim talks about a stock on CNBC TV, he waits 72 hours after issuing a trade warning before making a trade. THE PRIVATE INFORMATION OF THE BURNING CLUB IS SUBJECT TO OUR TERMS AND CONDITIONS AND PRIVACY POLICY, AND OUR PRIVACY POLICY. NO LEGAL LIABILITY OR OBLIGATION EXISTS, OR IS CREATED, BY YOUR ACCEPTANCE OF ANY INFORMATION PROVIDED BY CONTACTING THE INVESTMENT CLUB. NO PARTICULAR RESULT OR INTEREST IS GUARANTEED.
