The next move for Amazon’s cloud leadership, and perhaps its stock price, will depend on whether growth in that side of the business continues to reaccelerate. Custom AI chips are key to that effort. Investors widely believe that Amazon’s stock price, which also dominates e-commerce, is driven primarily by the fortunes of its Amazon Web Services cloud division, which generates most of the company’s operating profits. That’s why AWS performance is the most important factor for Amazon’s market value, and despite last year’s worst Magicent Seven, its market value remains close to $2.54 trillion. And AWS will be the focus of Wall Street’s attention Thursday night, when Amazon reports its fourth-quarter 2025 results and looks ahead to this year’s forecast. A big part of Amazon’s strategy to further accelerate AWS growth is its own artificial intelligence custom Trainium chips. The latest version, Trainium3, launched in late 2020 and was announced at the AWS Re:Invent cloud conference in December. Trainium is designed in-house rather than relying solely on third parties. Amazon isn’t alone. Last week, Microsoft announced its next-generation custom AI chip, the Maia 200, nearly two years after its first version. Google has a custom AI chip called the Tensor Processing Unit, and since the launch of Gemini 3 in November, it has received a lot of attention as a winning AI model. Gemini 3 was trained by a TPU co-designed by Broadcom. By building its own chips, Amazon, like its cloud competitors, aims to achieve two key benefits when running AI workloads: lower computing costs for customers and lower power usage in data centers. Both are more important than ever as AI models grow larger and data centers consume record levels of power. According to AWS Vice President David Brown, price performance, or how much compute customers get for every dollar they spend, is now a key metric that customers care about. “If they can find chips and processors that can deliver more performance for less dollars, that’s a very strategic advantage for their business,” Brown said in an interview with CNBC. This “strategic advantage” goes both ways for Amazon, as Microsoft’s Azure cloud and Google Cloud (the second and third largest clouds after AWS, respectively) are experiencing relatively strong growth. The law of large numbers is certainly at play here, with estimated AWS revenue growth for calendar year 2025 at 19.1% to $177.78 billion, compared to Azure’s 26.1% growth at $120.85 billion and Google Cloud’s 35.8% growth at $58.7 billion, according to FactSet. That’s $10 million. But there’s no denying that Microsoft and Google are gunning for AWS, which they don’t want to cede to the throne anytime soon. Growth and balanced spending guidance will also be noted from Amazon, as Alphabet on Wednesday night reported a great quarter and a huge capex guide. Microsoft’s earnings report last week also showed significant capital spending in AI. Optimism about AWS’s re-acceleration continued with strong third-quarter revenue growth of 20.2% year over year, beating expectations of 18.1% and marking a return to above 20% growth for the first time since 2022. Amazon shares soared 9.6% during trading hours after third-quarter numbers were released late on Oct. 30, and rose another 4% on Nov. 3 to hit a recent high close of $254. Shares fell from there, down more than 8.5% as of Wednesday afternoon trading. AMZN MSFT, GOOGL 5Y Mountain Amazon, Microsoft, Alphabet: 5-Year Stock Performance AWS Bets on Custom Chips AWS’ push into custom chips dates back to its 2015 acquisition of then-startup Annapurna Labs, which became the foundation of its internal silicon efforts. Since then, AWS has built major hardware platforms including Graviton CPUs (central processing units) and AI accelerators such as Trainium and Inferentia (close to Nvidia GPUs). All of these are specifically designed for cloud and AI workloads. The custom chips, also known as application-specific integrated circuits (ASICs), aim to reduce hyperscalers’ dependence on Nvidia hardware, which has become the gold standard for general-purpose AI. Nvidia’s GPUs (graphics processing units) are in short supply and expensive. For Nvidia’s part, CEO Jensen Huang recently said he’s not worried about custom chips. Shortly after Google’s Gemini 3 launched in November, Jensen told Jim Cramer, “What Nvidia is doing is much more versatile. … Nvidia can address a much broader market than just chatbots.” Nevertheless, Brad Gastwirth, global head of research and market intelligence at Circular Technology, said Trainium’s cost advantage still makes sense in a market dominated by Nvidia’s premium chips. “NVIDIA is charging an astronomical amount of money for silicon,” Gastwirth said. He sold his previous company to Wedbush, where he served as chief technology strategist for several years, before going out on his own again. Custom chips are much cheaper, allowing AWS to offer lower prices to customers. “That’s a big advantage,” he added. Gastwirth explained how Trainium’s specialization allows it to run certain AI models more efficiently and at lower cost. “You can run a model exactly for what you want it to do. … If you build something specific to your needs, you can save a ton of money instead of buying a GPU that’s far more powerful than your needs.” For large companies like Amazon, Gastworth says, the economics improve over time. Custom chips require an upfront investment, but are cheaper when operated on a large scale. There are very few things you can do on a GPU that you can’t do with something like a Trainium accelerator. AWS Vice President David Brown AWS’s Brown emphasized this point, including that the performance gap between GPUs and custom accelerators is closing. “When it comes to AI, there are very few things you can do with a GPU that something like a Trainium accelerator can’t do,” Brown said, pointing to the increasing ability of specialized chips to handle the large-scale training and inference tasks that customers are interested in. Customers including AI startup Anthropic, which developed the chatbot Claude, aim to reduce training and inference costs by up to 50% using the latest version. This not only improves AWS utilization, but also improves economics for Anthropic, which appears to be showing in the company’s projections. The Information reported last week that Anthropic raised its internal revenue forecast for 2026 to at least $17 billion, up from a previous estimate of $15 billion. For 2027, the AI startup’s revenue is estimated at $46 billion, up from the previous estimate of $39 billion. As AWS’s largest cloud partner, Anthropic’s cost-effective growth should help re-accelerate AWS’s growth. What’s on the Street In an interview with CNBC, Roth MKM analyst Rohit Kulkarni said AWS is “positioned to do very well in this new world of AI clouds” as more customers demand AWS’ in-house silicon. He cited Amazon’s published multibillion-dollar profit margins for its Trainium chip line, with more than 1 million chips produced and more than 100,000 customers. Roth MKM also wrote in a note last month that the Trainium chip is “focused on categories of computing where there is arguably a greater market opportunity.” In response, analysts who rate Amazon stock as a buy raised their price target to $295 per share from $270. He said he expects cloud revenue to be positive if it exceeds 22% in the next quarter. The FactSet consensus forecast calls for 21.7% growth in the fourth quarter. This is an increase quarter-over-quarter and significantly higher than the 18.9% in the same period last year. Along the same lines, Mizuho said that the availability of Trainium “should drive AI revenue growth,” adding that the expansion of inference workloads for which the Trainium chip is designed “should further expand AWS’s revenue base.” Mizuho analysts predict that AWS revenue growth will accelerate by 23% in 2026. They cite the resurgence of AWS as a “key driver” of why Amazon stock is positioned for multiple expansions this year. Mizuho rates Amazon stock a “buy” and has a price target of $285. To be sure, AWS’ custom silicon strategy is still in its infancy, Baird said, and there will be some bumps in the road. The company said in a memo late last month that AWS’s focus on its own chips “could deliver long-term benefits through higher margins and reduced dependence on third parties,” but noted that scaling the strategy across customers and workloads would take time. Baird analysts cited early frictions surfacing as customers diversified their workloads, citing “growing pains” such as the “forced adoption” of AWS’s proprietary stack, and Anthropic also putting some workloads on Alphabet’s Google Cloud. Baird also said that AWS is “increasing orders for off-the-shelf chips,” including Nvidia’s Blackwell, to alleviate near-term capacity bottlenecks associated with increased production of custom silicon. Baird analysts rate Amazon stock as Outperform and have a price target of $285. Conclusion AWS’s quarterly results could drive the stock price in the near term, as investors watch to see if AWS’s re-acceleration story continues. While the cloud giant has the ability to deliver the growth investors expect, Jim Cramer said at a Tuesday morning meeting that “we are not sponsoring Amazon at this time.” He added that the cloud and e-commerce giant could fall into the “software sector.” The sector has been hit by concerns that AI will disrupt business models for enterprise software. Ahead of Thursday night’s earnings report, the club has set a price target for Amazon at $275 per share. We rate the stock a ‘1’, the equivalent of a ‘buy’, but Jim’s wait-and-see attitude is just good discipline as printing is imminent. (Jim Cramer’s charitable trust is long on AMZN, NVDA, and GOOGL. See here for a complete list of stocks.) As a subscriber to Jim Cramer’s CNBC Investment Club, you will receive trade alerts before Jim makes a trade. After Jim sends a trade alert, he waits 45 minutes before buying or selling stocks in his charitable trust’s portfolio. If Jim talks about a stock on CNBC TV, he will issue a trade alert and then wait 72 hours before executing the trade. The above investment club information is subject to our Terms of Use and Privacy Policy, along with our disclaimer. No fiduciary duties or obligations exist or arise from your receipt of information provided in connection with the Investment Club. No specific results or benefits are guaranteed.
