Analysts say AI infrastructure capex fears are overdone for Alphabet and Amazon

Analyst Nick Jones of BNP Paribas was recently out with a bullish note on Alphabet (GOOGL, GOOG) and Amazon (+2%), saying that fears over artificial intelligence (AI) infrastructure spending are 'overdone.' Data center spending has become a hot-button issue to start the year, as some investors have questioned the economics of this spending. Capex climb raises investor eyebrows However, this didn't keep Alphabet and Amazon from cranking up their capex plans for the year. In February, Alphabet said its capex would rise to between $175 billion and $185 billion this year, up from $91.4 billion in 2025. Amazon turned around and upped the ante a few days later, saying it would boost its capex to $200 billion from $131.8 billion last year. The announcements didn't sit well with investors, who sent the shares of both stocks lower. The custom chip advantage In my view, if there are any two companies that should be ramping up their AI infrastructure capex, it is Alphabet and Amazon. The reason for this is twofold. The first is that both companies have developed their own AI ASICs (application-specific integrated circuits). These are custom chips developed to handle specific AI-related tasks. Because ASICs are purpose-built, they tend to be more energy efficient, and the total cost of ownership tends to be lower. This gives both companies a structural cost advantage. Alphabet is the leader in this space with its tensor processing units (TPUs), which it developed more than a decade ago and has tightly integrated with its entire hardware and software ecosystem. Amazon, meanwhile, has developed more budget-oriented chips with its Trainium and Inferentia offerings. Both companies have been seeing momentum with customers deploying their chips within their data centers. Anthropic has placed a $21 billion TPU with Alphabet partner Broadcom for this year and recently extended its partnership for additional capacity to both be deployed through Google Cloud and supplied directly by Broadcom. Meanwhile, Amazon also opened a large data center exclusively for Anthropic at the end of last year, powered by its Trainium chips. In addition to the cost advantage of having its own AI chips, the other reason why both Alphabet and Amazon should be pressing their advantages and investing heavily in AI infrastructure is that both companies also have heavy internal usage. They have both developed large language models (LLMs), and both are using AI to help drive growth in other parts of their businesses, such as search for Alphabet and e-commerce with Amazon. This ability to internally absorb capacity lessens the risk of overbuilding. With both stocks off their highs, now is a great time to buy these AI giants. In fact, they are two of my favorite stocks at the moment, and I own both.