There’s a lot of talk about the current phase of artificial intelligence: infrastructure building. This has clear benefits for chipmakers like Nvidia and hyperscalers like Amazon, Microsoft and Alphabet. Hyperscale enterprises provide the vast amounts of cloud computing power needed for AI applications, and analysts predict that demand for more data centers will grow as data centers accommodate the massive amounts of computing power required for AI workloads. But technology analysts say the next bottleneck in artificial intelligence infrastructure — and the one worth investing in — is the Internet. A network in normal technical terms refers to a network of devices that can transmit and share information through physical or wireless communications. However, in the field of artificial intelligence, the requirements are higher as large language models and other artificial intelligence applications require very high bandwidth and low latency. Morningstar analysts said in a June 2024 report: “While Nvidia and its graphics processing units capture most of the headlines for generating artificial intelligence, we believe that the network is the hardware that supports models and applications such as ChatGPT. “So far, the focus has been on (graphics processing units), real AI chips. Of course, those are the most important pieces of the puzzle. But networking is where we see the next bottleneck emerging. place,” Pleydell-Bouverie, a portfolio manager at Liontrust Asset Management in Clare, told CNBC Pro Talks in May. That’s because the “larger systems” coming to market, such as Nvidia’s rack-scale systems, require “much more” infrastructure content, such as networking, she said. Morningstar technology equity analyst William D. Kerwin and technology equity strategist Brian Colello said they believe the demand for fast networks from generative artificial intelligence will directly translate into “strong, long-term growth for well-positioned network vendors.” . The company said increased investment in generative AI model training and inference will drive 34% growth in AI network spending over the next five years. That means spending will reach $34 billion in 2028, up from Morningstar’s 2023 forecast of $8 billion. Morningstar analysts said: “Internet creates performance bottlenecks for the development of generative AI models.” They added: “Well-positioned Internet companies are excellent second-order derivatives for investing in generative AI. “Most of the generative AI spending will go to GPUs, but networking is the critical infrastructure to enable GPU performance.” Stocks that capture this trend Marvell Technology is Morningstar’s top pick on the AI generative networking trend. The company said it is currently “undervalued” and offers investors an “immediate opportunity” to capitalize on growing investments in generative artificial intelligence networks. Morningstar said other major winners from this networking trend include Arista Networks, Nvidia and Broadcom. However, its analysts believe the generative AI opportunity in these three stocks is “largely priced in” as their share prices have experienced “strong” appreciation. “However, patient investors can wait for a pullback as long-term fundamental opportunities are strong,” Morningstar said. It added that it is bullish on the adoption of Ethereum in generative artificial intelligence networks, referring to a type of network standard. According to Morningstar, Arista will be a major beneficiary of the transition to Ethernet. The currently commonly used technology is InfiniBand. “There are very few players that can actually provide this infrastructure,” Pleydell-Bouverie added, referring to network infrastructure. She named Meta and Broadcom as stocks riding the trend. Broadcom is a “leader” in networking chips and will benefit as Ethernet emerges, she added. “Ethernet is becoming the de facto standard for scaling these artificial intelligence workloads. Broadcom has best-in-class chips to power Ethernet,” Pleydell-Bouverie said.
