Nvidia is one obvious beneficiary of artificial intelligence — companies have been buying its chips and investor interest has sent its shares soaring to record highs. In the past year, Nvidia sales have skyrocketed as companies such as Alphabet, Microsoft , Meta , Amazon and OpenAI buy billions of dollars of its graphics processing units, which are advanced and pricey chips required for developing and deploying artificial intelligence applications. Many companies are in the AI infrastructure buildout phase right now. But will their millions — and billions — of dollars of investments pay off? Clare Pleydell-Bouverie, portfolio manager at Liontrust Asset Management, said on CNBC Pro Talks last week that more infrastructure needs to be built for companies that want to provide AI applications such as ChatGPT for their customers. That’s because, in order to enable AI applications, companies have to make the switch from “general purpose computing to accelerated computing,” she said. “You can’t run AI on traditional compute, it would be prohibitively expensive, and far too energy intensive,” said Pleydell-Bouverie. To meet that demand, an estimated 100,000 data centers would need to be built, she said. “This arms race to build out AI infrastructure is really underway,” she said, adding that Amazon, Google, Meta and Apple combined will spend $200 billion on capital expenditure this year alone. That’s a 35% increase from last year, she said, and all this incremental investment is being directed to AI initiatives. “The sort of natural pushback here is — is this sensational investment worth it?” Pleydell-Bouverie said. But, she said, the “ROI [return on investment] on accelerated compute is actually very compelling if you’re a cloud service provider – so AWS [Amazon Web Services] or Microsoft is here.” “And if you’re spending $1 on Nvidia’s accelerated computer architecture, this translates to $5 in instant revenue over a four-year time horizon,” she said. Those capital expenditure increases will be funded by productivity gains, she said. Meta is one example, with 50% of the content seen on Instagram being generated by AI, she said. “So you can actually assume quite a decent return on that,” Pleydell-Bouverie said. And the world is “only in the first five minutes of this AI infrastructure buildout,” she added. However, there’s a “lower-than-average return” on certain investments such as Meta’s generative AI assistant and its Llama AI models as they require “a lot of investment,” according to her. However, she said it would pay off in the long run. “When we’re thinking about the long term, no other company within Meta’s competitive landscape probably has the capability to develop these AI clusters,” she said. Meta’s Llama 3 models are built on computer clusters using Nvidia’s 24,576 GPUs — and that has allowed Meta to make a “phenomenal breakthrough,” said Pleydell-Bouverie. “And so we are already starting to see the fruits of the labor coming through for these tech giants,” she said. “But there are so many other companies that are benefiting from immediate productivity gains as a result of these cloud service providers providing access to AI.” Companies using AI for productivity gains range from banks such as JPMorgan, which uses an AI-powered cash management tool, to consumer beauty brands like L’Oréal which are going into beauty tech, according to her. Pleydell-Bouverie co-manages Lionstrust’s global tech, innovation and dividend funds. For the year to March, all three funds have beaten their benchmark indexes, with the Liontrust Global Technology Fund rising 51.9%, higher than the MSCI World Information Technology Index’s 39.1%.
