Sam Yeh/AFP via Getty Images
Nvidia shares dropped as much as 8% on Monday, extending their post-earnings decline to 13% as investor questions linger around the return on investment from huge artificial intelligence spending by companies.
According to JPMorgan's Chairman of Market and Investment Strategy, Michael Cembalest, Nvidia's second-quarter earnings results were extraordinary.
"Phenomenal, and the antithesis of the dot-com era," Cembalest said in a note on Monday. "NVIDIA bears no resemblance to dot-com market leaders like Cisco whose P/E multiple also soared but without earnings to go with it."
But while Nvidia has benefited immensely from technological advancements of its GPUs, there are still lingering questions about how much its customers will financially benefit from the hundreds of billions of dollars they're spending on AI-enabled GPU chips.
If profits don't arrive quickly for the cloud hyperscalers, spending on Nvidia's core product could eventually taper off.
"Within the next two years, corporate AI adoption trends…need to move higher to avoid a 'metaverse' outcome for all the capital deployed," Cembalest said, referencing the chart below.
JPMorgan
A recent report from The Information suggests OpenAI could lose $5 billion this year, the CEO of Anthropic said it will cost $100 billion to train a single AI model in 2027, and the largest tech companies might have $500 billion in "missing revenues" that are needed to break even on their massive AI investments, Cembalest highlighted.
"Barclays estimates that in 2024, enough GPUs were built to generate ~$100 bn of revenues at maximum utilization rates. Actual 2024 payments by GPU users: ~$10 bn. How long will it take for this gap to close?" Cembalest asked.
To be sure, every new computing platform has required massive investments in infrastructure upfront before monetization strategies gained scale.
Yet, the ultimate question is just how patient investors will be — and that patience, or lack thereof, will be directly reflected in Nvidia's share price.
"Every computing cycle entails infrastructure first, then platforms and then applications. It's probably too soon to worry that there's no killer generative AI app yet akin to the enterprise resource planning software of the 1990s or the search and e-commerce applications of the 2000s," Cembalest said.
He added: "But the clock is ticking: NVIDIA data center revenues as a share of market-wide capital spending are projected to reach heights seen at the peak of the mainframe era in 1969 and the dot-com boom. As a result, stakes for investors are high."
For his part, Nvidia CEO Jensen Huang addressed the return on investment question on his company's earnings call last week.
But rather than point to a killer app that is generating a lot of revenue for its customers, Huang highlighted the immense cost savings its customers are experiencing thanks to the immense computing power of its accelerated GPUs compared to traditional CPUs.
"The amount of computing demand continues to grow quite significantly. You could maybe even estimate it to be doubling every single year. And so if we don't have a new approach, computing inflation would be driving up the cost for every company, and it would be driving up the energy consumption of data centers around the world," Huang explained.
He added: "When they build out Hopper-based infrastructure and soon Blackwell-based infrastructure, they start saving money. That's a tremendous return on investment. And the reason why they start saving money is because data processing saves money."
